The Seat-Count Collapse

Mads Srinivasan

27 Feb 2026 — 1 min read

Howard Marks' new AI memo buries its most important observation in a quote from Claude: if 30-50% of structured knowledge work migrates to AI compute, that's $150-250B in annual labor value shifting.

The market already priced this in, violently. The SaaSpocalypse erased roughly $1T in enterprise software market cap in February. Anthropic launched Claude Cowork on February 3rd and Thomson Reuters dropped 16% that day. The S&P software index had its worst month since October 2008.

But Marks frames this as a labor story. It's also a pricing model story.

If 10 AI agents do the work of 100 people, you don't need 100 Salesforce seats. The per-seat subscription model that powered two decades of SaaS growth faces its first structural decline. Not slower growth. Structural decline. The revenue base shrinks even if the software stays essential.

The winners won't be the companies that bolt agents onto existing seat-based products. They'll be the companies that price on work completed, not humans employed. This means outcome-based pricing, consumption-based pricing, value-per-task pricing. The entire commercial architecture of enterprise software needs to be rebuilt.

Marks correctly identifies AI as labor replacement, not just labor saving. The investment implication he doesn't draw: every software company's pricing model is now a strategic vulnerability.

If you are not already reading Howard Marks' insights, you should start reading them at oaktreecapital dot com.

Claude Wins on Its Own Skills — GPT-5.4 Costs Half as Much

Anthropic publishes skills for Claude Code: structured task definitions that tell the model what to build and how. They're designed for Claude. I ran them on GPT-5.4 and Gemini 3.1 Pro too. Claude won. Claude Opus 4.6 averaged 0.825 across four Anthropic skills, against

Agent-First API Design (Agent Pickability 2/2)

Most companies treating agent-pickability as a distribution problem are solving the wrong layer. Getting into training data, appearing in MCP registries, reducing integration friction to five lines of code: that's necessary. It's also table stakes within eighteen months. The durable advantage is further down: whether your

The Agent's Default (Agent Pickability 1/2)

Microsoft gave Office to students for nearly nothing, not as charity but as customer acquisition cost arbitrage. Four years of muscle memory did the selling, and enterprise licenses followed for three decades. Something structurally similar is happening to dev tools, except the conversion funnel has collapsed from years to minutes

The Code Works Fine. Nobody Can Explain It.

Vibe coding gets you to a working prototype faster than anything before it. The part that trips teams up is the handoff. The pattern shows up consistently now. Someone spends a weekend prompting their way to a functional app. The demo lands well, leadership sees velocity, and engineering gets the

Read more

Claude Wins on Its Own Skills — GPT-5.4 Costs Half as Much

Agent-First API Design (Agent Pickability 2/2)

The Agent's Default (Agent Pickability 1/2)

The Code Works Fine. Nobody Can Explain It.