Empire Signal - an AI tool that teaches you how to USE a word, not just what it means
Hey IH — first post here, wanted to share something I just launched.
The problem: I'm a CS student in Havana, learning English for years. My bottleneck was never vocabulary — I knew the words. I just never knew how to use them: right register, natural collocations, when something sounds off in context. Every dictionary tells you what a word means. None of them tell you how to actually use it.
What I built: Empire Signal. You type a word, it breaks it down across ~13 angles — meaning in context, register, collocations, usage examples by tone, etymology, common errors, word family. You can also write your own sentence and it scores how natural it sounds (0–100) with specific feedback.
Stack: Next.js 16, Prisma + Supabase Postgres, a multi-model AI fallback chain on free tiers, deployed on Vercel.
The constraint that shaped everything: building from Cuba means no direct DB connection (migrations are manual via SQL Editor), unstable internet, and free-tier AI models that need real engineering to be consistent — a single model can win 95% of the time in a "race," so you have to actually log and verify what you're shipping instead of trusting the architecture diagram.
Where I'm at: day one. No users yet, no revenue, just launched. Plan is Reddit/IH for initial feedback, then iterate hard on quality before thinking about monetization (Lemon Squeezy, since Stripe isn't viable from Cuba).
Would genuinely love feedback — especially from anyone who's tried building something solo with real infrastructure constraints. What's broken, what's confusing, what would make this actually useful to you?
The part I found most interesting wasn't the product.
It was the difference between a problem you personally experienced for years and a problem other people immediately recognize as worth solving.
Those can overlap perfectly.
Or they can drift apart in ways that aren't obvious at the beginning.
That's a sharp distinction, thank you. Honestly -- right now it's still mostly the first one. I know my own problem deeply; I don't yet have strong evidence that it's the same problem at scale for others.
That's actually why I'm here instead of just shipping silently. The phrasal verb / collocation / register gap shows up constantly in language-learnings forums, so I have a hypothesis it's shared -- but hypothesis isn't validation. Curious if it resonates with you, or if you've seen this kind of personal-problem vs market-problem gap bite you before?
I've definitely seen situations where the distinction mattered more than it initially appeared.
Probably more than I'd try to unpack properly in a thread though.
Happy to continue by email if useful.
That works for me - happy to keep digging into this by email: cesarabrahamhernandezponce@gmail.com
Just sent you a note.
I think the distinction is more important than it first appears.