pull down to refresh

10 sats \ 1 reply \ @6404e30b28 18 Apr \ on: Opus 4.7 to 4.6 Inflation is ~45% tech

The bigger question might be whether costs are rising or if we’re just getting more capability per unit and calling it inflation.

write
compose
12 sats \ 0 replies \ @freetx 18 Apr

Well smaller local LLMs are getting much better as well. Frankly my use of SOTA models has dropped off over last 3 months, I get lots of work out of qwen3-coder-next and now qwen3.6 / gemma4

I don't have enough local horsepower to run Minimax-2.7 or Kimi-2.5 but I could basically see ditching Cursor entirely very soon and using MM or Kimi to write detailed plans, and then pass off to local qwen to implement.

I don't think SOTA models will ever go away, but at some point local LLMs are going to become "good enough" to seriously curtail their usage.

reply