Brad Abrams retweeted
i'm happy to share some recent work on safely scaling a team of 1,000 Fable 5.2 agent to tackle ambitious, long-horizon work. the team was able to formalize a proof of Fermat's Last Theorem through coordination within our safety boundaries.
read it here:
claude.dev/blog/safely-scali…
Brad Abrams retweeted
if you want to work on building the safest agents in the world, or if you just want to call me and Mark your teammates, come work on Claude Managed Agents
I recently joined @AnthropicAI to work on our platform for production agents.
I'm excited to help teams safely take on long-running tasks with thousands of agents working together.
If you're building with agents, I'd love to hear from you.
What we’ve learned about agent memory, and why agents should dream: anthropic.com/research/compo…
Your agents should sleep on it.
Agents get should get smarter over night as they learn from all their previous runs!
Build your own dreaming service with dynamic workflows in Claude Managed Agents.
platform.claude.com/cookbook…
love to hear what you think!
Looking forward to @SlushHQ in Helsinki in the winter :-) Please join me on Nov 19 for a hands-on workshop on safely building multi-agent systems. platform.slush.org/public/sl…
1
7
630
Brad Abrams retweeted
Looking for a developer (who happens to be great at comms) to lead comms for the Claude developer platform:
job-boards.greenhouse.io/ant…
Great latency win for long agentic work
Message threads are now in public beta on the Messages API.
The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
6
1,031
Brad Abrams retweeted
AI is rough if you're a software engineer who loves the engineering more than the software. It’s incredible for those that care more about the end result than how it was built. Make the shift.
We asked Claude to explain how a toilet works.
It built this with a dynamic workflow in Claude Managed Agents.
youtu.be/1EhEAvAV3lE
Try it yourself in Claude Code:
run /claude-api managed-agents-onboard explainer-video-maker in Claude Code
Brad Abrams retweeted
⛑️ safety tip for Claude Managed Agents
Run /claude-api managed-agents-secure in Claude Code to compare your agent's settings with the security guide. It's read only and lists each setting to change.
platform.claude.com/docs/en/…
Brad Abrams retweeted
We've raised the max request size for Messages and Token Counting on the Claude API from 32 MB to 128 MB.
Image-heavy requests like long computer-use sessions and big PDFs were previously hitting 32 MB with plenty of context window left. Now you get ~4x more room.
Brad Abrams retweeted
Brains <> Hands is the way to go for so many reasons. The upsides to reliability, security, latency, and scalability far outweigh any perceived downsides, which are extremely solvable with just good engineering practices (or just telling Claude "pls fix")
Super happy to have played a smol role in popularizing this architecture alongside @gcemaj.
How Anthropic ships Claude with Claude
latent.space/p/opus55/
Opus 5.5 is the #1 LLM in the world right now. We go behind the scenes of the launch, with @katelyn_lesse and @angjiang, heads of Eng and Product behind the platform used to build everything from the first party harnesses like Code/Tag/Duet and by >50% of Anthropic's top customers!
Trust your agent. But read its logs.
A few resources to help if your agent is on Claude Managed Agents:
Read the guide: platform.claude.com/docs/en/…
Run in Claude Code: /claude-api managed-agents-secure
Review its sessions: platform.claude.com/cookbook…
Brad Abrams retweeted
We’ve added a new way to access Claude on Google Cloud.
You can now use the full Claude API from your Google Cloud account, including Claude Managed Agents, code execution, Opus 5.5 fast mode, and new features as they launch.
Brad Abrams retweeted
Seeing lots of struggles with API keys.
Started using WIF in more places recently so wanted to share in case this helps: platform.claude.com/docs/en/…
Do you know what your agents did last night?
New cookbook: point Claude at your agent transcripts and it flags the sessions worth a second look. Works for CMA, Agent SDK and your own agent transcripts
platform.claude.com/cookbook…
I built a set of strategies with dynamic workflow for Claude Managed Agents: Advise, Grade, Fan-out, dream.
linkedin.com/pulse/give-your…
What other patterns would you like too see?
1
6
611
Building secure cloud-hosted agents for models this smart and persistent is not a weekend project.
We’ve put that work into Claude Managed Agents. For example, your API keys stay in a vault and the sandbox only ever sees a placeholder.
We just published a guide to locking down a deployment. platform.claude.com/docs/en/…
How confident are you in your agent sandbox?