I want my real dev setup on my iPhone Duo. I mean the actual mac app for Claude Code or Codex with a local session. I don't even need to type we have great dictation nowadays.
I know you can already do something equivalent right now with the iphone app + cloud agents. But let's be real it's not the same.
I want my agent to be able to run a real linux terminal on my MacOS based iPhone.
BIG NEWS!
My ~8 hour course on AI data centers is now live on @freeCodeCamp!
From Watts to Tokens: 16 chapters covering grid power, 800V DC racks, liquid cooling, networking and GPU cloud economics.
https://t.co/Bolpv7M1se
I think most people are trying to larp anything that is an agentic loop as RSI.
The bar is low because for 99% of experiments, RSI is actually not working. We still need some human element to steer the LLM loop, if we reached RSI in any significant way we would have accelerated way more.
@poteto from my testing on DeepSWE, adding poteto mode usually just double my token usage + latency without actually moving the needle at all in term of output correctness
there’s no easy eay to do this right now with current toolings, but i’d suggest you to test blind a/b testing with the rest of sota models. I think it’s really great at explaining concepts and it’s nice to have a third option when the two other models are having issues (opus having a sloppy output and astra eating your tokens hyper fast)