@AnthropicAI + @Innovators_Room - Previously CEO @Zenjobcom, @WellingtonVC, @Citymapper, @McKinsey; CompSci @UniofOxford
Berlin
Joined March 2009
- Tweets3.8K
- Following999
- Followers2.7K
- Likes2.7K
Stephan von Perger retweeted
Claude Fable 5 will be available again globally tomorrow.
After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.
We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.
Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.
Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.
Read our full blog: anthropic.com/news/redeployi…
Stephan von Perger retweeted
Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in software engineering, knowledge work, scientific research, and vision.
The longer and more complex the task, the larger Fable 5’s lead over our other models.
ALT Benchmark table titled Mythos 5 & Fable 5, comparing Claude Mythos 5 and Fable 5 against Claude Mythos Preview, Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro.
Let’s go!
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
ALT Benchmark table showing how Claude Opus 4.8 compares to its predecessor and to other models on tests of coding, agentic skills, reasoning, and practical knowledge work tasks.
Time to build! 4.7
Stephan von Perger retweeted
WIP: Index × @AnthropicAI sold out. So we added a lock-in on Friday evening.
Claude Code Lock-In — April 17, London. Show up with a laptop and an idea, spend the evening building something with Claude Code (@claudeai) alongside Index and Anthropic's Applied AI teams, and actually ship it.
Space is limited — let us know if you'd like to join → luma.com/5sr1umwp
Stephan von Perger retweeted
We released Claude Cowork as a research preview 12 weeks ago. Since then, millions of people have made it part of how they work - and hundreds of thousands more are trying it every week. I'm really optimistic about its ability to make people more productive.
Today, we're removing the "preview" label, marking Claude Cowork as generally available - with new enterprise controls for businesses.
Looking forward to this in London on April 17th!
Nothing great starts finished.
The best code, companies, and collaborations all began as a draft. WIP is a day for the builders who know it's the work in progress worth celebrating.
Hear from builders at @wiz_io, @Revolut, @incident_io, @DeepLcom, @motorway, @NordNewsroom and @ElevenLabs — joined by @AnthropicAI's technical, research, and GTM team and the Index team, including @dannyrimer, @janatindex, @hannahlseal and many more.
Spaces are limited - request your spot here: indexventures.typeform.com/t…
Stephan von Perger retweeted
15 builders. Important problems.
From March 20th to 23rd, we’re bringing together 15 of the best builders across Europe in Oxford - people determined to build things that matter.
Join us for a weekend of focused building, Claude credits, and exclusive access to General Reasoning’s long-horizon RL environments platform.
Find out more and sign up here using the link in the comments.
@joinef
@IndexVentures
@GenReasoning
@AnthropicAI
@matthewclifford
@GeorgiaS_IV
@vonperger
@AAyman_1302
@_tobiaswagner
💬 My current setup to speak to #ClaudeCode in 🇬🇧🇩🇪:
Spokenly + @nvidia Parakeet Tdt 0.6B V3
• free, on-device
• pauses music playing while speaking
• supports in-sentence language switch
• fast and accurate
Has anyone in my network tried a USB pedal for @claudeai?
Exciting times to build!
This past month, we've been running an experiment. What happens when you give insanely talented engineers the capital and space to go deeper on an idea they've been working on?
That became The Engine Room - a one-week sprint where eight teams across the London + European ecosystem received £5,000 and time to build.
We wrapped the program with Demo Day this week, and seeing a room full of investors, founders, operators, and friends supporting the pitches was incredible.
The cohort:
- @Itsishan_ – Agendex
- @zeror404 – Flics.ai
- Jack Peck – Factal
- @TervitLuke – PrepShift
- Max Amos – Redacted Lab
- @Guanming717 – Aegis Robotics
- @AdirajuSumukh – Democratizing Evals
Another big part of it was the support from people we deeply admire: @giaccoangelo (@ElevenLabs), Max McGuinness and @vonperger (@AnthropicAI), @lkshaas (@GoogleDeepMind), @timmm_chong (@yondercard), @thomaspaulmann (@raycast), @danielreidcahn (Slingshot AI), Patrick Gilday, @Jackyhuang (@hardlaunching / @coinbase), @_RHoward (@awscloud), @mattzcarey (@Cloudflare), Raj P (@stripe), and @mandeepsinghuk. Thank you all so much for your time and insights.
All in all, the program delivered exactly what we hoped for - bringing together technical talent who, in our view, are right at the beginning of building things that matter.
So what's next?
Next batch coming in 2026.
More soon from @chapterone and @aiengine_hack 🔥
Thank you to @jmj and @zqwq333 for the incredible partnership and organization to make this happen.
Links: Follow our mailing list (aienginehack.beehiiv.com/sub…), Luma (luma.com/aiengine), and Substack (jamesin.substack.com/) and check out the website (engineroom.run/) for the latest updates!
Stephan von Perger retweeted
Introducing Claude Sonnet 4.5—the best coding model in the world.
It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.
ALT A benchmark table for Claude Sonnet 4.5 showing leading performance across many domains, including agentic coding, computer use, mathematics, graduate-level reasoning, and financial analysis.
🤖 If one AI can apply on your behalf, another one can screen your application, then another one can probably also do the job itself.
Jobs which the AI can’t do in coming years, can probably also not ai-applied or ai-recruited for without significant human intervention?
Inevitable that job seeking and hiring will just become instant auto matching of the best matching people to jobs by AI
Not dystopian to me because it'll put the right people in the right jobs
It can even auto negotiate salary and compensation
Me and @andreyazimov are building this now with getapplicantai.com
Has anyone figured out yet how to stop the double saving of @WhatsApp photos to the iOS photo roll? reddit.com/r/ios/comments/1b…
Stephan von Perger retweeted
I have a few subsidised one-on-one slots available for ambitious early to mid career women working on x-risk/GCRs (e.g. AI alignment, AI governance, biosecurity, nuclear security, climate or other neglected fields). Express interest: katieglass.com
