
Community, Claude, Code at @AnthropicAI
Colorado
Joined December 2011
- Tweets3.9K
- Following1K
- Followers35.3K
- Likes16.7K
Pinned Tweet
I turned the Advent of Claude into a blog post where I could provide additional context, resources, and re-ordered the tips into a more logical order going from startup to advanced capabilities.
Free the standards! So happy to see this.
We're donating Agent Skills to the Agentic AI Foundation (@AgenticAIFdn), the same foundation we donated MCP to.
Since December 2025, 70+ agent platforms have integrated the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code.
Here's what that means for you:
I've been having a blast redesigning and rebuilding my personal site (do people still have these?).
Now with more Clawd.
Also working on a CMS for Hugo called Victor that will power it and will be releasing it as OSS shortly.
Chatbots wait for you to ask.
Claude Duet doesn't.
It briefs you every morning, pings you only when something actually needs you, and keeps grinding after you close the laptop.
Your duet partner just dropped 🎶
Learn more: claude.com/resources/article…
Get on the waitlist: claude.com/product/claude-du…
Claude Duet is a proactive agent that has become my daily driver for everything non-code related. I dont know what I would do without it.
Thank you for trying Claude!
Benchmarking a http proxy built with Go and ported to Rust to see if I can squeeze even more performance out of it.

ALT Dark-themed evaluation note headed in red "✗ Goal could not be achieved (4h · 9 turns · 4m tokens)", explaining that the 5x performance requirement is not met: proxied throughput 1.13x on macOS (58.9k vs 66.5k rps) and 1.3x–1.7x on Linux (115k vs 149k rps at 64 connections; 121k vs 206k at 512), CPU per proxied request 2.2x on macOS and 1.7x on Linux (46 µs vs 27 µs), with only the WAF microbenchmark reaching about 5.2x (594 µs vs 113 µs). It concludes the assistant states the target cannot be met for proxying and asks the user to clear or restate the goal, and that the port is complete and its tests pass but the performance condition is unsatisfied.
The fastest way to learn Claude Code isn't the docs.
It's asking Claude.
Just ask how do hooks work. Then ask it to write you a hook. Then ask it to turn your release checklist into a skill.
It knows itself better than any tutorial.
A brain with no body just talks.
A brain with a body gets things done.
The LLM (model) is the brain. The harness is the body: the tools, files and loop around the model. The Agent SDK let's you define the body for your use case.
Learn more about the Claude Agent SDK: code.claude.com/docs/en/agen…
Ado retweeted
We’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new wholly owned entity dedicated entirely to the public benefit.
It will start off with 5% of our shares. For every 50% increase in the company’s value over a set baseline, we can contribute up to an extra 2%, subject to our board’s discretion.
In total, this means that roughly 20% of Anthropic’s shares are committed to the public good. Our co-founders have already pledged to donate 80% of their equity, and around half our employees are donating a portion of their equity as part of a donation matching program we’ve run since our founding.
In the post below, we share more details about our new fund.
anthropic.com/news/public-be…
Claude Code mods change how Claude Code looks and behaves. They are functions that hook into events, change what it does, and can draw their own UI in the terminal.
This one holds risky shell commands so you can see what it would delete.
Check out out more examples and learn to build your own: claude.dev/mods/
Adding automations to my smart home app is as easy as just describing what you want to happen. Claude figures out how to set up the rules properly and we're off (or in this case on). :)
Claude is building Claude Code.
Claude writes about 90% of Claude Code's code.
From June to September, merged PRs more than doubled while the share reverted fell by more than 40%.
We wrote about how we changed code review to get there, and how to adopt it:
claude.dev/blog/solving-code…
New in Claude Code: idle compaction
The prompt cache lives 1h. Leave a 200K+ token session idle ~54 min and it gets auto-compacted, so you come back to a warm summary, not a cold re-cache.
Set the Threshold: CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS
Turn Off: "idleCompaction": false
Great way to ensure I never purchase a Withings product ever again.
support.withings.com/hc/en-u…






