EP 04 - Warp Speed: When the News Cycle Can't Keep Up With the AI Releases

Anthropic dropped Claude Design. OpenAI dropped GPT updates. A new Chinese open model dropped. Figma's stock fell 7%. And it all happened in a single week. Harry's newsletter called it: "The news cycle can't absorb it. Enterprise buyers can't evaluate it. The development community can't benchmark it fast enough. And the next wave is already loading."

This episode is about what it means to make decisions — as a capital allocator, a business owner, or an executive — when the pace of change has outrun any normal framework for evaluating it.

Three things this episode gets into:

The model benchmarks don't matter as much as you think. Jeff Dean at Sequoia made the case that even if you made the model infinitely fast, you'd only get a 2-3x improvement end-to-end. The real gains — 10x, 50x — come from the harness, the business logic wrapped around the model. Anthony calls it "the harness is the business." It's the only place the competitive edge actually lives.

Figma is on the wrong side of the equation. Claude adding design wasn't about killing Figma — it was about going full-stack on programming. When AI can go from concept to rendered code in one session, the design-to-developer handoff layer disappears. Capital allocators should be asking: which of my portfolio companies are that handoff layer?

The Sequoia thesis: the next trillion-dollar company sells outcomes, not software. For every $1 of software, $6 goes into services. The AI-native firm that captures that $6 doesn't sell you a seat — it sells you a closed sale, a filed tax return, a completed legal brief. Harry and Anthony discuss what that means for VC models, PE rollups, and the companies actually positioned to make that shift.

Read the CO/AI newsletter at getcoai.com

Follow Anthony: @djabatt | Follow Harry: @hdemott