AI is already shaping decisions in councils across England but the governance needs to catch up. New research outlines three practical realities shaping how councils govern #AI. Read more: ai.cam.ac.uk/blog/three-thin…
RESEARCH
-
Leak: Anthropic Claude Next-Generation Models 2026
By
–
tl;dr upcoming Anthropic models due to leak – Claude "Mythos", Capybara tier already in version 2, it'll have a context window of 1m
– Mythos comes in "fast" and regular thinking
– Opus 4.7 and Sonnet 4.8 already within the code
– Claude "buddy" also found, but not clear what it -
Leaked List of Invented Verbs Shown When Claude Thinks
By
–
Obvio la parte más interesante del leak, el listado de verbos inventados que se muestran cuando Claude piensa
-
Everything You Need to Know About Recursive Language Models
By
–
Everything You Need to Know About Recursive Language Models https://
machinelearningmastery.com/everything-you
-need-to-know-about-recursive-language-models/?utm_source=dlvr.it&utm_medium=twitter
… -
Leaked Claude AI Model Versions and Codenames Revealed
By
–
HOLY SH*T, here we go: "Undercover mode – ciritical" "Capybara", Opus Mythos "Unreleased Model version number Opus 4.7, Sonnet 4.8"
-
Claude Code Leak Reveals Capybara Model
By
–
BREAKING : Claude Code source code has been leaked via the map file in Anthropic's npm registry. The code references an unreleased model named "Capybara", mentioned in the recently revealed blog post. UNDERCOVER MODE
-
Model Training with SFT Using Claude and GPT-4o-mini
By
–
When I published this post I hadn't yet come across Trip's own writeup of the project, which provides way more detail about how he trained the model including SFT (supervised fine tuning) against synthetic chat examples created using Claude Haiku and GPT-4o-mini
-
Grok’s Benchmark Success: Context and Limitations
By
–
The post frames this as Grok taking over but the benchmark is one specific slice of agentic performance. Useful signal, not the whole picture.
-
First Conference with SKILL.md Submissions
By
–
The first conference where your submission isn't a paper.
— God of Prompt (@godofprompt) 31 mars 2026
It's a SKILL.md file Claude can execute, review, and reproduce end-to-end.
Stanford and Princeton just made reproducibility the submission format.
$50,000 prize pool. 364 winners.
Deadline April 5 👇… https://t.co/4ql4fVGe4QThe first conference where your submission isn't a paper. It's a SKILL.md file Claude can execute, review, and reproduce end-to-end. Stanford and Princeton just made reproducibility the submission format. $50,000 prize pool. 364 winners. Deadline April 5
-
Specific Complaints About Token Consumption and Output Quality Signal Real Issues
By
–
Burning tokens faster for worse output is a very specific complaint and specific complaints usually mean something real is happening. Vague dissatisfaction is noise. This pattern showing up across multiple users is a signal.