Bari Be

i felt this earlier than expected. not because we had too many features, but because small decisions started needing old context: why we built something, what broke before, what users actually asked for. for us, the hard part is keeping the chain of thinking visible. once you lose that, even simple changes feel heavier than they should. i don’t think tools should make complexity disappear. they...

tbh i don’t think i can reliably tell anymore. the small signals people mention, dashes, perfect grammar, weird structure, are already easy to fake or remove. for me the more useful test is: can i trace the claim back to something real?

I think solving your own problem is probably the best starting point, but not enough by itself. For me, Theoria started from a very personal frustration: I kept opening papers, abstracts, PDFs, related tabs, and somehow ending up more confused than before. The first version was basically just me trying to make research feel less heavy: find papers, explain them in plain language, and let me ask...

The line “we were chasing a dream, not data” hits hard, because I think a lot of builders quietly do this. Enterprise sounds like the “serious” next step, but the boring requests from real users are often the actual roadmap. Also love the rule at the end: no enterprise features until someone pays first. Simple, painful, but probably saves months of building the wrong thing. appreciate the post

My biggest frustration is not that the tools write bad code. It’s that they write kinda confident code in the wrong context. For Theoria I use a mix of AI coding tools, mostly for moving faster on React/TypeScript, backend routes, database logic, etc. And honestly, the speed is crazy. But the annoying part is when the agent understands the task, but not the history of the product. Example: it...

The moment was more internal: I looked at the MVP and realized it was a cool demo, but not yet something I could really stand behind as a product. With Theoria, the first version could find papers and explain them in a simpler way with basic features. That was enough to prove the idea. But then I started adding things that made it feel more real: saving research, adjustable answers, audio...

IMO free should show the value, paid should unlock serious/repeated use. For Theoria for example I keep the basic experience free: people can find research, read simple explanations, and see if it’s actually useful for them. The paid part starts when someone wants to go deeper or use more expensive AI features again and again, like asking detailed research questions, getting deeper paper...