Entries for July 27, 2026
-
I've noticed GPT line-up of models have a hard time keeping their math notation consistent during formulation (including Pro) Interestingly, Fable is better at this Yet another skill (Consistent notation) for something that should be common sense: github.com/osolmaz/tools/…Image hidden -
A brave new world -
Models don't have common sense when it comes to spending money. They can waste compute resources, API tokens, $$$ easily I created a skill to always be on top of my resource usage, this one is for doing ML experiments on Hugging Face Jobs If it's above 20 bucks, it will always ask me to authorize it before launching it Paid Compute Launch:Image hidden -
-
Use Kimi, GLM, Qwen and other models from Hugging Face Inference Providers @FireworksAI_HQ, @togethercompute, @novita_labs and so on, easily and directly in Pi pi install npm:pi-huggingface-oauth Then in Pi, /login -> choose Hugging Face -> go through oauth Dynamically loads the model list, so you get the latest model releases automatically, like Kimi K3 tomorrow @pidotdev extension repo: github.com/osolmaz/pi-huggingface-oauth Available models on @huggingface Inference Providers: -
I have not been as excited for a model release since o1 came out Pivotal moment for AI. Marks the end of the duopolistic OpenAI/Anthropic chokehold on the industry Like the model on HF and get notified: huggingface.co/moonshotai/Kim…Image hidden