1. As other comments have already pointed out, the fact that you have AI plastered all over (including your demo video) does not bode well for the "real developers" claim. I would suggest doing some internal testing/surveys with real developers and get them in the demo instead. Best case, your product is a hit and people genuinely love it. Worst case, you learn about your own pitfalls.
The way this is priced and presented feels it's not a alpha/beta product. But, I need the confidence that it actually is not still in alpha/beta.
2. Why this and not Cursor/Claude/Zed? In fact, your slogan ("Code at the Speed of Thought") is a DIRECT rip-off from Zed (https://github.com/zed-industries/zed)
3. Why should I bother with the switching costs? When you mention "better understanding of your codebase", do you mean Cursor-style code indexing and context windows? Do you try anything new here?
Thanks, that's helpful. It's not in alpha/beta but definitely a new product with flaws and stuff we could have done better.
We actually tried to do a lot of stuff differently! it's 100% online, connected to Github, handle deployments and hosting and remain open to other platform with Docker.
I meant in comparison to other frontend products that can't do anything about your back-end. $200 for 2000 messages may seems a lot but it includes every tool use (even if the agent uses 100K tokens editing your code it would count as a single message), web deployment with hosting and everything.
Feel free to try it, the free tier is here for that!
AppJet is directly integrated with GitHub and has a better understanding of your codebase, with more full-stack functionalities. It also includes all the deployment of your code and a direct integration with Supabase that Claude or Jules doesn't have.
The agent itself is always limited by context inside a question but it also draw a map of your codebase and use it as additional data to better serve you.
I guess most ads are lying, but when that AI woman says, she now "gets stuff done". There isn't even a hint of believability anymore.
A real developer, or maybe even a customer would be more believable if they would actually use the product.
We made AppJet.ai for real developers, not just vibe-coders. AppJet helps you navigate your GitHub repository, including inside branches, fixes bugs and create new features. Let me know what you think!
FYI, the layout for the navigation component is broken on many smaller screen sizes. It's a small issue that I wouldn't mention normally, but it does undercut your messaging about the "future of software development" and "serious developers" when the first element on the page doesn't render properly.
Sorry I have to explain. By putting text that you wrote into AI generated video format, you are showing a lack of care for my time. if you have text just put the text there, a professional will be convinced by facts, Not by some artificially generated images of a lady suck puppet. hard pass.
AppJet.ai can help you in several ways with TLA+:
Language fundamentals: Basic syntax, operators, temporal logic, actions, and specifications
Modeling techniques: How to represent system state, define actions, specify invariants and temporal properties
PlusCal: The higher-level algorithmic language that transpiles to TLA+ TLC model checker: Running specifications, interpreting results, debugging models Common patterns: Modeling mutual exclusion, consensus algorithms, message passing, etc.
Practical examples: From simple counters to complex distributed systems
If your code is on GitHub you can use our app with it, it will create a safe independant branch "appjet" to work on it. Give it a try and let me know what you think!
> Common patterns: Modeling mutual exclusion, consensus algorithms, message passing, etc. Practical examples: From simple counters to complex distributed systems
I don't know how you composed this response. But it reads more like a laundry list of TLA+-related buzzwords than a response based on any kind of actual experience with TLA+.
TLA+ is a weird and obscure niche, easily obscure enough to give most models very serious headaches.
There it is. Like their "AI" companions, the people using them are unable to conceive not knowing something, so they might as well invent anything that sounds plausible, unable to even learn a new concept.
Donning-Kruger as a service. I guess that's what defines the 'serious developers' from the rest of us.
"AI" in shovelware products like this is a shibboleth that means either "I don't know what I'm talking about" or "I'm trying to ride the hype train before it kills us all".
Constructive criticism:
1. As other comments have already pointed out, the fact that you have AI plastered all over (including your demo video) does not bode well for the "real developers" claim. I would suggest doing some internal testing/surveys with real developers and get them in the demo instead. Best case, your product is a hit and people genuinely love it. Worst case, you learn about your own pitfalls.
The way this is priced and presented feels it's not a alpha/beta product. But, I need the confidence that it actually is not still in alpha/beta.
2. Why this and not Cursor/Claude/Zed? In fact, your slogan ("Code at the Speed of Thought") is a DIRECT rip-off from Zed (https://github.com/zed-industries/zed)
3. Why should I bother with the switching costs? When you mention "better understanding of your codebase", do you mean Cursor-style code indexing and context windows? Do you try anything new here?
Funny, reading that, made me think of Bill Gates' book, Business at the Speed of Thought: https://www.amazon.com/Business-Speed-Thought-Succeeding-Dig...
Thanks, that's helpful. It's not in alpha/beta but definitely a new product with flaws and stuff we could have done better.
We actually tried to do a lot of stuff differently! it's 100% online, connected to Github, handle deployments and hosting and remain open to other platform with Docker.
Well damn, I'm not a serious developer. I guess I'm not included.
edit: 2000 messages per month for $200 doesn't really seem worth it.
edit edit: even less so when JetBrains lets you plug in self-hosted models. This business model is headed for a crash.
I meant in comparison to other frontend products that can't do anything about your back-end. $200 for 2000 messages may seems a lot but it includes every tool use (even if the agent uses 100K tokens editing your code it would count as a single message), web deployment with hosting and everything.
Feel free to try it, the free tier is here for that!
Nice that it has a free tier. But how does this compare to Jules/Claude/etc? What makes it different?
AppJet is directly integrated with GitHub and has a better understanding of your codebase, with more full-stack functionalities. It also includes all the deployment of your code and a direct integration with Supabase that Claude or Jules doesn't have.
Isn't it limited to a context max token window? How can it handle the whole code base, not mentioning LOC or anything?
The agent itself is always limited by context inside a question but it also draw a map of your codebase and use it as additional data to better serve you.
> has a better understanding of your codebase, with more full-stack functionalities
Can you elaborate on what this means?
Presumably there is a custom harness on a SOTA model. The harness is what you are paying for. They call it "RobertoAI".
When even the demo video has an AI generated persona to explain the business.. you know it's vibes.
I guess most ads are lying, but when that AI woman says, she now "gets stuff done". There isn't even a hint of believability anymore. A real developer, or maybe even a customer would be more believable if they would actually use the product.
Curious to know how this possibly made it to the front of HN.
Upvotes from the serious developers.
Hi,
We made AppJet.ai for real developers, not just vibe-coders. AppJet helps you navigate your GitHub repository, including inside branches, fixes bugs and create new features. Let me know what you think!
Alex.
FYI, the layout for the navigation component is broken on many smaller screen sizes. It's a small issue that I wouldn't mention normally, but it does undercut your messaging about the "future of software development" and "serious developers" when the first element on the page doesn't render properly.
Yeah, mobile devices are not easy since there is so much to display for the app to be usable.
Real developers like the AI woman on your demo video? Come on....
yeah man, we're an AI company, we're using AI! is that so shocking?
Sorry I have to explain. By putting text that you wrote into AI generated video format, you are showing a lack of care for my time. if you have text just put the text there, a professional will be convinced by facts, Not by some artificially generated images of a lady suck puppet. hard pass.
I use TLA+ on a project that requires lots of compliance checks and tests. How will this product help me vibe-code the code base?
AppJet.ai can help you in several ways with TLA+: Language fundamentals: Basic syntax, operators, temporal logic, actions, and specifications Modeling techniques: How to represent system state, define actions, specify invariants and temporal properties PlusCal: The higher-level algorithmic language that transpiles to TLA+ TLC model checker: Running specifications, interpreting results, debugging models Common patterns: Modeling mutual exclusion, consensus algorithms, message passing, etc. Practical examples: From simple counters to complex distributed systems
If your code is on GitHub you can use our app with it, it will create a safe independant branch "appjet" to work on it. Give it a try and let me know what you think!
> Common patterns: Modeling mutual exclusion, consensus algorithms, message passing, etc. Practical examples: From simple counters to complex distributed systems
I don't know how you composed this response. But it reads more like a laundry list of TLA+-related buzzwords than a response based on any kind of actual experience with TLA+.
TLA+ is a weird and obscure niche, easily obscure enough to give most models very serious headaches.
I immediately thought this sounded like an LLM reply.
It reads like they asked the AI how it could help and just pasted whatever it said
exactly.
I stopped reading mid-sentence after seeing a cacophony of buzzwords shoved into that AI generated reply.
Very disappointed.
Well, I don't have any idea what's TLA+ is so I just asked the agent and that's it's response yep
There it is. Like their "AI" companions, the people using them are unable to conceive not knowing something, so they might as well invent anything that sounds plausible, unable to even learn a new concept.
Donning-Kruger as a service. I guess that's what defines the 'serious developers' from the rest of us.
"AI" in shovelware products like this is a shibboleth that means either "I don't know what I'm talking about" or "I'm trying to ride the hype train before it kills us all".