Diffusion seems like a much more natural way to get good reasoning to me. And you can directly scale as many more steps in diffusion as you want – no problem!
@jeremyphoward
-
KV Caching and Pricing for AI Model Competitiveness
By
–
They really need KV caching though – otherwise the pricing is just not competitive. (OTOH if they *do* add caching at a good price point, they will be amazing!)
-
HuggingFace bf16 Positional Embeddings Bug Went Unnoticed
By
–
Remember how HF precalculated positional embeddings in bf16 for a long time and the impact was small enough that it slipped through unnoticed? :O
-
Positional Embeddings Were Less Important Than Originally Thought
By
–
It actually turned out the particular spectrum of the chosen waves often is pretty horrible, but it also turned out positional embeddings weren't actually that important anyway so no-one really noticed for years.
-
Search Providers for AI Code Tools: Brave, SerpAPI, Gemini
By
–
What search providers are you all using with openclaw/pi/opencode/etc? Brave; serpapi; gemini; …? Got any favorites?
-
DeepSeek V4 Prefill Support Praised Amid Provider Trend
By
–
This is great – @deepseek_ai V4 supports prefill! 😀 Most other providers have been dropping support for this critically important capability, so wonderful to see at least one company stepping up.
-
LiteLLM Fixes Bug with Recent Pull Request Merge
By
–
Looks like they recently landed a PR fixing that https://
github.com/BerriAI/litell
m/pull/26157
… -
V4 Performance Improvement: Fast Speed Achievement
By
–
Yeah although I found it too slow. But v4 is fast! 😀
-
DeepSeek v4 Now Supports Fill-in-the-Middle Capability
By
–
Also the new @deepseek_ai v4 supports FIM now! 😀 (Requires using the beta header.)