I'm building my own cloud (I actually typed claude instead of cloud there... wow). There's no IaaS or PaaS; it's much simpler. I wanted my own way of connecting to machines and the TCP services on those machines without having to install Tailscale (not allowed on a locked-down corporate PC) or pay for Azure or AWS or GCP or even Hetzner or Linode. I've got 10gbps fibre and a huge workstation at home, and I've got lots of laptops and VMs and other outboard stuff that I want to work in concert with that workstation, so I started building something I call Tela (Filipino for fabric; I was sitting in Ninoy Aquino International Airport waiting for a flight when I had the idea, and it's implemented as a network fabric).
I am not sure WireGuard existed at the time, and I used SoftEther and based it all on doing outbound tunnels to TCP/443* to avoid firewall blocks in corporate networks.
Thanks! Yes, Tela already does UDP hole-punching. I made Tela because I wasn't allowed to install Tailscale on my new corporate laptop, and no other available solution seemed to tick the right boxes. It started as a simple way to RDP to my home workstation, but then I realised that if I could do that, I could finally pull my ad-hoc home cloud into one tool. The hub model is very much by design, for organisational purposes. The hole-punching feature gives me the P2P speed (and even STUN, if available). An upcoming version will allow hub-to-hub topologies.
It should have occurred to me that tela is also Spanish, since about every third word I hear in a Tagalog sentence seems to be of Spanish origin.
Does tela create an L3 network? if that's the case, what do you do to avoid IP addressing clashes? In Wormhole I decided by default to use CGNAT addressing (100.64.0.0/10)
I did not go too far unfortunately, so I did not face problems such as discoverability (do you have to know/remember all the IP addresses from the devices connected? DNS? etc).
I am building a guitar learning platform where the main feature is an RPG game called Guitar Quest. You plug in your real guitar into an audio interface and the game listens to what you play based on whats shown on the screen. Play the right note, the monster takes damage.
The goal is: fretboard memorization, scales, music theory, chord construction, and performance. There's also a tuner, practice tracking, backing tracks you can jam over, and a community for sharing your gear collection.
No AI anywhere in the stack. Just a guitar, a microphone, and some monsters.
Working on an audio streaming platform powering an indy internet radio. Looks like Icecast & friends show its age and a similar product can be easily built with the functionality cast down to simply robust streaming & handling "timed playlists". I enjoy every bit of knowing exactly what happens in the code. It's not open source atm, but will be. It's in Go, is a pleasure to write and the deployment takes minimal amount of resources.
Other project is to continue a bit stalled progress of a configuration language BCL - add functions, more structures and fix some hidden scoping issues. Making languages is an endless fun.
https://github.com/wkhere/bcl
Was recently looking for such a streaming thing; just streaming from a set of MP3s in a folder, nothing fancy. The majority looked too complex, with too many moving parts, for my idea.
Found one in Go that might interest you too: GoFM. Although I dropped my idea for now, I'd love to see yours come to life, too.
And then, modified a lot. At some moment I will open it back. (Author's MIT license allowed closing it; I did it actually because I embedded a number of idiosyncrasies related to the radio service that shouldn't be disclosed; but with some amount of work it can be divided into an open and closed part).
The broadcasting skeleton from that original blog/github project is good, though! It might work for your case.
Please keep in mind it's better to stream AAC than MP3. Basically any format you'd probably want to use can be converted to AAC with ffmpeg.
AAC has a simple frame format and it's easy to decipher it; I use it to always send full frames, even when one would want to skip to the next song - by doing that the client behaves more smoothly.
The idea came after I finished a permanent piece for a museum using MaxMsp and python. I always had this thought in the back of my mind that "I could express this so much easier in a few lines of code.."
I'm building an alpaca breeding and show management SaaS which is the first one to combine animal register with a strong approval workflow (think like a pull request), herd management and show management.
No Ai involved but might be something I'll put on the roadmanp to help with breeding decisions. (alpacuna.com)
My friend and I are building Strandfall, a highly physical outdoor larp (live action role playing game) that uses custom spatial computers: https://strandfall.com
Players are survivors of a global disaster that has unleashed mysterious, deadly storms. For three hours, they investigate the origin of the storms and make fateful decisions about their future as individuals and as a community.
We received Immersive Arts funding, which means we can run it in Edinburgh later this year. Here's an excerpt from our 2025 grant application about exactly what those are:
--
Our “storm sensors” are novel spatial computers designed for outdoor usage over long distances. They will house ePaper displays, LoRa (long range) radios, accelerometers, gyroscopes, and GPS chips in a 3D-printed enclosure to provide a low-tech way to augment the reality of the park. These computers will be cheaper, more rugged, longer-lived, and more capable than smartphones, deployable to locations with zero cellular service and no battery charging options.
The sensors will be mounted on top of camera tripods for deployment. Runners will carry them through the park, then position and aim them in the correct direction, as co-ordinated by “operators” using walkie talkies. This will let players feel like they are really setting up important equipment, scanning historical sites for clues (like surveyors), and establishing laser communication links. Lacking colourful touchscreens, the sensors will be less distracting for runners, helping them focus on their surroundings. Essentially, they are a highly tactile and deeply realistic way of immersing players in a post-apocalyptic setting, since such devices – not smartphones – are the most likely to be used.
this is totally not going to cause spontaneous reports to authorities about suspicious _insert anything here_ and ultimately cost taxpayers more to respond to situations that otherwise wouldn't be happening.
It might, if we weren't planning to put prominent stickers on the tripods making clear they're part of a larp, and also getting permission from the park and council and consulting with local parkgoers.
I'm working on ghidra-delinker-extension [1], which is a relocatable object file exporter for Ghidra.
The algorithms needed to slice up a Ghidra database into relocatable sections, and especially to recover relocations through analysis are really tricky to get right. My MIPS analyzer in particular is an eldritch horror due to several factors combining into a huge mess (branch delay slots, split HI16/LO16 relocations, code flow analysis, register graph dependency...).
The entire endeavor requires an unusual level of exacting precision to work and will produce some really exotic undefined behavior when it fails, but when it works you feel like a mechanic in a Mad Max universe, stripping programs for parts and building unholy chimeras from them, some examples I've linked in the readme. It has also led to a poster presentation to the SURE workshop at ACM CCS 2025 in Taiwan as a hobbyist, an absolutely insane story.
Mad respect. I tried extracting a clean .o file out of a statically linked ELF once, and it's an absolute nightmare. How are you handling switch tables and indirect jumps? Without dynamic analysis, it's sometimes physically impossible to figure out what a register is actually pointing to
I have analyzers that resynthesize relocations from the contents of the Ghidra database, no custom annotations required. They evaluate relocation candidate spots through primary references and pointers/instructions and emit warnings if the math doesn't check out.
It does require a reasonably accurate Ghidra database to work properly, but I've had users delink megabytes of code and data from a program successfully (as in, relinking it at a different address results in a functionally identical executable) once they've cleaned it up. The accuracy warning in the readme is mostly because it's really complicated to describe exactly what inaccuracies you can get away with, there's a fair amount of wiggle room in reality as long as you know what you're doing.
I'm building OSS structural biology libraries in Rust (+ CUDA), and a GUI application that ties them together. It is kind of an integration of functionality traditionally in separate tools: Viewing, editing, therapeutic properties + pharmacokinetics, molecular dynamics etc.
I'm building an E2EE encrypted Journaling app with option to upload photos and voice recording (that too is encrypted)
The files and content are encrypted on the device and then gets to sent on server and stored encrypted and when you load the app it fetches E content and decrypts on device, with E Key (which is always stored on device, if you lose the key, you lose all data cause we don't store it)
A bunch of square panels with a grid pattern to mount hold on, with the panels hanging on french cleats (with a locking system, #TODO) so the panels are easily removable so I can hang something like planters on the wall as well with the same french cleats.
No AI, a bit of computers to draw things out in CAD, but otherwise just manual building stuff.
https://imgur.com/a/2TIye34 not much to see yet: square panels with a lot of M10 holes and edges (also the edges of the holes...) sealed. I just got the holds, bolts etc the other day. I sanded the panels & primed them for paint last night.
The larger panel is going to be a volume with a bit of a hang. Panels are all tough 18mm outdoor rated 'betonplex' with phenol? resin or the surfaces.
I'm working on the next update of Omiword[1], an ongoing daily word game previously discussed on HN[2]. I'm building an alternate stand-alone app version with access to all of the archived puzzles. It's slow going since it's just one of many side-projects I like to work on, but that's the tinkerer's dilemma.
Yes, I know that [insert LLM here] could do a lot of that conversion for me in mere minutes. No thank you. I'm doing it, in part, for the doing.
I'm still working on Nottawa (https://nottawa.app) - free live audioreactive visuals for everyone.
It's a node based workstation for generating live visuals using shaders, webcams, local files, and more. 100+ effects and source types.
Nottawa's aimed somewhere between the Resolume and TouchDesigner market. VJs and musicians who are interested in live visuals but don't have the time or funds to dive into a pro setup.
It's built in C++ with OpenFrameworks, ImGui, BTrack, and more. Right now I'm doing a front-end rewrite in SwiftUI which bridges back to C++ while keeping all the rendering in OpenGL.
I’ve talked about it many times here, but I’m slowly building golfcourse.wiki, a golf wiki, because the golf world focuses all it’s attention on “the best” golf course, and the vast, vast, vast majority of interesting courses get zero attention.
I thought it was a website that should exist during the pandemic, so I built it. It’s been slowly growing ever since, because most of the time you really only need to add the info once.
JDBC does not allow pipelining (a Postgres only feature).
It can reduce the number of db round-trips a lot, especially when using Supabase+RLS (or other systems that require frequent setting of configuration values that are basically fire-and-forget).
Meet Bpdbi, a library with first-class pipelining, which provides a Postgres db driver (that's binary only, as the legacy text-based protocol is no longer needed, it just takes up space) and exposes an API that's more close to Jdbi's that to JDBC's (developer friendly).
So great to see how much stuff there is going on outside of AI!
I've been working (on and off for over a decade) on a way to manage my unwieldy photo and video collection. I experimented with so many tools before but ultimately nothing ticked all my boxes so I wrote my own. Admittedly lots of niche stuff like proper support for UTC time, stereo photos and partial dates and things like 100% offline face detection and custom tag hierarchies and querying without a database. Each time I add a new feature I'm having a blast :)
I'm building a good intermittent fasting app for iOS using SwiftUI and native Apple APIs for the most "Apple" experience possible. The app is 100% free and without ads.
I love Elixir but I had been a JS developer by trade. Bridging that interface keeps my brain focusing on building instead of splitting between JS and Elixir.
Still a lot to do, but I love the progress so far. This puts the joy of building something back to me.
I recently started with Paperless-ngx and wanted to also include archive serial numbers (ASNs) for all paper documents using small label stickers, so I built a small tool to create and print ASN label sheets. It's free, no sign up, no ads and just runs in the browser:
Question, do you "build LEGO" as in "make what's on the box it came in", or do you use LEGO to build things yourself and not following instructions?
(I am old, but as a kid, I remember building what the kit was supposed to build once or _maybe_ twice, but using the parts from the kit and various other kits I had to make things purely out of my head pretty mush all the time. I _think_ that's not how people "play with LEGO" these days?)
Both, I enjoy the process of building, so following instructions. But then it's done and boring. I.e. LEGO supercars sets are nice, they're so nice until you marriage chassis with drivetrain - you can play with all the gears and see them, after the marriage it's just boring. So now I'm working more towards reusing parts and building things of my own.
Edit: LEGO Architecture Set (not sure if they still make it) is super nice for building random stuff - it's a set of pure white bricks, super fun.
I just built a Google Chrome chess extension. It is a new tab page full of chess puzzles. It has various difficulties, and a speed run game (solve as many puzzles as you can in 60 seconds). I personally think it's the coolest browser extension I have ever seen (being objective as possible).
It works on Chrome and Brave for now. Planning a Firefox release soon.
If you're a chess fan, you'll probably love it, but if not, let me know why (I'll try to improve it).
I'm building my woodworking workshop. It's gonna take a long long time for AI to invade this area if ever. It's very therapeutic and works as an escape from all the AI craze.
I’m building websites designed to feel as hand-coded as possible! A random assortment of games, digital instruments, off-kilter tools, technical experiments, and more. Complete with a homepage that’s literally handmade ~ featuring tons of hand drawn GIFs! It’s a good place to waste some time :)
Been tinkering on this Olmsted-inspired map of all the trees in NYC for a long time - just added some more features (seasonal colors and better search/filtering) recently.
I'm building a flash card learning tool. I was frustrated with the slowness and complexity of other solutions. Animations and loaders with tools like Duolingo, and unnecessary extra steps to move onto the next card in tools like Anki. If you give it a try I'd love to hear your feedback.
I’m making a digital ESP32 powered synthesizer at https://subalpinecircuits.com/. It’s been a ton of fun, learning so much about every part of the process (and I don’t use LLMs). Currently I’m learning FreeCAD and figuring out what my case design will be. Woodworking, CAM and DFM is a whole other world for a software guy!
still working on https://stringscales.com - fun sideproject to visualize guitar scales on a configurable fretboard, with interactive note highlighting to a backingtrack.
The backingtrack is what I'm actively improving right now. It's just a pad running now, but it will turn into a full track with bass/drums/piano/... and will feature a comprehensive chords based editor so you can add and save your own progressions with a logged in account.
that would certainly be a very interesting thing to add, but in the current iteration that is not possible.
The biggest issue here is that there's not really "one" microtonal system out there. The entire fretboard of a conventional guitar is mapped to work in 12tet - and the libraries I'm using to do all the musical operations also only supports 12tet.
to accommodate all the microtonal temperaments out there would be a pretty daunting task. But I'm not saying never!
Built a SQL interface for DynamoDB. I was tired of constantly trying to find workarounds for querying on a NoSQL database, so I built a direct interface for DynamoDB, which I'm calling DynamoSQL. With it, you can use standard SQL (even JOINs) on DynamoDB. I'm really excited about it, and I'm starting beta trials this week.
So instead of people using the right tool for the right job and learning single table design when using DDB, they are going to use the completely wrong tool. I can just imagine the sub optimal in memory aggregations and expensive use of read and write units.
Valid points. The read cost and in-memory aggregation concerns are real and worth flagging. DynamoSQL is index-aware and will use your partition keys, sort keys, and GSIs where it can, but you're right that a well-designed native access pattern will be cheaper.
That said, not everyone using DynamoDB has the luxury of designing perfect access patterns from the start. Real production tables are often messier than the textbook examples. And for analytical or exploratory work (e.g. understanding your data, debugging a production issue, running ad-hoc reports) a slightly more expensive read is a reasonable trade-off for writing SQL in seconds rather than SDK code in minutes.
I appreciate the feedback. The cost visibility question is something worth being clearer about in the docs.
For analytical are exploratory work, it still would be much more sensible to use DynamoDB streams with a simple Lambda and write it to an Aurora Serverless Postgres/MySQL table.
The tradeoff of doing aggregations and joins in memory just isn’t worth it.
I agree for cases that need real-time replication and are querying at high frequency. There is a trade-off, though. You're adding three services to maintain, there's replication lag, and Aurora's running continuously whether you're querying or not. For teams that need occasional analytical access to their DynamoDB data without building and operating a replication pipeline, I still think the in-memory trade-off is worth it. Different teams, different trade-offs.
An open-source TypeScript microservices framework I've been building solo for the 8 months. It generates and deploys an entire production-grade cloud infrastructure (VPC, gateway, WAF, observability, CI/CD) from a single config file. Multi-cloud across GCP, AWS, and Azure. Just shipped v0.2.0.
Built it because I got tired of writing the same Terraform, gateway config, and CI/CD for every TypeScript project. The application code came together quickly, the infrastructure took weeks every time.
Relevant to your "not AI" framing: the framework itself isn't an AI product, though it does include an MCP server so AI agents can manage the stack if you want them to.
I'm building a macOS image editor -- no AI here, just Metal textures, compositing, blend modes, all that mathematical raster goodness. It's getting closer to a Mac App Store release and has a preview version on https://skullrocksoftware.com
I am building nOne, the first subjective layer of the Internet, or more accurately, a permission layer underlying the engines/platforms/automation. Although it does pertain to AI, its more than that... it is a permission system for email, AI, and cognitive inputs (undisclosed automation, what you permit in your feeds, etc). It encompasses the entirety of the digital realm.
We have created a permission system within our social worlds... time to apply the same standards to the digital realm.
I have realized that playing board games with other people makes my mental health better and started work on a community platform and at some point physical space to connect: https://GamesAgainstLoneliness.org
Also working on a journaling app but it’s still in the works.
Do you worry people might avoid your site/shop due to stigma around loneliness? I am curious is people might end up being embarrassed to be around because they would get designated as "lonely". If you find that people are slow to sign up, you could potentially do some A/B testing and advertise the same idea with a name that is related but feels like it has a more "positive" twist to it?(Could be called "Ludo League" [Ludo being the Latin root for 'game'] or something. Sorry I am crap at coming up with decent names lol)
Also, since you are in Germany, have you considered going to the next Spiel Essen board gaming convention? I haven't gotten the chance to go yet, but I imagine it would be a good way to make a lot of connections and get your gaming group known to a bunch of people.
Also, boardgamegeek/rpggeek are places where you could get your name out there(just in case you don't know about those).
Lastly, not sure how common board gaming shops/communities are over there but they are quite popular and common over here in the US so if there is any extra stuff you need to know regarding gaming shops you could probably get a lot of useful info by googling (random big US city like New York City) + "board gaming shop". Star City Games is a super popular one here and they have a high quality website that might be worth studying.
Best of luck from an aspiring board game designer!
I’m not particularly worried about any stigma or anything. So far most people understand that it’s not “yet another another gaming club” but a community/place that is also for this particular purpose. Worst case scenario I can rebrand it later into something else.
love the website, but sounds like the community mostly needs places to actually go be together, feel like maybe the website and members thing could have happened after maybe?
Maybe over the capacity allocated for VPN originating traffic? I would never interpret a user-facing error message like that on a public site to be a reliable claim of what's actually going on in the underlying infrastructure. For any service.
That's very charitable. I can't imagine that they would segregate bandwidth like that. I've never had Imgur work on VPN no matter the endpoint or time of day or week.
Well PocketBot won't be AI related in a few years. We're using AI with it's only job being to eliminate itself. Automations, using a single simple prompt, fully private, with a massive user created library (~1000 testers so far). People create automations - our LLM makes a template out of it -> goes to our shared user library -> when someone asks for an automation we check if it already exists and if so then deterministically adjust it for their use. If no automation exists then request sent to our Tier 2 LLM which creates it. Maybe not exactly what you wanted to hear, I am just excited about our idea...
Edit: Sorry missed the most important point. So the more users and the bigger this library gets, the less LLM usage we have, until eventually we are able to run almost fully deterministic.
A crossword puzzle generator, just for fun. Grid generator in Python because it's easy to hack around and grid generation doesn't need to be particularly fast. Go for populating the grid with words because with large grids there are combinatorial explosions, and Go's speed is beneficial.
Not source-available yet because it's a bunch of hacks (particularly the Python) but maybe one day.
I've built Bubbles, a reader-curated frontpage for discovering articles from thousands of personal indie blogs. No AI, no algorithm, no signup. Auth and comments are handled via Fediverse handles.
https://bubbles.town
Broke my scapula and had never built myself (in Product, not dev) So decided to try to use Wispr Flow and keep productive since I couldn't type.
Built an open source chrome extension to build reports to provide back to our dev team. I promote it as AI related but it's not really. It just lets me grab elements on local files visually, comment on them, and provide them to our devs as a structured brief. There's others on the market but it wasn't worth paying for something so simple so I just made it free for all!
I'm working on a focus / pomodoro timer again that I started a few years ago.
The redeeming qualities, are that it doesn't require login, and you can just use it as a simple pomodoro timer that syncs to youtube / spotify music.
The additional features are all optional. Configuration, timeblocking, task management, etc... The features that are paid are just 1 (relatively low) price (with PPP per country). No subscriptions.
I started this project after pomofocus added giant ad banners, and had been missing some quirky things I had always wanted in a timer.
I started looking into AT Protocol recently and find it very interesting, so I started collecting a list of decentralized products built on top of AT that are alternatives to mainstream (popular) products.
One question, I tried sharing your site with a friend and the share URL is coming up as http://localhost/ which is strange and I've never seen before. Is this something that's possible to customise via meta tag, or is this a bug in my browser (Vanadium 147 on GrapheneOS)?
I’m reimplementing the Nix parser. I don’t know if it’ll morph into another formatter, or an LSP server, or what. I guess I’d like to try and improve on error messages and build it from scratch. Rust has really brought ergonomics to systems programming, and I wrote very few parsers for real languages, and never one with good errors.
Might add JSON feeds, RSS for now. Main goal was to validate if anyone is using these releases 2 years later (forks?), and the answer seems to be no. The FSL premise for delayed releases is primarily around risk mitigation (if company goes bankrupt or bought out), and that is also yet to be tested. Most FSL/BUSL forks are at the moment of license change (OpenTofu for eg).
As another datapoint, Oxide is maintaining their fork of Cockroach, but has ignored all future releases (that became FOSS).
MusicLibrarian, a macOS app that deep-cleans Apple Music libraries. iCloud sync duplicated my playlists up to 540x (77,542 ghost entries). Built it in Swift, uses ITunesLibrary framework + AppleScript. https://thevaultdj.com
I'm building this site to display behavioral therapy data better for parents / caregivers. The site itself is just Next.js but the essence is ETL and data visualizations using the Observable Framework[1]. There is an opportunity to apply AI for Q&A.
I'm building a rss client with an extra layer to have user comments/threads. This lets you create your own feed of articles entirely but with a social aspect.
I want it open source and free, just building an app that I'd like to use myself.
I built the exact same thing 2 years ago, and then figured out that Inoreader actually already had it in their product. They are disabled by default, probably because this feature is not popular at all.
I'm going with my fourth (or fifth) attempt to create a digital twin of my apartment in a game engine.
This time it's in Godot and i want to add more interactive stuff in the virutal version, rather than just redraw the rooms with some lights synced with Home Assistant.
I'm building a website integrity and security monitor. The backend is written in Java/PostgreSQL. The front end is written in JS/React. It will allow for interactive use via front end or API driven.
I initially was using SSE to push events down to the front end during long scans but decided to switch over to plain old HTTP polling for better reliability across different browsers.
Going for a 100% self-service model. No corporate sales cycles, no slide decks, no meetings. If the tool works for you, you buy it.
Built/building godom, a framework that lets me build local apps in Go, with the browser serving as a dumb view layer. I don't hate JavaScript or React, but my primary motivation was to eliminate NPM as much as possible.
https://github.com/anupshinde/godom
I used AI to create the first POC, and once it was proven, it was improved, and AI handled a lot of grunt work where it could. The framework was built primarily to solve my pain points
And building Fractiz, a customizable pre-coded backtests platform.
Uses government IDs that a lot of people already have and saves signed credentials to your phone. There is a server element for the verification process but it runs all in-memory (follow the Mullvad model and not even have persistent storage in the servers).
It's fun. Get to practice mobile development (Flutter), use some local GPUs, learn about the changes to JWTs for signed, selective disclosure, Bitstrings for progressive disclosure lists etc.
Landing page is https://agora.gdn/ in case anyone wants to try a beta in the future.
I love watching YC Videos. So i vibecoded an app which found interesting clips from YC videos and show in like tiktok format. I get new content every other week, so makes scrolling fun again.
I have an invoice/order taker https://www.bestsellerapp.net, it syncs to many ERPs and allow people to operate everywhere, that I'm upgrading to be a more general mini-ERP.
On the side, retaking https://tablam.org, that I have procrastinated because wanna provide a parser that work for editors (that is way harder than normal batch parsers!)
I'm making more and more art. I'm meeting friends, and we sit together and doodle and paint. I've made so much progress in a few months simply by practicing more, experimenting more, and allowing myself to waste paper and make mistakes. A few friends have framed my art.
It's nice to do something away from the screen, without any goals besides having fun.
Very wholesome! I find that using cheap paper, and basic supplies from the dollar store really gets rid of hesitation when making art. After all, you're definitely not "trying seriously" when you use low-end supplies...
I've got a few things on my plate that I bounce between building for fun:
1. A point-and-click adventure game: making it with incredibly heavy technical constraints "just because"
2. A coding puzzle game of rapidly escalating difficulty
3. Part of #2 had me needing to craft some JSON Path queries and I felt like there wasn't anything nice to build and test them with, so I built this tool for it (inspired by the amazing regex101): https://jsonpath101.com/
4. A website where I write about text-based browser games
Also working on a Korean guide, so far it introduces Hangul (한글), the Korean alphabet: https://tolearnkorean.com/
Also making an app / web app for memorizing Korean words, takes inspiration from Anki and Duolingo. Words go through 4 stages: 1) matching up pairs 2) Multiple choice answer 3) Writing word through blocks 4) Free-form writing.
I'm scratching my own itch: building a yet another audio signal generator for smartphone. I need some extra functionality that is not available elsewhere, and impose some limitations "just because": it must be bare minimal PWA.
But actual app does not matter, the main take away for me is: it is easy and fast to write bloatware (esp. with AI), but not that easy to distill to what is really needed. And what looked like a weekend project, a couple of hours max (with help of AI), now lingers for 2+ weeks on-and-off on evenings (of manual effort).
Working towards building a network fault attribution system oriented towards NaaS platforms that manage their own hardware. In the experimentation phase with existing monitoring frameworks, anomaly detection, and fault localization for sub-optimal routes.
Possibly see a use case in home labs too, but not there yet.
I recently built Cranki [0], a free little PWA that generates crosswords using your Anki flashcard lists. It's aimed at language learners (who find flashcards boring and crossword fun!). To be honest I built it just for me and then decided it might be useful for others.
It's all local, no server, no database, etc. Mobile and desktop friendly.
I did build it with the help of AI, but no AI inside the actual thing.
I am building a web application for learning math. I want it to be something between Khan Academy and Math Academy. Here is a demo of fourth grade https://demo.numerikos.com/ Currently the best part about it is one of my kids is using it. I have some more lessons ready, waiting to be released and I am currently working on Trigonometry.
Flora Carta, design and keep track of your gardens. Originally a project for myself to keep track of every rose variety (I have over a hundred) in my garden, and apple/pear/plum variety in my orchard.
I've been working on a super-straightforward volunteering platform -- https://handsup.barryvan.com.au/ . Mainly because I got frustrated with my kids' school and sailing club activities being (mis)managed across WhatsApp, Google Forms, pieces of paper, and the like.
Building automated analysis tool that help identify the use of GPL-licensed SDKs in mobile apps, promoting license compliance and supporting sustainable open-source development.
Knowii, a community for knowledge workers who want to excel and thrive. We explore the frontier of knowledge management, personal organization, zen productivity, clarity, personal development, project management... And AI
Every time I have to open Acrobat it makes me want to puke, and Bluebeam is quite expensive, so I'm working on a full-featured PDF editor you can run in the browser.
I am building something similar to Windows Task Manager but it also works on Linux, launches/monitors web servers, and provides a visual dashboard for docker containers.
I'm currently working on a small tool that helps me setup new personal computers in a purely descriptive way.
A little bit like ansible, but then totally not like ansible.
It generates SSH-Keys. It clones repositories, installs uv and rust. It removes Snap from my Ubuntu machines and installs firefox from the mozilla repositories.
iCloud-photos-downloader is unmaintained and starting to crack due to Apple API changes so I’m working on my own fork/rewrite/take on iCloud sync, with ambitions to add Google Takeout, Immich, and other services too.
I wrote a book about software engineering mental models that barely mentions AI on purpose. Only in the Appendix section I explain the rationale for not talking about AI, pasting it here:
---
Thoughts on AI
I consciously decided not to talk about AI throughout this book. Not because I don't believe in the benefits of using AI... I do.
I believe AI will keep bringing a lot of value to society.
I believe AI will keep changing our profession in many profound ways.
But whatever happens, I believe the principles here are still going to be incredibly valuable, even if the software engineer profession ceases to exist with its current name.
We might be working at a totally different level of abstraction, but the values, principles, mental models, patterns of communication and behavior described in this book will still make a huge difference. I believe they are atemporal.
Having said that, here are my recommendations regarding AI:
- Think of AI as another tool you have to create value, just like your IDE, the code you write or the emails you send.
- Keep yourself long enough in the problem space before throwing an AI api to solve a problem that might not exist
- Unless you are working at the frontier of AI development, ignore the noise. AI is now the shiny object everyone wants to be up to date with. JavaScript frameworks were used as a joke between value-driven software engineers because there would be a new one every 6 months and a lot of people would move to it without the real need. AI changes and improves every day or week. Being on top of it is probably a full time job. Instead of following all the news on a daily basis, level up your AI game in batches every 3 or 6 months. It will be more than enough.
- The most interesting use of AI for me has been in finding out my unknowns-unknowns. One question I frequently ask is "What are the building blocks of this problem/system/piece of knowledge?"
Took some good ideas of Pascal and making it more modern. Minimal runtime, manual memory management, single (small) executable, no dependencies. Compiler itself is written in Swift and I am using QBE as a backend ATM.
Thanks for this thread. I'm so tired of reading about yet another "AI wrapper for sending emails" raising a $5M seed round that I almost forgot what actual engineering looks like. The whole industry feels like a Gold Rush right now where 99% of people aren't even trying to mine gold, they're just reselling each other the exact same shovels (LLM APIs) with different logos slapped on them. It's incredibly refreshing to read about people writing Rust parsers or soldering hardware synths
I'm building the CAD app I have always wanted. Free, Parametric, STEP export for 3d printing. The input is just a javascript editor. It uses OpenCascade under the hood. Lots of fun.
I have been building cross platform QT desktop apps with PureScript and NodeGUI. PureScript is truly underrated as a language and as a gateway to the entire JS ecosystem!
I've been building the games I've always wanted to play. It used to take months to make something like Medieval Building Simulator.
Now you tell some idea of what you want and AI gives you a better thing than what you originally imagined. Then you go into a rabbit hole... simulate productivity, sales, customer, cash flow, materials for the building, make it curvy, make it pop up in 3D, populate it with NPCs, give your NPCs wages and background, write a dialogue with the ghosts of Christmas. It ends up being a game in itself.
It's good fun. It costs me $20 and I can do it and deploy it from my phone.
i am building a chaos testing CLI for distributed node systems in Go, started it because every time I wanted to test how a consensus protocol behaves when nodes start dying, I ended up writing throwaway bash scripts, Got annoying fast.
i’m currently avoiding committing/releasing a bunch of changes i did last week because people are actually using the library already (the curse of writing something useful lol)
i made my own personal website [0] that loads in less than 512kb; with a mini bio, a blog (with RSS support), images, links to friends' websites, stuff i made/maintain, and some fluff[y boys] like a walking Ralsei [from Deltarune] that follows the cursor, a music player, and some secrets/easter eggs ;)
then there's another project i have, which is a download-and-run version of Minecraft, where you just download one binary (most likely an AppImage) and can run the full game. it will be for personal use so i shouldn't worry about copyright.
i'm also making my "own" minimal tiling window manager for Linux [1] (which is a fork of [2]) with my custom keybinds and [planned] controller (joystick) navigation support.
and lastly, i made yet another fetch software: jotafetch (*jota*lea's implementation of neo*fetch*). it should be available to read and download a
on my website [3] (i haven't packaged it for mainstream distros yet, and i probably won't).
Since the moment "AI" became a thing, I've been rolling my eyes and looking the other way since I know it's just a half-assed hype buzzword. With that in mind, pretty much everything I work on is not AI related. As a hobby(and to a degree to be prepared for WW3), building a ton of drones and drone equipment, flying them and finding more and more ways to push their range to the absolute limit.
I don't think it qualifies as AI in the modern day and age but NLP in general. It's truly amazing how easy it is to spot troll farms online and no one is doing anything about it, be it individuals, private sector or even on national level, given that those should be considered a risk for national security.
The goal is to make every recipe foolproof on the first try, similar to when you walk into a restaurant and just pick what you want to eat without thinking about the details. The goal is to have the same experience, just pick what you want to eat, with recipes that tells you exactly what to do with no magic involved.
Technically it is probably the most advanced recipe app ever made. The database is a huge graph that captures the relations between ingredients and processes. Imagine 'raw potato'->'peeled potato'->'boiled potato'->'mashed potato'. It is all the same ingredients but different processing. The lines between the nodes define the process and the nodes are physical things. Recipes are defined as subsets of the graph. The graph can also wrap around into itself, which is apparently needed to properly define some European dishes in this system. The graph also has multiple layers to capture different relationships that are not process related.
Why was it designed it in this way? Because food/cooking is extremely complex. This design is the only way I have found that can capture enough of these complex relationships that the computer can also 'understand' what is going on.
My favourite thing about this is that each recipe is strictly defined in the graph. If the recipe skips a step, or something is undefined, the computer knows that the recipe is incomplete. It won't ask you to do 10 things at the same time and then have something magically appear out of nowhere. It is like compile time checking but for recipes.
It also enables some other superpowers, for example:
• Exclude meat part of the graph = vegetarian. Same thing works with allergies.
• Include meat part of graph = only show me recipes that contain meat.
• Recursive search: search for 'potato' and the computer will know that french fries are made from potato. It can therefore tell you that you could make the hamburger meal, but you will need to complete the french fries recipe first, which should take 60 minutes.
• Adjustable recipe difficulty (experimental): It knows which steps can be done in parallell, and which can't based on how the nodes connect. A beginner can get a slower paced recipe with breathing room between steps, while someone more experienced can do a faster pace and do more things in parallell.
If I knew what it would take to build this, I would never have gotten started. I completely underestimated the complexity of the problem I was trying to solve. But here we are, and now it is basically done and working.
The website is slightly outdated but captures the key points from a non-technical point of view, and you can enter your email and get notified when it will launch in your country.
I'm building a no bs esign product, with api access by default. At a previous company, automating this stuff cost us an arm and a leg and we got locked into a bad contract. I've always been on the engineering side of things and never had the opportunity to learn how to find customers and sell. It's been a fun and rewarding experience for me. https://www.sign-n-post.com/
https://www.adashape.com/ - a humanistic CAD for 3D printing. Early but has profile editor, extrusions, booleans, and imports STL, STEP, SVG and bunch of visual mesh formats.
This seems a little bit of a duplicate of [1], but I can repeat my answer here as any views help!
My wife and I continue to work on Uruky, a EU-based Kagi alternative [2].
Since last month we finally got our production API Key for EUSP/STAAN (it was certainly the slowest and most complicated search provider to adopt, so far), and that brought us to 5 search providers you can choose from and sort as you prefer.
We already have got over 40 paying customers (excluding family and friends, we’re guessing these paying customers came from some privacy listings and HN comments) and have exited beta last month!
Customers seem to really enjoy the simple UI (search can be used without JS) and search personalization (from choosing the providers to the domain boosting and exclusion). We also have hashbangs (like "!g", "!d", or “!e”) when something doesn’t quite give you what you’d expect, though.
You can see the main differences between Kagi and Uruky in the linked page, but one huge difference is that with Uruky, after being a paying customer for 12 months, you get a copy of the source code!
One thing we’re struggling with is outreach because we want to do it ethically, and it’s hard to find communities or places to sponsor which are privacy-focused and don’t require €5k+ deals. Ideas are welcome!
Because of bots there isn’t a free trial easily available, but if you’re a human and you’d like to try it for a couple of days for free, reach out with your account number and we’ll set that up!
Thanks.
P.S.: Because people have asked before, our tech stack is intentionally very "boring" (as in, it generates and serves the HTML + bits of JS to enhance settings and such), using Deno in the backend (for easier TypeScript), PostgreSQL for the DB, and Docker for easier deploying.
P.P.S.: Because this has been also brought up before, the name has no special meaning but we read it like "Euro-key" in English. Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
P.P.P.S.: Another frequent question here is “how does it work?” When you search, we query the first search provider on your list, and if it yields less than X results (only Mojeek really gives us a total count, we have to try + estimate for the others), we try the second, and so on. We then merge the results in a round-robin fashion (first of first, first of second, second of first, second of second, and so on). There’s a bit of more nuanced logic to also properly rank the results with the pin/exclude/raise/lower preferences, because it works differently across providers and not all of them support that, for example.
Regarding the bots: since you're building a privacy-first product, you should look into a Proof-of-Work captcha (like Hashcash or mCaptcha). Just have the user's browser mine hashes for a couple of seconds before issuing the trial token. A normal human won't even notice it, but it'll burn so many CPU cycles for bot farms that abusing your API becomes economically unviable
I love the idea! It's so good to see people advancing the european tech landscape. Just out of curiosity: Are there more reasons for building Uruky than Kagi being located in the US and closed source ?
That’s a good point, thanks for mentioning it. I think we’re safe since Uruk can’t be copyrighted as it’s an ancient city’s name and the name of a time period, and there are no intentional Tolkien references on the product.
I recently made a 'notification-driven RPC' app for Android to help me help others:
Notifactor (Android App)
A lightweight native Android app that intercepts notifications on the device and triggers actions based on configurable rules.
The app uses Android's NotificationListenerService API. Once granted notification access it receives a callback for every notification posted on the device. It then checks each notification against configured rules and runs the configured action.
Why would you want this?
I don't know why you would want this but I can tell you why I made it: to make it easier to control some functions on Android devices used by people I often help using them in some way. My mother's Android TV (which I use to communicate with her through Linphone), phone and tablet sometimes stop doing the right thing. I live about 1300 km to the north of where she lives so I can't just hop on my bike to fix things. Thus far I relied on a set of Termux scripts on these devices to keep a reverse ssh tunnel open to an endpoint on my server but this has a number of drawbacks: the tunnel is not always there when I need it due to WiFi dropouts and other similar problems and the constant connection uses battery power on the phone and tablet. If only I could cause the tunnel to be created when I need it and brought down when it is not needed... Well, that is possible using Notifactor by sending a notification on a specific channel (ntfy refers to these as topics) whereupon Notifactor runs a Termux script which manages the tunnel (etc.).
I'm building my own cloud (I actually typed claude instead of cloud there... wow). There's no IaaS or PaaS; it's much simpler. I wanted my own way of connecting to machines and the TCP services on those machines without having to install Tailscale (not allowed on a locked-down corporate PC) or pay for Azure or AWS or GCP or even Hetzner or Linode. I've got 10gbps fibre and a huge workstation at home, and I've got lots of laptops and VMs and other outboard stuff that I want to work in concert with that workstation, so I started building something I call Tela (Filipino for fabric; I was sitting in Ninoy Aquino International Airport waiting for a flight when I had the idea, and it's implemented as a network fabric).
https://github.com/paulmooreparks/tela
You are rebuilding tailscale but requiring a centralised hub. I did the same in 2016 (pre-tailscale?) https://web.archive.org/web/20160304013451/https://wormhole....
I am not sure WireGuard existed at the time, and I used SoftEther and based it all on doing outbound tunnels to TCP/443* to avoid firewall blocks in corporate networks.
You could explore full P2P by leveraging UDP hole punching: https://cloudnetworking.pro/firewall-bypass-series-1-2/ https://cloudnetworking.pro/firewall-bypass-encapsulating-tr...
(WireGuard may already do it, dunno)
Also, fun fact, tela is also Spanish for fabric. Given the Filipino history, I guess it comes from there.
* I know I know, TCP in TCP is a bad idea https://cloudnetworking.pro/tcp-over-tcp-is-a-bad-idea/
Thanks! Yes, Tela already does UDP hole-punching. I made Tela because I wasn't allowed to install Tailscale on my new corporate laptop, and no other available solution seemed to tick the right boxes. It started as a simple way to RDP to my home workstation, but then I realised that if I could do that, I could finally pull my ad-hoc home cloud into one tool. The hub model is very much by design, for organisational purposes. The hole-punching feature gives me the P2P speed (and even STUN, if available). An upcoming version will allow hub-to-hub topologies.
It should have occurred to me that tela is also Spanish, since about every third word I hear in a Tagalog sentence seems to be of Spanish origin.
Does tela create an L3 network? if that's the case, what do you do to avoid IP addressing clashes? In Wormhole I decided by default to use CGNAT addressing (100.64.0.0/10)
I did not go too far unfortunately, so I did not face problems such as discoverability (do you have to know/remember all the IP addresses from the devices connected? DNS? etc).
I am building a guitar learning platform where the main feature is an RPG game called Guitar Quest. You plug in your real guitar into an audio interface and the game listens to what you play based on whats shown on the screen. Play the right note, the monster takes damage.
The goal is: fretboard memorization, scales, music theory, chord construction, and performance. There's also a tuner, practice tracking, backing tracks you can jam over, and a community for sharing your gear collection.
No AI anywhere in the stack. Just a guitar, a microphone, and some monsters.
https://openfret.com/game/demo if you want to try it (free demo, no signup).
Working on an audio streaming platform powering an indy internet radio. Looks like Icecast & friends show its age and a similar product can be easily built with the functionality cast down to simply robust streaming & handling "timed playlists". I enjoy every bit of knowing exactly what happens in the code. It's not open source atm, but will be. It's in Go, is a pleasure to write and the deployment takes minimal amount of resources.
Other project is to continue a bit stalled progress of a configuration language BCL - add functions, more structures and fix some hidden scoping issues. Making languages is an endless fun. https://github.com/wkhere/bcl
Was recently looking for such a streaming thing; just streaming from a set of MP3s in a folder, nothing fancy. The majority looked too complex, with too many moving parts, for my idea.
Found one in Go that might interest you too: GoFM. Although I dropped my idea for now, I'd love to see yours come to life, too.
I started with this one:
https://medium.com/@icelain/a-guide-to-building-a-realtime-h...
And then, modified a lot. At some moment I will open it back. (Author's MIT license allowed closing it; I did it actually because I embedded a number of idiosyncrasies related to the radio service that shouldn't be disclosed; but with some amount of work it can be divided into an open and closed part).
The broadcasting skeleton from that original blog/github project is good, though! It might work for your case.
Please keep in mind it's better to stream AAC than MP3. Basically any format you'd probably want to use can be converted to AAC with ffmpeg.
AAC has a simple frame format and it's easy to decipher it; I use it to always send full frames, even when one would want to skip to the next song - by doing that the client behaves more smoothly.
Thank you for the pointers.
A programming language to hack music, and anything else really https://github.com/audion-lang/audion
The idea came after I finished a permanent piece for a museum using MaxMsp and python. I always had this thought in the back of my mind that "I could express this so much easier in a few lines of code.."
here's the language spec: https://github.com/audion-lang/audion/blob/main/docs/LANGUAG...
I really liked how objects came out, I don't think it needs any more since I can do object composition.
There are some nice functions to generate rhythms and melodies with combinatorics, see src/sequences.rs and melodies.rs
Its a WIP but you can use it now to create music with whatever you want: hardware/daws/supercollider
supercollider is tightly integrated but not required. I havent had time to develop userland libraries yet but I'm working on it
I'm building an alpaca breeding and show management SaaS which is the first one to combine animal register with a strong approval workflow (think like a pull request), herd management and show management. No Ai involved but might be something I'll put on the roadmanp to help with breeding decisions. (alpacuna.com)
My friend and I are building Strandfall, a highly physical outdoor larp (live action role playing game) that uses custom spatial computers: https://strandfall.com
Players are survivors of a global disaster that has unleashed mysterious, deadly storms. For three hours, they investigate the origin of the storms and make fateful decisions about their future as individuals and as a community.
We received Immersive Arts funding, which means we can run it in Edinburgh later this year. Here's an excerpt from our 2025 grant application about exactly what those are:
--
Our “storm sensors” are novel spatial computers designed for outdoor usage over long distances. They will house ePaper displays, LoRa (long range) radios, accelerometers, gyroscopes, and GPS chips in a 3D-printed enclosure to provide a low-tech way to augment the reality of the park. These computers will be cheaper, more rugged, longer-lived, and more capable than smartphones, deployable to locations with zero cellular service and no battery charging options.
The sensors will be mounted on top of camera tripods for deployment. Runners will carry them through the park, then position and aim them in the correct direction, as co-ordinated by “operators” using walkie talkies. This will let players feel like they are really setting up important equipment, scanning historical sites for clues (like surveyors), and establishing laser communication links. Lacking colourful touchscreens, the sensors will be less distracting for runners, helping them focus on their surroundings. Essentially, they are a highly tactile and deeply realistic way of immersing players in a post-apocalyptic setting, since such devices – not smartphones – are the most likely to be used.
this is totally not going to cause spontaneous reports to authorities about suspicious _insert anything here_ and ultimately cost taxpayers more to respond to situations that otherwise wouldn't be happening.
It might, if we weren't planning to put prominent stickers on the tripods making clear they're part of a larp, and also getting permission from the park and council and consulting with local parkgoers.
yeah but scale it up so it's everywhere suddenly but without common understanding in the public sphere
And? Not as if governments don't already waste money?
I'd rather my tax dollars be wasted on people having some harmless fun compared to the other things they can be spent on.
I'm working on ghidra-delinker-extension [1], which is a relocatable object file exporter for Ghidra.
The algorithms needed to slice up a Ghidra database into relocatable sections, and especially to recover relocations through analysis are really tricky to get right. My MIPS analyzer in particular is an eldritch horror due to several factors combining into a huge mess (branch delay slots, split HI16/LO16 relocations, code flow analysis, register graph dependency...).
The entire endeavor requires an unusual level of exacting precision to work and will produce some really exotic undefined behavior when it fails, but when it works you feel like a mechanic in a Mad Max universe, stripping programs for parts and building unholy chimeras from them, some examples I've linked in the readme. It has also led to a poster presentation to the SURE workshop at ACM CCS 2025 in Taiwan as a hobbyist, an absolutely insane story.
[1] https://github.com/boricj/ghidra-delinker-extension
Mad respect. I tried extracting a clean .o file out of a statically linked ELF once, and it's an absolute nightmare. How are you handling switch tables and indirect jumps? Without dynamic analysis, it's sometimes physically impossible to figure out what a register is actually pointing to
I have analyzers that resynthesize relocations from the contents of the Ghidra database, no custom annotations required. They evaluate relocation candidate spots through primary references and pointers/instructions and emit warnings if the math doesn't check out.
It does require a reasonably accurate Ghidra database to work properly, but I've had users delink megabytes of code and data from a program successfully (as in, relinking it at a different address results in a functionally identical executable) once they've cleaned it up. The accuracy warning in the readme is mostly because it's really complicated to describe exactly what inaccuracies you can get away with, there's a fair amount of wiggle room in reality as long as you know what you're doing.
I'm building OSS structural biology libraries in Rust (+ CUDA), and a GUI application that ties them together. It is kind of an integration of functionality traditionally in separate tools: Viewing, editing, therapeutic properties + pharmacokinetics, molecular dynamics etc.
I'm building an E2EE encrypted Journaling app with option to upload photos and voice recording (that too is encrypted)
The files and content are encrypted on the device and then gets to sent on server and stored encrypted and when you load the app it fetches E content and decrypts on device, with E Key (which is always stored on device, if you lose the key, you lose all data cause we don't store it)
archivist.anasdew.com
Ive been working on the HN Arcade :) https://hnarcade.com
Its always fun to see what games people are building - and some of the lesser known ones are amazing!
I enjoyed this! Thanks for the index
Nice!
I am working towards a big new release of my web scripting language, hyperscript:
https://hyperscript.org
Hoping to release next Monday
Big fan.
Looking forward to (hopefully) Monday.
Building a bouldering wall at home, for the kids.
A bunch of square panels with a grid pattern to mount hold on, with the panels hanging on french cleats (with a locking system, #TODO) so the panels are easily removable so I can hang something like planters on the wall as well with the same french cleats.
No AI, a bit of computers to draw things out in CAD, but otherwise just manual building stuff.
Pictures?
https://imgur.com/a/2TIye34 not much to see yet: square panels with a lot of M10 holes and edges (also the edges of the holes...) sealed. I just got the holds, bolts etc the other day. I sanded the panels & primed them for paint last night.
The larger panel is going to be a volume with a bit of a hang. Panels are all tough 18mm outdoor rated 'betonplex' with phenol? resin or the surfaces.
Cool, thanks for sharing! Don't forget maybe having some pads underneath for any falls :-)
Yeah, that table is going to be quite the natural selection filter I suppose. Crash pad is on the way.
I'm building a tool (no AI) helps you outsmart everyone with the methods of Sherlock Holmes: https://todofi.com/
I'm working on the next update of Omiword[1], an ongoing daily word game previously discussed on HN[2]. I'm building an alternate stand-alone app version with access to all of the archived puzzles. It's slow going since it's just one of many side-projects I like to work on, but that's the tinkerer's dilemma.
Yes, I know that [insert LLM here] could do a lot of that conversion for me in mere minutes. No thank you. I'm doing it, in part, for the doing.
[1] https://omiword.com
[2] https://news.ycombinator.com/item?id=43654350
"Cars on Mars" ...a scifi racing/shootout/exploring game with portals into stylized worlds - all content is purely made from math+code https://store.steampowered.com/app/4269050/Cars_on_Mars/
I'm still working on Nottawa (https://nottawa.app) - free live audioreactive visuals for everyone.
It's a node based workstation for generating live visuals using shaders, webcams, local files, and more. 100+ effects and source types.
Nottawa's aimed somewhere between the Resolume and TouchDesigner market. VJs and musicians who are interested in live visuals but don't have the time or funds to dive into a pro setup.
It's built in C++ with OpenFrameworks, ImGui, BTrack, and more. Right now I'm doing a front-end rewrite in SwiftUI which bridges back to C++ while keeping all the rendering in OpenGL.
Oh wow, that looks sooo awesome!
Why only mac? I would love to see Windows support! The girhub link in the header doesn't work, it's 404.
I am working on a network scanner in Go. No AI there, just probes and packets
https://github.com/ostefani/subnetlens
I’ve talked about it many times here, but I’m slowly building golfcourse.wiki, a golf wiki, because the golf world focuses all it’s attention on “the best” golf course, and the vast, vast, vast majority of interesting courses get zero attention.
I thought it was a website that should exist during the pandemic, so I built it. It’s been slowly growing ever since, because most of the time you really only need to add the info once.
JDBC does not allow pipelining (a Postgres only feature).
It can reduce the number of db round-trips a lot, especially when using Supabase+RLS (or other systems that require frequent setting of configuration values that are basically fire-and-forget).
Meet Bpdbi, a library with first-class pipelining, which provides a Postgres db driver (that's binary only, as the legacy text-based protocol is no longer needed, it just takes up space) and exposes an API that's more close to Jdbi's that to JDBC's (developer friendly).
https://github.com/bpdbi
It has an extensive benchmark that shows it's on par or faster compared to other db connectivity stacks.
So great to see how much stuff there is going on outside of AI!
I've been working (on and off for over a decade) on a way to manage my unwieldy photo and video collection. I experimented with so many tools before but ultimately nothing ticked all my boxes so I wrote my own. Admittedly lots of niche stuff like proper support for UTC time, stereo photos and partial dates and things like 100% offline face detection and custom tag hierarchies and querying without a database. Each time I add a new feature I'm having a blast :)
https://gitlab.com/geekysquirrel/memo2
Does it store thumbs separately from the photos? I see external usb was mentioned, can it support several (possibly disconnected) drives?
I'm building a good intermittent fasting app for iOS using SwiftUI and native Apple APIs for the most "Apple" experience possible. The app is 100% free and without ads.
https://apps.apple.com/si/app/fast-slow-fasting-tracker/id67...
Between job searches, I'm working on implementing the WHATWG Web APIs on top of Elixir.
https://github.com/sntran/web
I love Elixir but I had been a JS developer by trade. Bridging that interface keeps my brain focusing on building instead of splitting between JS and Elixir.
Still a lot to do, but I love the progress so far. This puts the joy of building something back to me.
Tusk - An open-source Postgres desktop app for macOS and GNOME - https://shapemachine.xyz/tusk
I recently started with Paperless-ngx and wanted to also include archive serial numbers (ASNs) for all paper documents using small label stickers, so I built a small tool to create and print ASN label sheets. It's free, no sign up, no ads and just runs in the browser:
https://asnlabels.com
If you're also using Paperless-ngx with ASNs or use them for something else, feedback welcome :)
LEGO. I build LEGO, because I like it and it puts a smile on my face :)
Question, do you "build LEGO" as in "make what's on the box it came in", or do you use LEGO to build things yourself and not following instructions?
(I am old, but as a kid, I remember building what the kit was supposed to build once or _maybe_ twice, but using the parts from the kit and various other kits I had to make things purely out of my head pretty mush all the time. I _think_ that's not how people "play with LEGO" these days?)
Both, I enjoy the process of building, so following instructions. But then it's done and boring. I.e. LEGO supercars sets are nice, they're so nice until you marriage chassis with drivetrain - you can play with all the gears and see them, after the marriage it's just boring. So now I'm working more towards reusing parts and building things of my own.
Edit: LEGO Architecture Set (not sure if they still make it) is super nice for building random stuff - it's a set of pure white bricks, super fun.
I was the same. My kids will only build things from the instructions
https://GetCozyCritter.com
It's an ESP32 board used as sleep trainer for children.
Uses colors, images, and arcs to help them understand what time it is.
Full prototype is done but waiting for 3d printed enclosures.
I'm building a YouTube for recipes. Basically it is YouTube clone, but improved to follow the actual recipe (ingredients, steps, etc.)
Was frustrated with discovering and following online recipes (blogs are fludded with SEO content, YouTube is not optimized for recipes).
So I gave it a try to build something I wanted.
https://www.cookcook.it
What do you mean by "YT is not optimized for recipes"?
I like this but your hosting is also through YouTube?
I just built a Google Chrome chess extension. It is a new tab page full of chess puzzles. It has various difficulties, and a speed run game (solve as many puzzles as you can in 60 seconds). I personally think it's the coolest browser extension I have ever seen (being objective as possible).
It works on Chrome and Brave for now. Planning a Firefox release soon.
If you're a chess fan, you'll probably love it, but if not, let me know why (I'll try to improve it).
https://chromewebstore.google.com/detail/leetchess/hogbcffpf...
I'm building my woodworking workshop. It's gonna take a long long time for AI to invade this area if ever. It's very therapeutic and works as an escape from all the AI craze.
I’m building websites designed to feel as hand-coded as possible! A random assortment of games, digital instruments, off-kilter tools, technical experiments, and more. Complete with a homepage that’s literally handmade ~ featuring tons of hand drawn GIFs! It’s a good place to waste some time :)
https://barcoloudly.com/
VERDURE is still a creative plant-generation sandbox where you grow and sculpt stylized trees.
https://store.steampowered.com/app/4069810/VERDURE/
A replacement for CMake/Ninja using golang.
https://github.com/bradphelan/nuke-engine/blob/trunk/USERGUI...
After a day of hating on CMake generator expressions I just wanted a proof of concept that something better is possible.
An example build is https://github.com/bradphelan/nuke-engine/blob/trunk/example...
Playdate games! Super constrained/lo-fi/retro and really a joy to program without AI.
Hyper Vector is dropping on Catalog next week!
I'm working on a daily journal app using rust and iced. Where you can quickly jot down your thoughts or information in a single line
All data are stored in text file, following the most general markdown formats, so it's easy to just open the text file and change.
Also, it supports TODO, and looking to add support for reminders, scripting support, calculation, tabular data using CSV, etc.
UI is complete keyboard driven.
Core intention is, managing your journals in single line, with really fast keyboard access, and stored in text file
https://exploretrees.nyc/
Been tinkering on this Olmsted-inspired map of all the trees in NYC for a long time - just added some more features (seasonal colors and better search/filtering) recently.
If you're in NYC, try finding some cherry blossoms near you: https://exploretrees.nyc/?species=cherry
I'm building a flash card learning tool. I was frustrated with the slowness and complexity of other solutions. Animations and loaders with tools like Duolingo, and unnecessary extra steps to move onto the next card in tools like Anki. If you give it a try I'd love to hear your feedback.
https://flash.boon.gl
I am building educational 16-bit game console for low-cost FPGA chips.
https://github.com/true-grue/Brus-16
Thank you for this!
I’m making a digital ESP32 powered synthesizer at https://subalpinecircuits.com/. It’s been a ton of fun, learning so much about every part of the process (and I don’t use LLMs). Currently I’m learning FreeCAD and figuring out what my case design will be. Woodworking, CAM and DFM is a whole other world for a software guy!
still working on https://stringscales.com - fun sideproject to visualize guitar scales on a configurable fretboard, with interactive note highlighting to a backingtrack.
The backingtrack is what I'm actively improving right now. It's just a pad running now, but it will turn into a full track with bass/drums/piano/... and will feature a comprehensive chords based editor so you can add and save your own progressions with a logged in account.
Does it let you visualise quarter tone fretting? (Or is it just me who's obsessed with Angine de Poitrine right now?)
that would certainly be a very interesting thing to add, but in the current iteration that is not possible.
The biggest issue here is that there's not really "one" microtonal system out there. The entire fretboard of a conventional guitar is mapped to work in 12tet - and the libraries I'm using to do all the musical operations also only supports 12tet.
to accommodate all the microtonal temperaments out there would be a pretty daunting task. But I'm not saying never!
Built a SQL interface for DynamoDB. I was tired of constantly trying to find workarounds for querying on a NoSQL database, so I built a direct interface for DynamoDB, which I'm calling DynamoSQL. With it, you can use standard SQL (even JOINs) on DynamoDB. I'm really excited about it, and I'm starting beta trials this week.
https://dynamosql.com
So instead of people using the right tool for the right job and learning single table design when using DDB, they are going to use the completely wrong tool. I can just imagine the sub optimal in memory aggregations and expensive use of read and write units.
Valid points. The read cost and in-memory aggregation concerns are real and worth flagging. DynamoSQL is index-aware and will use your partition keys, sort keys, and GSIs where it can, but you're right that a well-designed native access pattern will be cheaper.
That said, not everyone using DynamoDB has the luxury of designing perfect access patterns from the start. Real production tables are often messier than the textbook examples. And for analytical or exploratory work (e.g. understanding your data, debugging a production issue, running ad-hoc reports) a slightly more expensive read is a reasonable trade-off for writing SQL in seconds rather than SDK code in minutes.
I appreciate the feedback. The cost visibility question is something worth being clearer about in the docs.
For analytical are exploratory work, it still would be much more sensible to use DynamoDB streams with a simple Lambda and write it to an Aurora Serverless Postgres/MySQL table.
The tradeoff of doing aggregations and joins in memory just isn’t worth it.
I agree for cases that need real-time replication and are querying at high frequency. There is a trade-off, though. You're adding three services to maintain, there's replication lag, and Aurora's running continuously whether you're querying or not. For teams that need occasional analytical access to their DynamoDB data without building and operating a replication pipeline, I still think the in-memory trade-off is worth it. Different teams, different trade-offs.
tsdevstack — Infrastructure as Framework
An open-source TypeScript microservices framework I've been building solo for the 8 months. It generates and deploys an entire production-grade cloud infrastructure (VPC, gateway, WAF, observability, CI/CD) from a single config file. Multi-cloud across GCP, AWS, and Azure. Just shipped v0.2.0. Built it because I got tired of writing the same Terraform, gateway config, and CI/CD for every TypeScript project. The application code came together quickly, the infrastructure took weeks every time. Relevant to your "not AI" framing: the framework itself isn't an AI product, though it does include an MCP server so AI agents can manage the stack if you want them to.
https://github.com/tsdevstack https://tsdevstack.dev https://youtu.be/6MJ4PPPjxH8 https://dev.to/gyorgy/i-built-a-typescript-framework-that-ge...
Very cool! Have you seen https://encore.dev/ ? Haven't used it personally but I saw it on HN last year and have been meaning to try it out.
Seems like your approach is a bit more "batteries-included" but I'd curious for your thoughts on the differences.
I'm building a macOS image editor -- no AI here, just Metal textures, compositing, blend modes, all that mathematical raster goodness. It's getting closer to a Mac App Store release and has a preview version on https://skullrocksoftware.com
Really cool - I love the retro look&feel of it. Like Photoshop 1.0
Is it a heavily customized SwiftUI app?
I am building nOne, the first subjective layer of the Internet, or more accurately, a permission layer underlying the engines/platforms/automation. Although it does pertain to AI, its more than that... it is a permission system for email, AI, and cognitive inputs (undisclosed automation, what you permit in your feeds, etc). It encompasses the entirety of the digital realm.
We have created a permission system within our social worlds... time to apply the same standards to the digital realm.
https://network1.site
i'm building a technology detection / security / licensing scanner tool with deep inseption: https://stackaudit.heckstet.com/
a full sample report: https://stackaudit.heckstet.com/scan/594?token=370cd52a-8b7f...
I have realized that playing board games with other people makes my mental health better and started work on a community platform and at some point physical space to connect: https://GamesAgainstLoneliness.org
Also working on a journaling app but it’s still in the works.
Do you worry people might avoid your site/shop due to stigma around loneliness? I am curious is people might end up being embarrassed to be around because they would get designated as "lonely". If you find that people are slow to sign up, you could potentially do some A/B testing and advertise the same idea with a name that is related but feels like it has a more "positive" twist to it?(Could be called "Ludo League" [Ludo being the Latin root for 'game'] or something. Sorry I am crap at coming up with decent names lol)
Also, since you are in Germany, have you considered going to the next Spiel Essen board gaming convention? I haven't gotten the chance to go yet, but I imagine it would be a good way to make a lot of connections and get your gaming group known to a bunch of people.
Also, boardgamegeek/rpggeek are places where you could get your name out there(just in case you don't know about those).
Lastly, not sure how common board gaming shops/communities are over there but they are quite popular and common over here in the US so if there is any extra stuff you need to know regarding gaming shops you could probably get a lot of useful info by googling (random big US city like New York City) + "board gaming shop". Star City Games is a super popular one here and they have a high quality website that might be worth studying.
Best of luck from an aspiring board game designer!
I’m not particularly worried about any stigma or anything. So far most people understand that it’s not “yet another another gaming club” but a community/place that is also for this particular purpose. Worst case scenario I can rebrand it later into something else.
love the website, but sounds like the community mostly needs places to actually go be together, feel like maybe the website and members thing could have happened after maybe?
I’ve just released v2 of https://kintoun.app which is something I’ve been working on for quite a while now.
It’s an iOS client for Cloudflare and it covers a lot of resources with this last release.
Next bit of work is to clean the swift sdk a bit and make it open source, it’s been heavily inspired by the python-cloudflare sdk.
I'm creating my first game!
- Janky screenshot of progress so far: https://i.imgur.com/4afs5lv.png
- 2D single player browser game
- infinite procedural generated world
- build your starship
- manage a crew
- explore, harvest, trade and plunder the universe
- Frontend: Phaser 3 + WebGL + TypeScript https://phaser.io
- Backend: Workerthread + EliCS + TypeScript https://elixr-games.github.io/elics
- I made a Discord for it and I post daily videos of progress: https://discord.gg/FZa6w2TP
- I might open source it but since I play to make this a commercial game, I'm not sure yet. But I'm glad to help anyone with similar projects if I can
My challenges with it are:
- Keeping FPS above 60
- Art. I'm very bad with art
- Having something playable within 20 days
Imgur lies and gaslights, please don't use them.
While using a VPN, they will say "Imgur is temporarily over capacity. Please try again later." But if you don't use a VPN, then it magically works.
Sorry about that. I'll search somewhere better to post it while I don't have a website for it.
Meanwhile I found this random image upload website which might work:
https://freeimage.host/i/image.B0iOHxt
Maybe over the capacity allocated for VPN originating traffic? I would never interpret a user-facing error message like that on a public site to be a reliable claim of what's actually going on in the underlying infrastructure. For any service.
That's very charitable. I can't imagine that they would segregate bandwidth like that. I've never had Imgur work on VPN no matter the endpoint or time of day or week.
Well PocketBot won't be AI related in a few years. We're using AI with it's only job being to eliminate itself. Automations, using a single simple prompt, fully private, with a massive user created library (~1000 testers so far). People create automations - our LLM makes a template out of it -> goes to our shared user library -> when someone asks for an automation we check if it already exists and if so then deterministically adjust it for their use. If no automation exists then request sent to our Tier 2 LLM which creates it. Maybe not exactly what you wanted to hear, I am just excited about our idea...
Edit: Sorry missed the most important point. So the more users and the bigger this library gets, the less LLM usage we have, until eventually we are able to run almost fully deterministic.
A crossword puzzle generator, just for fun. Grid generator in Python because it's easy to hack around and grid generation doesn't need to be particularly fast. Go for populating the grid with words because with large grids there are combinatorial explosions, and Go's speed is beneficial.
Not source-available yet because it's a bunch of hacks (particularly the Python) but maybe one day.
I've built Bubbles, a reader-curated frontpage for discovering articles from thousands of personal indie blogs. No AI, no algorithm, no signup. Auth and comments are handled via Fediverse handles. https://bubbles.town
Broke my scapula and had never built myself (in Product, not dev) So decided to try to use Wispr Flow and keep productive since I couldn't type.
Built an open source chrome extension to build reports to provide back to our dev team. I promote it as AI related but it's not really. It just lets me grab elements on local files visually, comment on them, and provide them to our devs as a structured brief. There's others on the market but it wasn't worth paying for something so simple so I just made it free for all!
getmarkup.dev https://github.com/noahomrilevin/getmarkup
I'm working on a focus / pomodoro timer again that I started a few years ago.
The redeeming qualities, are that it doesn't require login, and you can just use it as a simple pomodoro timer that syncs to youtube / spotify music.
The additional features are all optional. Configuration, timeblocking, task management, etc... The features that are paid are just 1 (relatively low) price (with PPP per country). No subscriptions.
I started this project after pomofocus added giant ad banners, and had been missing some quirky things I had always wanted in a timer.
https://tomatillotimer.com/
I built and maintain an email service that has a no ai policy (https://sciactive.com/2026/01/21/our-stance-on-ai-in-email/):
https://port87.com/
Also, all of my open source projects (https://forge.sciactive.com/sciactive) use my SciActive Human Contribution Policy (https://sciactive.com/human-contribution-policy/), which bans AI contributions.
I’m building ATProto Alternatives (https://atprotoalternatives.com).
I started looking into AT Protocol recently and find it very interesting, so I started collecting a list of decentralized products built on top of AT that are alternatives to mainstream (popular) products.
This is very cool! I too have been very curious about ATProto after reading Dan Abramov's blog posts on it: https://overreacted.io/a-social-filesystem/
One question, I tried sharing your site with a friend and the share URL is coming up as http://localhost/ which is strange and I've never seen before. Is this something that's possible to customise via meta tag, or is this a bug in my browser (Vanadium 147 on GrapheneOS)?
Seems like accidentally they uploaded a local development build of the HTML:
<meta property="og:url" content="http://localhost/"/>
<link rel="canonical" href="http://localhost/">
Yeah thought it might be something like that, I just couldn't recall off the top of my head which one would get passed to the share context.
This should be fixed now!
Definitely something on my end. Thanks for the heads up! I’ll try to fix it tonight.
I’m reimplementing the Nix parser. I don’t know if it’ll morph into another formatter, or an LSP server, or what. I guess I’d like to try and improve on error messages and build it from scratch. Rust has really brought ergonomics to systems programming, and I wrote very few parsers for real languages, and never one with good errors.
I made a tracker for delayed-source releases (BUSL and FSL licenses): https://captnemo.in/finally-foss/
Might add JSON feeds, RSS for now. Main goal was to validate if anyone is using these releases 2 years later (forks?), and the answer seems to be no. The FSL premise for delayed releases is primarily around risk mitigation (if company goes bankrupt or bought out), and that is also yet to be tested. Most FSL/BUSL forks are at the moment of license change (OpenTofu for eg).
As another datapoint, Oxide is maintaining their fork of Cockroach, but has ignored all future releases (that became FOSS).
MusicLibrarian, a macOS app that deep-cleans Apple Music libraries. iCloud sync duplicated my playlists up to 540x (77,542 ghost entries). Built it in Swift, uses ITunesLibrary framework + AppleScript. https://thevaultdj.com
https://behavior.today
I'm building this site to display behavioral therapy data better for parents / caregivers. The site itself is just Next.js but the essence is ETL and data visualizations using the Observable Framework[1]. There is an opportunity to apply AI for Q&A.
[1] https://github.com/observablehq/framework
I'm building a rss client with an extra layer to have user comments/threads. This lets you create your own feed of articles entirely but with a social aspect.
I want it open source and free, just building an app that I'd like to use myself.
I built the exact same thing 2 years ago, and then figured out that Inoreader actually already had it in their product. They are disabled by default, probably because this feature is not popular at all.
I've dreamed of something like this with a geographical bounding (hard part) for what comments/requests you encounter.
I'm going with my fourth (or fifth) attempt to create a digital twin of my apartment in a game engine.
This time it's in Godot and i want to add more interactive stuff in the virutal version, rather than just redraw the rooms with some lights synced with Home Assistant.
This is really cool
I'm building a website integrity and security monitor. The backend is written in Java/PostgreSQL. The front end is written in JS/React. It will allow for interactive use via front end or API driven.
I initially was using SSE to push events down to the front end during long scans but decided to switch over to plain old HTTP polling for better reliability across different browsers.
Going for a 100% self-service model. No corporate sales cycles, no slide decks, no meetings. If the tool works for you, you buy it.
Targeting a June launch.
Built/building godom, a framework that lets me build local apps in Go, with the browser serving as a dumb view layer. I don't hate JavaScript or React, but my primary motivation was to eliminate NPM as much as possible. https://github.com/anupshinde/godom
I used AI to create the first POC, and once it was proven, it was improved, and AI handled a lot of grunt work where it could. The framework was built primarily to solve my pain points
And building Fractiz, a customizable pre-coded backtests platform.
A non-profit identity service.
Uses government IDs that a lot of people already have and saves signed credentials to your phone. There is a server element for the verification process but it runs all in-memory (follow the Mullvad model and not even have persistent storage in the servers).
It's fun. Get to practice mobile development (Flutter), use some local GPUs, learn about the changes to JWTs for signed, selective disclosure, Bitstrings for progressive disclosure lists etc.
Landing page is https://agora.gdn/ in case anyone wants to try a beta in the future.
I love watching YC Videos. So i vibecoded an app which found interesting clips from YC videos and show in like tiktok format. I get new content every other week, so makes scrolling fun again.
Named it Rocketship (LOL).
I have an invoice/order taker https://www.bestsellerapp.net, it syncs to many ERPs and allow people to operate everywhere, that I'm upgrading to be a more general mini-ERP.
On the side, retaking https://tablam.org, that I have procrastinated because wanna provide a parser that work for editors (that is way harder than normal batch parsers!)
I'm making more and more art. I'm meeting friends, and we sit together and doodle and paint. I've made so much progress in a few months simply by practicing more, experimenting more, and allowing myself to waste paper and make mistakes. A few friends have framed my art.
It's nice to do something away from the screen, without any goals besides having fun.
Very wholesome! I find that using cheap paper, and basic supplies from the dollar store really gets rid of hesitation when making art. After all, you're definitely not "trying seriously" when you use low-end supplies...
I've got a few things on my plate that I bounce between building for fun:
1. A point-and-click adventure game: making it with incredibly heavy technical constraints "just because"
2. A coding puzzle game of rapidly escalating difficulty
3. Part of #2 had me needing to craft some JSON Path queries and I felt like there wasn't anything nice to build and test them with, so I built this tool for it (inspired by the amazing regex101): https://jsonpath101.com/
4. A website where I write about text-based browser games
Also working on a Korean guide, so far it introduces Hangul (한글), the Korean alphabet: https://tolearnkorean.com/
Also making an app / web app for memorizing Korean words, takes inspiration from Anki and Duolingo. Words go through 4 stages: 1) matching up pairs 2) Multiple choice answer 3) Writing word through blocks 4) Free-form writing.
It's testable here: https://game.tolearnkorean.com/
Feedback is very welcome.
I'm scratching my own itch: building a yet another audio signal generator for smartphone. I need some extra functionality that is not available elsewhere, and impose some limitations "just because": it must be bare minimal PWA.
But actual app does not matter, the main take away for me is: it is easy and fast to write bloatware (esp. with AI), but not that easy to distill to what is really needed. And what looked like a weekend project, a couple of hours max (with help of AI), now lingers for 2+ weeks on-and-off on evenings (of manual effort).
“Ask HN: What are you working on? (April 2026) (Non AI)”
https://news.ycombinator.com/item?id=47679021
18 points | 1 day ago | 37 comments
Whoops didn't know about this, should have searched first
Working towards building a network fault attribution system oriented towards NaaS platforms that manage their own hardware. In the experimentation phase with existing monitoring frameworks, anomaly detection, and fault localization for sub-optimal routes.
Possibly see a use case in home labs too, but not there yet.
A family. Hope to have a little one in the next year! Wish me luck!
Same here! Good luck
I recently built Cranki [0], a free little PWA that generates crosswords using your Anki flashcard lists. It's aimed at language learners (who find flashcards boring and crossword fun!). To be honest I built it just for me and then decided it might be useful for others.
It's all local, no server, no database, etc. Mobile and desktop friendly.
I did build it with the help of AI, but no AI inside the actual thing.
[0] https://cranki.app
I am building a web application for learning math. I want it to be something between Khan Academy and Math Academy. Here is a demo of fourth grade https://demo.numerikos.com/ Currently the best part about it is one of my kids is using it. I have some more lessons ready, waiting to be released and I am currently working on Trigonometry.
I made this one-page little tool to help me split bills when going out with friends:
https://trianguloy.github.io/githubPages/SplitBill/splitBill...
The UI is horrible, but I really liked doing the coding. I'm also aware of other similar sites, but this one contains the features I need.
A few of us are working on trustworthy tech: https://trustworthy.technology/ and looking for folks to join up at zulip.
Ultimately, we’d like to get the main FOSS hardware, software, and services into a single store. Maybe we should put AI in the name. ;-)
I have been working on two opensource tools:
https://dhuan.github.io/mock/latest/examples.html
^Command line utility that lets you build APIs with just one command.
https://github.com/dhuan/dop
^JSON/YAML manipulation with AWK style approach.
Recently started working on my app for self-tracking again, a customizable combination of habit tracker, health journal and diary.
It should be as powerful as a spreadsheet for self-tracking, but the daily usability should be more on par with a habit tracker app.
The website and waiting list for anyone interested: https://dailyselftrack.com/
https://floracarta.com/
Flora Carta, design and keep track of your gardens. Originally a project for myself to keep track of every rose variety (I have over a hundred) in my garden, and apple/pear/plum variety in my orchard.
I've been working on a super-straightforward volunteering platform -- https://handsup.barryvan.com.au/ . Mainly because I got frustrated with my kids' school and sailing club activities being (mis)managed across WhatsApp, Google Forms, pieces of paper, and the like.
Building automated analysis tool that help identify the use of GPL-licensed SDKs in mobile apps, promoting license compliance and supporting sustainable open-source development.
I'm still working on my PDF search engine as an offline desktop application. It's at https://www.docgoblin.com/ if you're interrested.
No AI at all just plain old Java, JavaFX, Lucene and enough code to bring them together. I would love to get feedback if you try it!
Knowii, a community for knowledge workers who want to excel and thrive. We explore the frontier of knowledge management, personal organization, zen productivity, clarity, personal development, project management... And AI
https://dsebastien.net/community
Every time I have to open Acrobat it makes me want to puke, and Bluebeam is quite expensive, so I'm working on a full-featured PDF editor you can run in the browser.
https://DraftPDF.com
I am building something similar to Windows Task Manager but it also works on Linux, launches/monitors web servers, and provides a visual dashboard for docker containers.
https://github.com/prettydiff/aphorio
This project makes no use of AI.
I'm currently working on a small tool that helps me setup new personal computers in a purely descriptive way.
A little bit like ansible, but then totally not like ansible.
It generates SSH-Keys. It clones repositories, installs uv and rust. It removes Snap from my Ubuntu machines and installs firefox from the mozilla repositories.
Like nixOS?
I'm building a photo app specifically because AI sucks at helping me find photos in my camera roll.
https://picpocket.io
1. Reviving an old project that lets users sync their google docs as blog posts.
2. Writing a rich text editor library powered by pretext for cheap pagination
3. A layout engine that understands html/CSS subset for lightning fast pdf generation
Although AI is the main reason why I'm able to work on all these projects concurrently.
iCloud-photos-downloader is unmaintained and starting to crack due to Apple API changes so I’m working on my own fork/rewrite/take on iCloud sync, with ambitions to add Google Takeout, Immich, and other services too.
https://github.com/rhoopr/kei
Named after my ‘96 Subaru Sambar kei truck, small and mighty. The other thing I’m constantly tinkering with :)
I’m building technographics for desktop apps at https://DesktopInsights.com (similar to BuiltWith.com)
Just crossed 15k Mac and windows apps analyzed!
https://findsubstack.com
I'm building this mostly to scratch my own itch.
A newsfeed for Substack posts from the past 24h. Its helping me discover writers other than just what the algorithm gives me.
I’ve been working on a site that makes it easier to follow the Postgres mailing lists, which can be a bit of a firehose
https://pginbox.dev
I wrote a book about software engineering mental models that barely mentions AI on purpose. Only in the Appendix section I explain the rationale for not talking about AI, pasting it here:
---
Thoughts on AI
I consciously decided not to talk about AI throughout this book. Not because I don't believe in the benefits of using AI... I do.
I believe AI will keep bringing a lot of value to society.
I believe AI will keep changing our profession in many profound ways.
But whatever happens, I believe the principles here are still going to be incredibly valuable, even if the software engineer profession ceases to exist with its current name.
We might be working at a totally different level of abstraction, but the values, principles, mental models, patterns of communication and behavior described in this book will still make a huge difference. I believe they are atemporal.
Having said that, here are my recommendations regarding AI:
- Think of AI as another tool you have to create value, just like your IDE, the code you write or the emails you send.
- Keep yourself long enough in the problem space before throwing an AI api to solve a problem that might not exist
- Unless you are working at the frontier of AI development, ignore the noise. AI is now the shiny object everyone wants to be up to date with. JavaScript frameworks were used as a joke between value-driven software engineers because there would be a new one every 6 months and a lot of people would move to it without the real need. AI changes and improves every day or week. Being on top of it is probably a full time job. Instead of following all the news on a daily basis, level up your AI game in batches every 3 or 6 months. It will be more than enough.
- The most interesting use of AI for me has been in finding out my unknowns-unknowns. One question I frequently ask is "What are the building blocks of this problem/system/piece of knowledge?"
This discussion is specially for not talking about AI.
fair point =)
New programming language.
Took some good ideas of Pascal and making it more modern. Minimal runtime, manual memory management, single (small) executable, no dependencies. Compiler itself is written in Swift and I am using QBE as a backend ATM.
I'm attempting to create a tool for the manual curation stage of genome assembly in Zig. Something like juicebox, but with a browser based UI.
Thanks for this thread. I'm so tired of reading about yet another "AI wrapper for sending emails" raising a $5M seed round that I almost forgot what actual engineering looks like. The whole industry feels like a Gold Rush right now where 99% of people aren't even trying to mine gold, they're just reselling each other the exact same shovels (LLM APIs) with different logos slapped on them. It's incredibly refreshing to read about people writing Rust parsers or soldering hardware synths
I'm building the CAD app I have always wanted. Free, Parametric, STEP export for 3d printing. The input is just a javascript editor. It uses OpenCascade under the hood. Lots of fun.
I have been building cross platform QT desktop apps with PureScript and NodeGUI. PureScript is truly underrated as a language and as a gateway to the entire JS ecosystem!
Still working on my two decades old forum. Still going strong. https://www.tractorfan.app
visited expecting some bedroom DJing topics, stayed for actual agricultural machinery :)
I've been building the games I've always wanted to play. It used to take months to make something like Medieval Building Simulator.
Now you tell some idea of what you want and AI gives you a better thing than what you originally imagined. Then you go into a rabbit hole... simulate productivity, sales, customer, cash flow, materials for the building, make it curvy, make it pop up in 3D, populate it with NPCs, give your NPCs wages and background, write a dialogue with the ghosts of Christmas. It ends up being a game in itself.
It's good fun. It costs me $20 and I can do it and deploy it from my phone.
i am building a chaos testing CLI for distributed node systems in Go, started it because every time I wanted to test how a consensus protocol behaves when nodes start dying, I ended up writing throwaway bash scripts, Got annoying fast.
I’m building a home woodworking workshop—hands-on, slow and completely non-AI. It’s therapeutic and a nice escape from all the AI hype.
Hell ya, so many different ways to make it fit. Shops rule. Do you have a good idea of projects you want to do?
reverse engineered Elektron Octatarack binary data files and now writing libraries to interact with/modify them.
https://gitlab.com/ot-tools/ot-tools
i’m currently avoiding committing/releasing a bunch of changes i did last week because people are actually using the library already (the curse of writing something useful lol)
A shopify app to sync witg Google Sheets. https://shopsheets.co
https://SupplementDEX.com - helping people find out if their supplements work
watgo (https://github.com/eliben/watgo) -- a WebAssembly toolkit for Go. Think something like wasm-tools or wabt, just in pure (0 dependency) Go.
I have gathered lots of public datasets and wrote a service that analyzes a car brand.
i made my own personal website [0] that loads in less than 512kb; with a mini bio, a blog (with RSS support), images, links to friends' websites, stuff i made/maintain, and some fluff[y boys] like a walking Ralsei [from Deltarune] that follows the cursor, a music player, and some secrets/easter eggs ;)
then there's another project i have, which is a download-and-run version of Minecraft, where you just download one binary (most likely an AppImage) and can run the full game. it will be for personal use so i shouldn't worry about copyright.
i'm also making my "own" minimal tiling window manager for Linux [1] (which is a fork of [2]) with my custom keybinds and [planned] controller (joystick) navigation support.
and lastly, i made yet another fetch software: jotafetch (*jota*lea's implementation of neo*fetch*). it should be available to read and download a on my website [3] (i haven't packaged it for mainstream distros yet, and i probably won't).
[0] https://jotalea.com.ar/
[1] https://github.com/Jotalea/jwm
[2] https://github.com/173duprot/wm
[3] https://jotalea.com.ar/files/jotafetch
A Chrome extension that helps you efficiently declutter stale tabs.
Since the moment "AI" became a thing, I've been rolling my eyes and looking the other way since I know it's just a half-assed hype buzzword. With that in mind, pretty much everything I work on is not AI related. As a hobby(and to a degree to be prepared for WW3), building a ton of drones and drone equipment, flying them and finding more and more ways to push their range to the absolute limit.
I don't think it qualifies as AI in the modern day and age but NLP in general. It's truly amazing how easy it is to spot troll farms online and no one is doing anything about it, be it individuals, private sector or even on national level, given that those should be considered a risk for national security.
Solo project since 4+ years. Zero AI, but algorithms that will probably make people believe otherwise: https://kastanj.ch/en?mid=hn47700460
The goal is to make every recipe foolproof on the first try, similar to when you walk into a restaurant and just pick what you want to eat without thinking about the details. The goal is to have the same experience, just pick what you want to eat, with recipes that tells you exactly what to do with no magic involved.
Technically it is probably the most advanced recipe app ever made. The database is a huge graph that captures the relations between ingredients and processes. Imagine 'raw potato'->'peeled potato'->'boiled potato'->'mashed potato'. It is all the same ingredients but different processing. The lines between the nodes define the process and the nodes are physical things. Recipes are defined as subsets of the graph. The graph can also wrap around into itself, which is apparently needed to properly define some European dishes in this system. The graph also has multiple layers to capture different relationships that are not process related.
Why was it designed it in this way? Because food/cooking is extremely complex. This design is the only way I have found that can capture enough of these complex relationships that the computer can also 'understand' what is going on.
My favourite thing about this is that each recipe is strictly defined in the graph. If the recipe skips a step, or something is undefined, the computer knows that the recipe is incomplete. It won't ask you to do 10 things at the same time and then have something magically appear out of nowhere. It is like compile time checking but for recipes.
It also enables some other superpowers, for example:
• Exclude meat part of the graph = vegetarian. Same thing works with allergies.
• Include meat part of graph = only show me recipes that contain meat.
• Recursive search: search for 'potato' and the computer will know that french fries are made from potato. It can therefore tell you that you could make the hamburger meal, but you will need to complete the french fries recipe first, which should take 60 minutes.
• Adjustable recipe difficulty (experimental): It knows which steps can be done in parallell, and which can't based on how the nodes connect. A beginner can get a slower paced recipe with breathing room between steps, while someone more experienced can do a faster pace and do more things in parallell.
If I knew what it would take to build this, I would never have gotten started. I completely underestimated the complexity of the problem I was trying to solve. But here we are, and now it is basically done and working.
The website is slightly outdated but captures the key points from a non-technical point of view, and you can enter your email and get notified when it will launch in your country.
Now that gasoline prices skyrocketed I developed https://pumperly.com in selected countries where data is publicly available
I'm building a no bs esign product, with api access by default. At a previous company, automating this stuff cost us an arm and a leg and we got locked into a bad contract. I've always been on the engineering side of things and never had the opportunity to learn how to find customers and sell. It's been a fun and rewarding experience for me. https://www.sign-n-post.com/
I am building a 3-way diff editor: https://codeinput.com/products/merge-conflicts/online-diff
https://www.adashape.com/ - a humanistic CAD for 3D printing. Early but has profile editor, extrusions, booleans, and imports STL, STEP, SVG and bunch of visual mesh formats.
This seems a little bit of a duplicate of [1], but I can repeat my answer here as any views help!
My wife and I continue to work on Uruky, a EU-based Kagi alternative [2].
Since last month we finally got our production API Key for EUSP/STAAN (it was certainly the slowest and most complicated search provider to adopt, so far), and that brought us to 5 search providers you can choose from and sort as you prefer.
We already have got over 40 paying customers (excluding family and friends, we’re guessing these paying customers came from some privacy listings and HN comments) and have exited beta last month!
Customers seem to really enjoy the simple UI (search can be used without JS) and search personalization (from choosing the providers to the domain boosting and exclusion). We also have hashbangs (like "!g", "!d", or “!e”) when something doesn’t quite give you what you’d expect, though.
You can see the main differences between Kagi and Uruky in the linked page, but one huge difference is that with Uruky, after being a paying customer for 12 months, you get a copy of the source code!
One thing we’re struggling with is outreach because we want to do it ethically, and it’s hard to find communities or places to sponsor which are privacy-focused and don’t require €5k+ deals. Ideas are welcome! Because of bots there isn’t a free trial easily available, but if you’re a human and you’d like to try it for a couple of days for free, reach out with your account number and we’ll set that up!
Thanks.
P.S.: Because people have asked before, our tech stack is intentionally very "boring" (as in, it generates and serves the HTML + bits of JS to enhance settings and such), using Deno in the backend (for easier TypeScript), PostgreSQL for the DB, and Docker for easier deploying.
P.P.S.: Because this has been also brought up before, the name has no special meaning but we read it like "Euro-key" in English. Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
P.P.P.S.: Another frequent question here is “how does it work?” When you search, we query the first search provider on your list, and if it yields less than X results (only Mojeek really gives us a total count, we have to try + estimate for the others), we try the second, and so on. We then merge the results in a round-robin fashion (first of first, first of second, second of first, second of second, and so on). There’s a bit of more nuanced logic to also properly rank the results with the pin/exclude/raise/lower preferences, because it works differently across providers and not all of them support that, for example.
[1] https://news.ycombinator.com/item?id=47679021
[2] https://uruky.com
Regarding the bots: since you're building a privacy-first product, you should look into a Proof-of-Work captcha (like Hashcash or mCaptcha). Just have the user's browser mine hashes for a couple of seconds before issuing the trial token. A normal human won't even notice it, but it'll burn so many CPU cycles for bot farms that abusing your API becomes economically unviable
Thanks for the suggestion, I’ll look into it!
I love the idea! It's so good to see people advancing the european tech landscape. Just out of curiosity: Are there more reasons for building Uruky than Kagi being located in the US and closed source ?
Those are the main ones, but also something more focused and not going into other tools or AI as it is right now.
> Names are hard, and we’re aware it can remind people of Uruk and Uruk-hai. That’s OK.
What I would worry about is if it reminds whoever holds Tolkien's copyrights of Uruk and Uruk-hai.
That’s a good point, thanks for mentioning it. I think we’re safe since Uruk can’t be copyrighted as it’s an ancient city’s name and the name of a time period, and there are no intentional Tolkien references on the product.
rawfeed.social, for the people who loves RSSs.
Builsing my web dev agency aurevow.com (@aurevow.com on x)
I recently made a 'notification-driven RPC' app for Android to help me help others:
Notifactor (Android App)
A lightweight native Android app that intercepts notifications on the device and triggers actions based on configurable rules.
The app uses Android's NotificationListenerService API. Once granted notification access it receives a callback for every notification posted on the device. It then checks each notification against configured rules and runs the configured action.
Why would you want this?
I don't know why you would want this but I can tell you why I made it: to make it easier to control some functions on Android devices used by people I often help using them in some way. My mother's Android TV (which I use to communicate with her through Linphone), phone and tablet sometimes stop doing the right thing. I live about 1300 km to the north of where she lives so I can't just hop on my bike to fix things. Thus far I relied on a set of Termux scripts on these devices to keep a reverse ssh tunnel open to an endpoint on my server but this has a number of drawbacks: the tunnel is not always there when I need it due to WiFi dropouts and other similar problems and the constant connection uses battery power on the phone and tablet. If only I could cause the tunnel to be created when I need it and brought down when it is not needed... Well, that is possible using Notifactor by sending a notification on a specific channel (ntfy refers to these as topics) whereupon Notifactor runs a Termux script which manages the tunnel (etc.).
Why would that be refreshing?