Knowledge retrieval. Something where I could ask a question, say in natural language, and I would get an answer. Of course the responses have to make sense and not be made up.
I opened my Obsidian directory with Augment Code, and it indexed it in seconds. Now all my written knowledge is retrievable. You could probably get the same effect with any other RAG-based indexed agent
Tried a few workflow tools but most either break often or need more setup than the task itself. Would easily pay for something that just quietly handled all of that in the background.
Mostly a mix of both. Some are dev related workflows while others are more operational like content pipelines, research tasks and internal automation stuff.
I still manually synthesize information from multiple sources(docs,blog posts and treads).LLM help with sumaries but merge into a reliable decision step is very human-heavy.
If you are selling to ‘enterprise’ that business model might work well as a per-seat price.
At retail, into small bespoke niches a business built on $20 problems is unlikely to be sustainable because it is not enough money to reliably reach small markets and it is not enough money to sustain high quality service to a small number of customers…sure 1000 customers would be $20,000/month, but you have to get there first and stay there second.
Getting there is probably more money and calendar pages than you think and staying is going to gobble up more of the $20,000/month than you want through churn and employees.
My advice: find some thing you want to work on or find actual customers who will pay you handsomely to solve their important problems. Good luck.
I have nothing that fits that criteria exactly. I still write tests for test automation manually and that should be automated but new tests are only needed as features are created or retired, which is irregular and infrequent.
I hear you. I also don't buy clothes unless really really needed. But I am hoping that if I do buy a Robot like that, it will do other things as well and not just folding laundry :)
Knowledge retrieval. Something where I could ask a question, say in natural language, and I would get an answer. Of course the responses have to make sense and not be made up.
I opened my Obsidian directory with Augment Code, and it indexed it in seconds. Now all my written knowledge is retrievable. You could probably get the same effect with any other RAG-based indexed agent
Tried a few workflow tools but most either break often or need more setup than the task itself. Would easily pay for something that just quietly handled all of that in the background.
"Curious what kind of workflows you're talking about – are these dev-related tasks or more general business processes?"
Mostly a mix of both. Some are dev related workflows while others are more operational like content pipelines, research tasks and internal automation stuff.
For dev workflows, Augment works smoothly for me, and I've heard the same about Claude Code and Codex. What specifically breaks?
The recently announced Cosmos Agent OS is built to handle everything explicitly in the background, but it's more for enterprise than for individuals
I still manually organize my gallery, especially screenshots and downloads
I still manually synthesize information from multiple sources(docs,blog posts and treads).LLM help with sumaries but merge into a reliable decision step is very human-heavy.
Do you think that can actually be automated? Seems like a judgement-heavy process to me.
Drawing vector graphics.
Image generators can make reasonable-looking raster images. LLMs are good at coding. But drawing SVGs sits at the worst of both worlds.
Drawing SVGs via LLMs is mid, but how about converting raster images to SVGs? That sounds like something that shouldn't be too hard
I thought they'd be good at vectorising raster image, but no.
You'd pay $10-20/month if something just worked
If you are selling to ‘enterprise’ that business model might work well as a per-seat price.
At retail, into small bespoke niches a business built on $20 problems is unlikely to be sustainable because it is not enough money to reliably reach small markets and it is not enough money to sustain high quality service to a small number of customers…sure 1000 customers would be $20,000/month, but you have to get there first and stay there second.
Getting there is probably more money and calendar pages than you think and staying is going to gobble up more of the $20,000/month than you want through churn and employees.
My advice: find some thing you want to work on or find actual customers who will pay you handsomely to solve their important problems. Good luck.
choosing and playing boardgames - hope it never changed
I have nothing that fits that criteria exactly. I still write tests for test automation manually and that should be automated but new tests are only needed as features are created or retired, which is irregular and infrequent.
Furniture design.
Give the measures of a bookself and get the pieces and materials needed.
I'd say if you could automate product sheet details with detailed image recognition would be really interesting
Folding my laundry
Some companies are building Robots for this already. for example https://figure.ai
Except that robot probably costs more than the prevent value of all clothes I’ll ever buy.
I hear you. I also don't buy clothes unless really really needed. But I am hoping that if I do buy a Robot like that, it will do other things as well and not just folding laundry :)
Ahahaha, good point! Some automation on that field would be precious.
Do the dishwasher and throw the rubbish out
preparing for a trip , travel, lots of moving pices like itenary , packing , tickets , IDs , cabs .no sync .huge mess
I believe this will get solved once agent protocol is adopted widespread.
Reviewing AI generated slop.