forkabl
forkabl is an AI lab.
Forkabl is a lab running a series of experiments in how AI can safely percolate into the tools and workflows of its next billion users.
Principles and bets fork into experiments. Lessons merge back into the principles and bets.
Each experiment starts with a bet and builds something concrete to test it. The result either continues as a product or stops there. In both cases, the lessons update the shared principles and bets.
Principles
7- 1
Speed above all else. Great results come from sprints of focused, undistracted work. Protect the sprint and move quickly.
- 2
Follow your curiosity. When something pulls at you, go after it and keep notes.
- 3
Make bets, not plans. Plans pretend the future is known. It is not. State what you believe, what it costs to find out, and what would prove you wrong. Then run the experiment.
- 4
Fix the time, not the scope. Time is the only thing you cannot get back. Give every experiment a deadline and ship the smallest version that teaches you something.
- 5
Don't trust LLM estimates. LLMs learned how long things take from people who built everything by hand. That is not how software gets built anymore. Measure your own cycles and trust those.
- 6
End things well. Kill experiments quickly and write down what they taught you.
- 7
Stay independent. Answer to no one you did not choose. Users are people, not the product.
Bets
5- 1
Turn-by-turn is not the end state. Chat taught everyone how to talk to models, but taking turns is a constraint, not the destination. Models will watch, interrupt, and act without being asked.
- 2
Non-engineers are next for the agent magic. Engineers already know what it feels like when an agent does an hour of work in minutes. The next wave is everyone else.
- 3
Backends get discovered, then crystallized. Instead of designing an API first and building it after, the model finds the path by doing the work. The paths that survive get hardened into real APIs. A new kind of framework comes out of this.
- 4
Micropayments come back through crypto. The web always wanted to charge cents but card fees made it pointless. Crypto makes paying pennies per read or per call practical for the first time.
- 5
Creator coins need a selling mechanism. For creator coins to work, a creator has to sell small amounts steadily without crashing the price. Build the mechanism that lets them withdraw just enough to fund the next project, and coins become funding instead of speculation.