Anthropic Is Coming for Everything — Matthew Berman on the Real Shape of Claude Tag

Matthew Berman opens by naming what most people saw: 'Anthropic released a new feature called Claude Tag — a convenient way to invoke Claude from inside my Slack instance. Cool.' That was Berman's first take too. The more he read into it, the more concerned he got.

The revised reading: Claude Tag is Anthropic's entry point for owning all knowledge work. This isn't coding-specific. It affects anyone whose job involves a computer. Looking at surface features misses the design intent; tracing the intent makes it visible. That's the structure of the rest of the video.

Day-to-day usage is simple. Tag Claude in Slack the way you'd tag a teammate. But what's happening underneath is qualitatively different. Claude Tag carries the full context of everything inside your company and can execute real work. It knows who you are, who your colleagues are, understands your internal documents, and continuously reads through Slack channels. In Berman's words, it's 'building a graph of your entire company.'

And it doesn't always need to be tagged. It's designed to act proactively. This is the starting point for the AI-native company vision Anthropic has been laying out for months — AI that understands every corner of your company, runs 24/7, supports growth, surfaces issues, and helps human employees become a thousand times more productive. YC general partner Ankit Gupta adds a useful corroboration: YC's internal teams keep building versions of nearly every product OpenAI and Anthropic launch 6–12 months later — a useful way to see where things are going. AI-native company architecture has been a known target.

The risk framing is direct. A single company holding all that power isn't acceptable. Berman doesn't reject the AI-native company vision itself — he rejects handing the foundation to a single vendor.

Three pieces to the solution. Open-source models as a parallel option. Real competition from other labs. And critically, choosing Claude-Tag-style providers that let you own your business context — you own the context, not them. With that design, you aren't locked into paying Anthropic prices for every token; a multi-model, multi-provider strategy gets you the best token at the best price. The closing read is a doubled one. What started as a Slack convenience is Anthropic's bid to own all knowledge work — that's a bad thing. But AI and humans collaborating inside companies, with AI living and breathing in the workflow rather than as a separate chat tab, is genuinely the future — that's good. What's missing is more competition. That's where his optimism lands.