Claude is in Slack. OpenAI built its own chip. Big week.

Here's what the circle is talking about this week:

Anthropic launched Claude Tag, and it's a genuinely different kind of product than anything the lab has shipped before.

The mechanic is simple. You type @Claude in a Slack channel, assign it a task, and it gets to work. It can access your company's approved tools, databases, documents, and codebases. It completes the work on its own and reports back when it's done. No chat window. No switching apps. No re-explaining context. If your team lives in Slack, Claude now lives there too.

Andrej Karpathy called it the "third major redesign of AI interfaces," which is not the kind of observation Karpathy throws around lightly. His read is worth taking seriously. First it was chat. Then it was AI embedded into tools like Cursor. Now it's AI embedded directly into the places where team work actually happens.

What's different about Claude Tag isn't just the Slack integration. It's the continuity. Claude doesn't answer one question and disappear. It can follow ongoing conversations across channels, build context over time, and operate in the background without needing fresh instructions every session. That's a meaningful step away from the prompt-response loop that has defined AI interfaces since 2022.

There's also a competitive angle worth noting. Hundreds of startups raised serious money to build AI coworkers, AI project managers, and AI workplace assistants. Anthropic just put one inside the tool most of those startups' target customers already use every day. Watch the demo here to see how it actually works in practice.

This isn't incremental. Bringing AI into Slack changes the adoption question from "will people use this tool" to "will they tag it when they're already typing." That's a much easier habit to build. The startups building standalone AI coworker products should be paying very close attention.

OpenAI just published details on Jalapeño, its first custom-built chip, co-developed with Broadcom. The project went from initial design to factory-ready in nine months. OpenAI says that's the fastest ASIC development cycle ever achieved, and its own AI models helped with the design and optimization work along the way.

A quick clarification on what Jalapeño actually does. It's an inference chip, which means it handles the compute required to run finished AI models for users. This is separate from the chips used during training. Inference is what happens every time someone sends a message to ChatGPT or uses Codex. Right now that runs on Nvidia hardware. Jalapeño is built to change that.

In internal testing, OpenAI reports Jalapeño delivers performance per watt substantially better than the current state of the art. The company is targeting 10 gigawatts of custom chip-powered compute by 2029. Nvidia isn't going anywhere from the training side, but the inference side is now a different conversation.

The nine-month timeline is the part that catches your attention. Custom silicon typically takes years. OpenAI compressing that cycle significantly, partly by using its own models in the design process, points to where this is headed. Faster chips built faster, with AI doing more of the chip design work each generation.

Every dollar OpenAI spends running ChatGPT and Codex on Nvidia hardware is a cost it doesn't control. Jalapeño starts changing that math. When you own the chip, the model, and the product, you can tune each layer for the others in ways that commodity hardware never allows. The cost and efficiency gains compound across the whole stack.

Your finance close runs in #finance. Stripe and QuickBooks reconciled, runway updated, posted Sunday night without you asking.

Engineering review lands in #eng. Viktor pulled the open PRs, left comments on auth-refactor, flagged a dependency blocking api-pagination.

Campaign brief lands in #growth: Meta CPA up 18%, recommendation to pause broad match, a draft landing page already deployed for the variant test.

You hired him on day zero. He lives in Slack and Microsoft Teams alongside your contractors and investors, connects to 3,000+ tools, pushes back when you ship something dumb.

"Viktor is now an integral team member, and after weeks of use we still feel we haven't uncovered the full potential." Patrick, Director, Yarra Web.

Get Started for Free

Meta launched Meta Glasses, a new line of AI-powered smart glasses starting at $299, making them one of the most affordable entries in a category that has mostly lived at premium price points.

The timing is deliberate. Smart glasses shipments are growing and the competitive field is filling up fast. Apple, Google, OpenAI, and Snap are all moving in this direction. Meta's answer is to get there first at a price point that actually reaches mainstream consumers rather than early adopters with high hardware budgets.

The Ray-Ban Meta glasses already demonstrated there's real appetite for AI glasses that don't look ridiculous. This new line extends that bet with more styles through the EssilorLuxottica partnership, which gives Meta access to established eyewear brands and distribution that a pure tech company couldn't build overnight.

The smartphone comparison Meta is leaning into is bold but not crazy. The argument is that a device you already have on your face, connected to AI, accessible without pulling anything out of your pocket, could eventually replace some of what your phone does. At $299 that's a consumer product, not a developer toy. Whether it gets there depends on what the AI can actually do on your face in daily life, and that's still being figured out.

Three stories and one theme underneath all of them. AI is moving off the screen and into the places where you actually work.

Claude is in your Slack. OpenAI is building the chip that runs the inference for everything you use. Meta wants AI sitting on your face between your work and the world.

The interfaces are changing faster than the models are. A year ago the question was which AI was smartest. Today the question is where AI shows up and whether you notice it or not.

People inside this circle tend to notice. That's the point of being here.

Stay sharp. The cap means you're still paying attention when most people have stopped.