New research result: we use Claude to make fully autonomous progress on scalable oversight research, as measured by performance gap recovered (PGR).
Claude iterates on a number of different techniques and ends up significantly outperforming human researchers for $18k in credits.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
Claude Opus 4.8 showing up the Agent SDK published to npm 20 mins ago:
Claude Opus 4.8 — The most capable Claude model to date.
Highly autonomous and state-of-the-art for long-horizon agentic work, knowledge work, and memory, with clearer, warmer writing.
Uses the same API surface as Opus 4.7:
Adaptive thinking only
Sampling parameters removed
budget_tokens removed
1M context window at standard API pricing
No long-context premium
See shared/model-migration.md → Migrating to Opus 4.8
A 4.7 → 4.8 migration is a model ID swap plus prompt re-tuning, with no new breaking changes.
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
New post: "Generalization Dynamics of LM Pre-training"
Most people (including me) assume that LMs smoothly mature from pattern-matching to generalizing.
This mental model is wrong. The true dynamics are stranger, and far more fascinating!
We call it Mode-Hopping.
New work with @AlecRad and @DavidDuvenaud:
Have you ever dreamed of talking to someone from the past? Introducing talkie, a 13B model trained only on pre-1931 text.
Vintage models should help us to understand how LMs generalize (e.g., can we teach talkie to code?). Thread:
New Anthropic Fellows research: developing an Automated Alignment Researcher.
We ran an experiment to learn whether Claude Opus 4.6 could accelerate research on a key alignment problem: using a weak AI model to supervise the training of a stronger one.
https://t.co/OAxCjOiWTm
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.
It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
https://t.co/NQ7IfEtYk7
Mistakes happen. As a team, the important thing is to recognize it’s never an individuals’s fault — it’s the process, the culture, or the infra.
In this case, there was a manual deploy step that should have been better automated. Our team has made a few improvements to the automation for next time, a couple more on the way.
📰 Excited to share the results of this Anthropic Fellows project! We built an automated alignment agent (A3) for safety finetuning.
The project was great fun, and @jifan_zhang was an amazing collaborator to work with - excited to see what he does next!
New in Claude Code: Remote Control.
Kick off a task in your terminal and pick it up from your phone while you take a walk or join a meeting.
Claude keeps running on your machine, and you can control the session from the Claude app or https://t.co/er6Blrr63e
@Kimi_Moonshot used policy mirror descent (PMD) for RL in Kimi k1.5/k2. Most take it simply as PG+KL with an updating anchor, but this is not the full story. Check our blog for some interesting findings about this algorithm: https://t.co/6PgwAcjgp5