Post #3563957
2026-07-03 19:14 UTC
@peach@phpc.social The explosion in token costs that everyone is expecting flat-out won't happen because too much money is being dumped into making inference more efficient (because the hardware is so expensive and the profit margins on inference are so good). IIRC Junie runs at "API rates" which is as expensive as tokens will get, and e.g. OpenAI's newer stuff is cheaper than their older stuff for the same performance.
2/3
Replies (1)
-
@ian@phpc.social 2026-07-03 19:17
@peach@phpc.social LLM viability for dev would *absolutely* survive the implosion of Anthropic and OpenAI at this point if the reasons for the implosion are "burning too much money." There'd have to be some catastrophe from another direction (which I'm not ruling out) to kill viability, and the legal avenue would require Microsoft to decide that throwing their legal team at the issue wouldn't be worthwhile. Which...TBD if we get to that point. 3/3