JunShern

Wake up babe, new agent benchmark dropped! MLE-bench - or, as I like to say it, EMILY BENCH 🙆🏻‍♀️. Main summary on Neil’s thread, but adding some highlights of my own:

With the help of Claude Mythos Preview, the Firefox team fixed more security bugs in April than in the past 15 months combined.

at the intersection of i've basically been replaced and i've never worked so hard in my life

Back in ~November, our team picked a stretch goal of seeing if we could find and fix vulnerabilities in Firefox with Opus 4.6. In 2 weeks, we found 22, and ~1/5th of all high severity CVEs in a year. For our team, this feels like a rubicon moment.

Nicholas Carlini at [un]prompted. If you know Carlini, you know this is a startling claim.

The increasingly-hyperbolic METR graph is actually good news for safety. We just have to survive a brief singularity in March, and then afterwards the models will never be able to do more than undo a few hours' worth of work

Across all mini-SWE-agent + <model> runs, SWE-bench Verified's current "ceiling"? - 87.4% (0.874 - 0.8) * 500 = another *37* instances that aren't solved consistently. If you recalculate this number across all official SWE-bench Verified submissions? - 95% from SWE-bench site

We’re publishing a new constitution for Claude. The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process. anthropic.com/news/claude-ne…

sent claude opus 4.5 a christmas card :-) then i asked if it wanted to send any of its own, and it made a bunch (thread)

New on our Frontier Red Team blog: We tested whether AIs can exploit blockchain smart contracts. In simulated testing, AI agents found $4.6M in exploits. The research (with @MATSprogram and the Anthropic Fellows program) also developed a new benchmark: red.anthropic.com/2025/smart-con…

Remarkably, prompts that gave the model permission to reward hack stopped the broader misalignment. This is “inoculation prompting”: framing reward hacking as acceptable prevents the model from making a link between reward hacking and misalignment—and stops the generalization.

@Miles_Brundage @idavidrein Yes, I thought it was supposed to be “Google proof” 😂

Title: Advice for a young investigator in the first and last days of the Anthropocene Abstract: Within just a few years, it is likely that we will create AI systems that outperform the best humans on all intellectual tasks. This will have implications for your research and career! I will give practical advice, and concrete criteria to consider, when choosing research projects, and making professional decisions, in these last few years before AGI. This is my current go-to academic talk. It's mostly targeted at early career scientists. It gets diverse and strong reactions. Let's try it here. Posting slides with speaker notes... -- The title is a play on a very opinionated and pragmatic book by the nobel prize winner ramon y cajal, who is one of the founders of modern neuroscience. To get you in the right mindset, on the right we have a plot of GDP vs time. That is you, standing precariously on the top of that curve. You are thinking to yourself -- I live in a pretty normal world. Some things are going to change, but the future is going to look mostly like a linear extrapolation of the present. And the plot should suggest that this may not be the right perspective on the future. This plot by the way looks surprisingly similar even if you plot it on a log scale. We didn't stabilize on our current rate of growth until around 1950.

We’re hiring someone to run the Anthropic Fellows Program! Our research collaborations have led to some of our best safety research and hires. We’re looking for an exceptional ops generalist, TPM, or research/eng manager to help us significantly scale and improve our collabs 🧵

@MariusHobbhahn @TIME Huge, congrats Marius!

Honored and humbled to be in @TIME's list of the TIME100 AI of 2025! time.com/collections/ti… #TIME100AI

Ender, that wasn’t an RL environment with a verifiable reward, those were real Amazon orders you placed.

Launching now — a new blog for research from @AnthropicAI’s Frontier Red Team and others. > red.anthropic.com We’ll be covering our internal research on cyber, bio, autonomy, national security and more.

Humanity has prevailed (for now!) I'm completely exhausted. I figured, I had 10h of sleep in the last 3 days and I'm barely alive. I'll post more about the contest when I get some rest. (To be clear, those are provisional results, but my lead should be big enough)

A simple AGI safety technique: AI’s thoughts are in plain English, just read them We know it works, with OK (not perfect) transparency! The risk is fragility: RL training, new architectures, etc threaten transparency Experts from many orgs agree we should try to preserve it: 🧵

We are at ICLR, come chat!

To interpret AI benchmarks, we need to look at the data. Top-level numbers don't mean what you think: there may be broken tasks, unexpected behaviors, or near-misses. We're introducing Docent to accelerate analysis of AI agent transcripts. It can spot surprises in seconds. 🧵👇