AI is becoming training data for AI.
I looked into model distillation, the controversy around Claude and Chinese AI labs, and new research showing models can even learn from their own outputs.
This rabbit hole gets interesting fast. https://t.co/Y4yGasDcHC
Really appreciate everyone who followed today. Met a lot of interesting builders already.
If you want to connect, just say hi below. Iโm following people back, but X temporarily slowed me down after I followed too many accounts too quickly.
Iโll catch up with everyone.
AI agents are getting good at doing work.
Iโm more interested now in what happens when they fail.
Can they explain what happened?
Recover?
Learn from it?
Avoid doing the same stupid thing tomorrow?
What are you building right now?
SaaS, open source, AI agents, mobile apps, dev tools - anything.
Drop it below. Iโm looking for more builders to follow.
Still under 100 followers, so Iโm looking for more builders to connect with.
AI, SaaS, open source, agents, mobile/web apps, dev tools - if youโre working on something, say hi and share it.
The more I learn about AI engineering, the more I realize the model is only one piece.
Evals, retrieval, tools, memory, permissions, observability, cost, latency.
Thatโs where a lot of the real engineering starts.
AI made it incredibly easy to start building software.
The advantage now is probably not who can generate the most code.
Itโs who can choose a real problem, stay focused, and actually ship.