The tools we build on, reviewed honestly.
We ship on these every week, so these reviews are about what the tool is like in production rather than what its landing page claims. Each one says where it breaks and what we reach for instead.
Bubble
Real products ship on it. Watch the workload, not the drag and drop.
Read →
LangGraph
Agents that survive a restart. The state model is the product.
Read →
OpenAI API
Easy to call, harder to run. Rate limits decide your architecture.
Read →
Supabase review
Worth betting a product on. Row-level security is where teams get hurt.
Read →
Claude Agent SDK
The agent loop, already written. Permissions are the reason to use it.
Read →
LangChain
Useful for the first month. Judge it on what you keep after.
Read →
Make.com
Fast to build, and it discards failed runs unless you say otherwise.
Read →
Stripe
The default for taking money. Marketplaces are where it gets hard.
Read →
n8n
Good enough to run a client's operations on. Self-host it.
Read →
Vercel
Deploys you stop thinking about. Know the limits before you hit them.
Read →
