Claude Tag moves into Slack channels

THE AI BRIEF

Today's signal: Anthropic moved Claude from a tool a user opens to a teammate that lives in a Slack channel.

In today’s issue:

• Main story: Anthropic put a persistent AI agent inside your company's Slack channels

• Also worth knowing: OpenAI's first inference chip, Meta glasses at $299, Engram's $98M for enterprise memory, and a Mythos disclosure on classified US systems

THE READ

Claude Tag replaces the old Slack app with one shared Claude per channel that remembers context, runs work asynchronously, and can act on its own. The procurement surface for AI agents just moved from a vendor tab into the room where decisions get made.

Anthropic launched Claude Tag yesterday in beta for Enterprise and Team customers, replacing the existing Claude in the Slack app. An administrator pairs Claude with a Slack workspace, grants access to specific tools and data sources, sets a spending limit, and defines which channels Claude can operate in. Anyone in those channels can tag @Claude with a request, and Claude executes the task across whatever tools the admin connected. Anthropic says 65% of its own product team's code is now created through its internal version of the system.

Three things make this different from the previous generation of Slack-AI integrations. It is multiplayer: there is one Claude per channel, not one per user, so everyone sees what it is working on, and anyone can pick up where the last person left off. It learns over time inside the channels it monitors, so projects do not get re-explained at each turn. And it works asynchronously on tasks that run for hours or days. The model behind it is Claude Opus 4.8, which Anthropic shipped less than a month ago.

AI READY PRO

I will get you AI trained in 30 days for free

After several months of running it quietly with top AI operators and teams, I’m excited to launch AI Ready Pro today to our newsletter subscribers like you.

It’s a 30-day personalized AI training program that is personalized to you and where you are in your AI journey. It’s the culmination of 100’s of hours spent teaching AI to top Fortune 500 AI teams and operators (and working with Mark Cuban).

Each day, you receive a 10-15 minute exercise to complete to improve your AI skills.

This isn’t vague AI theory or a way to pitch you a tool. It’s 30 days of learning to use AI in your real work. So you come out the other side with the skills to use AI to actually improve you and your teams output.

If you lead a team, AI Ready Team is open today too, also for readers first. It’s the same engine, but to train your whole organization. You’ll get a detailed view of:

• Your team’s AI readiness level + detailed strengths & weaknesses

• Ranked list of AI automation opportunities

• Get tactical advice on how to advance AI efforts in your org

Leading a team? Take the Team assessment instead.

What I keep coming back to in client conversations this quarter is where the AI surface actually lives. For two years, most enterprise AI has been a tab in a browser or a sidebar in an IDE. Claude Tag moves the surface into the channel where the team already coordinates, which changes who can use it and who decides what it can touch. The administrator question stops being "which seats do we license" and starts being "which channels does this agent sit in, what tools does it hold credentials for, and what is the budget cap." Anthropic put those choices in the admin's hands by default, which is the right shape. The operator question is whether the channels currently in scope match the data-access posture the security team would write down if asked.

The honest hedge: a 65% internal-code claim from the vendor that sells the product is a marketing number, not an audited one. Read Claude Tag as the clearest signal yet that the enterprise AI surface is moving from per-user tools to per-channel teammates, not as a guarantee of the productivity numbers Anthropic is quoting from inside its own building.

SPONSORED BY CLUTCH

Hire secure AI teammates that work 24/7.

Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.

Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.

Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.

ALSO WORTH KNOWING

OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom inference chip, developed from design to production in nine months with Broadcom and Celestica. The accelerator targets LLM inference and is planned for gigawatt-scale deployment. OpenAI now has a path off pure Nvidia dependency for the workload that actually serves ChatGPT and the API.

Meta launched its own-brand Meta Glasses starting at $299 with EssilorLuxottica, powered by Muse Spark on day one. The launch extends Meta's wearable AI line below the Ray-Ban price point. A consumer story for now, with the same form factor likely landing in field-service and frontline-worker procurement conversations within a year.

Engram raised $98 million to build compact organization-specific memory models that sit alongside frontier LLMs and reduce token waste on repeated context. Enterprise AI memory is now a funded startup category, separate from vector databases and from fine-tuning.

Anthropic's Mythos model found vulnerabilities in classified US government systems during Project Glasswing, a US official told SecurityWeek. The disclosure lands the same month Mythos and Fable 5 were briefly blocked under export-control review. The same model that the government restricted is the one being run inside classified environments to find defense vulnerabilities.

Gong launched Revenue Harness, a governed-agent layer that lets revenue teams deploy AI agents on top of Gong's call and account data with audit trails and human approval gates. The vertical-SaaS pattern of turning a proprietary data graph into an agent execution surface, applied to GTM.

Seltz raised $12.5 million to rebuild web search for AI agents rather than humans. The bet is that agents need structured, machine-readable retrieval at a different shape than the indexed page results a person clicks. A new category if it works, and the first venture round to name the problem directly.

WATCHING TOMORROW

Friday brings the Week in AI roundup. The pattern worth tracking is the cluster of agent-surface launches inside seven days: Claude Tag in Slack, Gong Revenue Harness for revenue teams, NeoXam for investment ops. Three different shapes of the agent moving into the surface where work already happens.

REPLY

If Claude moved into one of your Slack channels tomorrow, which channel would you actually grant it, and what tools would you let it hold credentials for? I read every reply.

If a colleague is about to evaluate the Claude Tag beta for your team, forward this issue to them before that call.

Back tomorrow,
Haroon