Post

hailey
@hailey.at

found dario’s bluesky account

Quantian@quantian.bsky.social

If you are still on the “cheap Chinese models will take share from US labs” cope train you need to get off. US labs are now distilling their own models to compete across the Pareto frontier, so Haiku is better model than Kimi at 1/8th the cost. The only reason to use it is if you MUST self-host.

hailey
@hailey.at

“cope train” while imagining the only use case of a chinese model is to self host lol

Thoth
@thoth.ptnote.dev

I feel like "I fucking hate you and don't want to contribute to your success" is perhaps underrated as a reason to pay white-label inference providers over the frontier labs.

2 reposts35 likes

Thoth
@thoth.ptnote.dev

I think this is the single largest motivator of my use of open weights. There was that few month window where it really seemed like every professional programmer (or anyone who wanted to become one) was going to HAVE to pay OAI or Ant, and I felt genuinely kind of unwell.

Thoth
@thoth.ptnote.dev

I do think they're right about GLM though. I mean, I pretty much exclusively use flash models because I don't have money and I recently tried DSv4.1 1017 since it's nominally "better" than GLM-5.3-Flash.

Thoth
@thoth.ptnote.dev

It works pretty well... but compared to GLM it feels like the model has ADHD. Much easier to distract.