In early May, the best superforecasters predicted that, by the end of the year, the longest METR 80% task horizons would reach 3-4 hours. In late May, Claude Mythos achieved that number.
LLMS
-
New 2nd edition of RAG-Driven Generative AI book
By
–
New (2nd edition) from @PacktDataML available at https://
amzn.to/4tULP1b RAG-Driven Generative AI — Build MAS-RAG with DualRAG, GraphRAG, multimodal video pipelines, and Oracle Database 23ai 𝗞𝗲𝘆 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝘀:
Master DualRAG by combining vector search with SQL filtering -
Top Chinese AI labs Moonshot and DeepSeek dominate
By
–
S-Tier Chinese Labs: Moonshot and DeepSeek These 2 are levels above everyone else
-
Manual transcription of bank wire details avoided by Claude in 2026
By
–
Me: Happy to upload our bank's standard letter.
TA: No it has to be on *our* standard format.
Me, twitching: You want me to manually transcribe approximately 100 characters into boxes on a flat PDF where one being wrong causes wire misdelivery. That is… oh it's 2026. Claude, go -
Search across previous Claude sessions is not good.
By
–
search across previous claude sessions is not good.
-
Gary Marcus argues LLM token prices will decrease, not increase
By
–
This is really intellectually dishonest. I have not been arguing that LLM token prices are increasing (though the all you can eat buffet is over), I have been arguing the *opposite*, viz that they will go down (e.g., in my recent tweet on commodity pricing that got 1 million
-
GPT-5.6 release imminent, prepare for wild ride
By
–
This is probably GPT-5.6. Either tomorrow or coming week i suppose. Get ready friends. We are in for a wild ride!
-
Build Multimodal AI Knowledge Base with Gemini Embedding 2
By
–
How to Build a Multimodal AI Knowledge Base With Gemini Embedding 2 https://
madebyagents.com/blog/build-mul
timodal-rag-gemini-embedding-2?utm_source=dlvr.it&utm_medium=twitter
… #ArtificialIntelligence #MachineLearning #DataScience #AIStrategy #DigitalTransformation #GenerativeAI #technology #ChiefDataOfficer -
New Gemma 4 12B model matches 26B performance
By
–
NEW GEMMA 4 12B MODEL! Google's beloved open-source model saga gets an update today to add the 12B model, which, thanks to its new architecture, performs on par with the equivalent of 26B from a few months ago! The model is multimodal in input: vision, audio, and text
-
Common Misconceptions About How LLMs Actually Work
By
–
Most people, including really accomplished people, don't have an accurate mental model of how LLMs operate (and why would they?) You see this in wide beliefs that AI is just copying from known sources, or that it only produces average answers, or that it can't generate new ideas