· 3 items
Latent Space ↗
[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut…
- What
- Anthropic released Claude Fable/Mythos 5.1 with SOTA performance, a 75% cache price cut, and 70% higher output tokens.
- Why
- Massive price reduction on prompt caching combined with higher output makes complex agent loops significantly cheaper.
- Impact
- Reduces token costs for building autonomous agents.
arXiv ↗
Language Models Can Control Their Own Attention
- What
- New research demonstrates that language models can actively control their own attention mechanisms over the KV cache.
- Why
- This bypasses the need to scan full long contexts for every single token, drastically speeding up retrieval.
- Impact
- Improves latency for long-context voice agents.
OpenAI ↗
ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT
- What
- ATV Big Air Tour used ChatGPT Work to build an inventory website from merchandise photos in just 15 minutes.
- Why
- Demonstrates rapid, non-developer prototyping and workflow automation using standard enterprise AI tools.
- Impact
- Speeds up prototyping and no-code product iteration.