Kimi Code, our open-source coding agent, just got a major upgrade!
🔹One-line CLI install, zero setup, fast startup
🔹Drag in videos as coding context: reference-to-LUT, long-video-to-short, screen-recording-to-code, and more
���Plugins for stocks, financial reports, academic papers, with more coming
🔹Supports the ACP protocol, and works with JetBrains, Zed, and more
🔹Hooks for custom tools and workflows
Try it with Kimi K2.6 ���� https://t.co/1owb8LUBMQ
Issues, plugin ideas, and PRs welcome! Community feedback helps shape what ships next.🚀
🔴 LIVE NOW: @Letta_AI and @browserbase are demoing how they build agents on Claude Opus 4.8, hosted by @AnthropicAI.
Real products, real lessons, live Q&A. Register to join now 👇
https://t.co/18vFjdLbT5
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.
As agentic workflows scale,
the challenge is no longer just compute,
it’s reasoning cost.
Context. Retries. Reviewers.
Planners. Verifiers.
Every step consumes tokens.
“How smart should this step be?”
…is becoming an architecture decision.
Contrarian thought from #TheSpeedOfTrust in the Agentic AI era:
As AI reduces the cost of execution, trust becomes the bottleneck.
Anyone can generate code or content with AI.
But credibility, judgment and accountability stay human advantages.
#AgenticAI#AI#Trust
A question from The Speed of Trust:
If people who know you had to sort your picture into stacks,
“trusted” or “not trusted”
where would it go?
Credibility is built quietly:
consistency, honesty, accountability, follow through.
Trust compounds over time.
#TheSpeedOfTrust
Changing how you see, speak, and behave isn't just about shifting your mindset, it is a loop. Change one, and the rest follow. 🔄
Trust starts with the small, deliberate shifts we make in our daily actions.
#TheSpeedOfTrust#GrowthMindset#Leadership
Finally, a way to run AI agents that doesn't feel like a black box.
With the new self-hosted sandboxes and MCP tunnels in @claudeai Managed Agents, you get all the power of an agent while keeping your data and security exactly where they belong: in your own infrastructure. 🔒
Live from Code with Claude London: we're launching self-hosted sandboxes (public beta) and MCP tunnels (research preview) in Claude Managed Agents.
Run agents inside your own perimeter, with your security controls applied by default.
Live from Code with Claude London: we're launching self-hosted sandboxes (public beta) and MCP tunnels (research preview) in Claude Managed Agents.
Run agents inside your own perimeter, with your security controls applied by default.
Trust is the ultimate cheat code for life and work. Reading The Speed of Trust and realizing that when trust goes up, everything else just gets
easier. 📖
#TheSpeedOfTrust#GrowthMindset
🗯️💬🗯️📬📭📩✍️ Join our Community Feedback Session on Sept 18th (19th in some APAC time zones)! We’re collecting thoughts on what’s working (and what’s not) in the new alpha.
Come ready with notes on:
• create_agent
• Middleware API
• Structured output logic
• New docs site
• Standard content blocks
Your feedback shapes the future — see you there!👉RSVP: https://t.co/9GVOU8UHCs
🫡 Assistants
We’re winding down the Assistants API beta. It will sunset one year from now, August 26, 2026. We’ve put together a guide to help you migrate to the Responses API: https://t.co/cu0Kby8LKV.
Assistants were our early take on how agents could be built (before reasoning models). In the Responses API announcement, we said we’d follow up on deprecating the Assistants API once Responses reached feature parity — and it now has. Based on your feedback, we’ve folded the best parts of Assistants into Responses, including code interpreter and persistent conversations.
Responses are simpler, and include built-in tools (deep research, MCP, and computer use). With a single call, you can run multi-step workflows across tools and model turns. And with GPT-5, reasoning tokens are preserved between turns.
The Responses API has already overtaken Chat Completions in token activity. It’s our recommended path to integrate with the OpenAI API today, and for the future.
A few GPT-5 updates heading into the weekend:
- Now rolled out to 100% of Plus, Pro, Team, and Free users
- We’ve finished implementing 2x rate limits for Plus and Team for the weekend; next week we’re rolling out mini versions of GPT-5 & GPT-5 thinking to take over until your limits reset
- GPT-5 thinking and GPT-5 pro now in main model picker
- GPT-4o is now also available to Plus and Team users. To use it across platforms, go to settings on ChatGPT web and toggle on “show legacy models.”
���GPT-5 is now in LangSmith Playground w/ built-in cost tracking.
Available models:
• gpt-5
• gpt-5-mini (now default)
• gpt-5-nano
Start building with the latest models → https://t.co/fRKTx58e2V
gpt-oss is out!
we made an open model that performs at the level of o4-mini and runs on a high-end laptop (WTF!!)
(and a smaller one that runs on a phone).
super proud of the team; big triumph of technology.