The Hamburger Nobody Ordered
The numbers come first. Yahoo Finance's AlphaSpace product team shipped over 100 new features in roughly two months. Sprints that used to run two weeks now run 24 hours — planning in the morning, shipping by end of day, retrospective at night. Users are spending 3x more time in AlphaSpace than the average Yahoo Finance user. In one instance, feedback arrived at 4:45 p.m. on a Friday; by 8:30 that same evening the fix was reviewed and live.
'In the old days,' said George Leimer, Yahoo Finance's product chief, 'that would have been, We can't launch it next week.'
The Sandwich Framework
The conceptual scaffolding behind these results comes from Arvind Narayanan, a Princeton computer scientist who keynoted the International Conference on Machine Learning in Seoul. Narayanan's model — the 'decide-execute-deliver sandwich' — breaks knowledge work into three layers: a decide layer on top (understanding what to build and why), an execute layer in the middle (the actual implementation), and a deliver layer on the bottom (integration, testing, long-term accountability).
AI agents, Narayanan argues, are compressing that middle layer. But the execute slice was only ever about a third of the job. The buns, he told the ICML audience, are 'arguably expanding as AI compresses the middle layer — because once building gets cheap, it becomes easier to start projects and harder to keep up with deciding and verifying them.'
Leimer's visual shorthand: work used to be a fat hamburger with skinny buns. Now the patty has shrunk to a sliver and the buns have swollen. More bun to chew over.
Role Lanes Are Dissolving
At Yahoo, the practical consequence is that traditional job-function silos are breaking down. Principal product designer Nick Lockington began prompting Google's Vertex AI directly, generating structured JSON logic for new features. An engineer on the team later noted that Lockington had effectively created a functional API by prompt.
'Nick became immediately the most leveraged engineer on our team, because he was working at the highest level of abstraction,' said David Grandinetti, a distinguished software apps engineer on the project.
On the team behind Yahoo Scout, the company's AI answer engine, one designer prolific in code commits was labeled a 'design engineer' internally. An engineer with a strong eye for usability became an 'engineering designer.' Yahoo's leaders say the point is not that job titles vanish, but that roles are increasingly defined by the decisions people own rather than the code they can manually produce.
What the Market Should Take From This
Narayanan's framework and Yahoo's execution data together point to something that free-enterprise readers should take seriously: AI is not eliminating human work — it is repricing it. The commodity is execution. The scarce input is judgment.
For businesses that embrace this shift, the payoff is compounding. Faster iteration cycles mean cheaper experimentation, which means more bets placed and failures discarded before they consume capital. Teams that once tested five ideas can now test a hundred. That is not a threat to productive workers; it is a multiplier for anyone who owns a decision, not just a task.
The firms that treat AI as a headcount-reduction tool rather than a decision-velocity engine will find themselves holding a very expensive bun with nothing inside. Capital rewards clear rules — and the clearest rule emerging from this data is that judgment, accountability, and creative direction are not going to be automated away anytime soon. The market has already voted.
