This is going to have an opposite effect that the decels want. It's a huge open source AI accelerant, an accelerant for corporations, and enterprises
Now it's a real race. You either make your own AI infrastructure or you don't have a seat at the table
@saketkr_ Ha true. The open models are the thing I’m watching here. The day you can run something locally through Ollama as good as today’s Claude Code or Codex, it’s game over in the best way. No bills, no caps, no flip-flops!!
There are engineers right now asking AI questions they already know the answer to, just to run up their token count, because their company put it on a leaderboard.
I keep thinking about that. It might be the dumbest metric in tech, and it's finally starting to fall apart.
@KaiXCreator For well defined problems, Codex. Claude Code is the better explorer, but when the task is clear and I just need it solved, Codex lands it more often for me. Was stuck for ages in a long running harness with Claude on Maestro mobile testing, and Codex sorted it in ~10 min
@simonw Makes sense. What makes you read it as a value hint though? A flat cap feels more like a budget limit than a readout of what the tools are actually worth. Curious how you're seeing it.
The bit that stuck with me is that being efficient and being effective turned out to be the same thing here. Keeping the context tight is what makes these tools good and what makes them cheap, at the same time.
So now it's basically this. One project file I keep short. Small specific asks instead of one giant one. And a fresh chat the moment it starts drifting.
🚨🚨 38 balls 69 for Virat Kohli
- man, how the hell is this guy able to do it? He literally scores 50 for RCB in every 2nd game without getting tired!
- last 3 seasons, all 600+, here he comes & starts with a 50. He is the beast of highest order🐐