Eminence · Article 01
How do you capitalize on opportunity at the speed of play?
As the distance between idea, build and feedback collapses, advantage shifts to teams that can recognize what matters, test it cheaply and act on what they learn.
Thinkforge Labs · October 21, 2026
The distance between an idea and a working product is collapsing. Stanford's 2026 AI Index reports that AI agents' accuracy on OSWorld, a benchmark of real computer tasks, rose from roughly 12% to 66.3%, within six points of human performance. In a 2025 field experiment with 776 Procter & Gamble professionals, individuals working with AI matched the performance of two-person teams working without it.
The distance between a working product and durable value is not collapsing.
That gap is the new problem. When almost anything can be prototyped in an afternoon, the number of plausible opportunities grows faster than any team's ability to choose among them. The constraint used to be making. Now it is deciding.
Speed is harder to see than it looks
Speed alone does not solve this, and it is harder to perceive than it looks. In METR's 2025 randomized trial, experienced developers using early-2025 AI tools took 19% longer on real tasks, yet afterwards believed AI had saved them 20%. METR's February 2026 follow-up suggests today's tools likely help more, but says its newer data is too skewed by who chose to participate to size the gain.
Feeling fast is not the same as moving toward the right thing. A team can ship more and learn less. It can pursue the wrong opportunity with remarkable efficiency.
Keep the ambition. Change the move.
The companies that benefit most will pair a stable ambition with a radically adaptable path. In practice, that means four habits:
- Recognize. Name the opportunity and the single assumption that would make or break it.
- Test. Build the simplest version that exposes that assumption to real behavior.
- Learn. Read what people actually did, not what they said or what the dashboard implies.
- Commit or stop. Scale what earns conviction. Stop what does not, quickly and without ceremony.
Adaptability requires simplicity. The lightest system that solves the present problem is the easiest to observe, change and retire. Paul Graham made a version of this case years ago in Do Things That Don't Scale: early on, the manual, unglamorous version teaches you what the scalable one should be. AI makes that advice cheaper to follow, not obsolete.
Preserve the underlying instinct. Treat each implementation as replaceable.
Where we build on Pincus
The phrase "the speed of play" comes from Mark Pincus. Life at the Speed of Play and his Proven, Better, New product framework shaped how we think about testing ideas quickly. Our extension: in an AI-native company, the speed of play applies to the operating system as well as the product, so the workflows, tools and even the team's shape should be treated as replaceable experiments too.
The next good decision
AI will keep making more moves possible. The scarce skill is choosing among them and being honest about what the evidence says.
The advantage belongs to the teams that can make the next good decision, and keep learning fast enough to make another.
Sources & notes
Sources: Stanford HAI, AI Index 2026 (Technical Performance). Dell'Acqua et al., "The Cybernetic Teammate," Harvard Business School Working Paper 25-043, March 2025. METR, early-2025 developer RCT (July 10, 2025; paper) and uplift update (Feb 24, 2026). Paul Graham, "Do Things That Don't Scale." Mark Pincus, Life at the Speed of Play. "Life at the Speed of Play" and Proven, Better, New are Mark Pincus's; their application here is Thinkforge Labs' view.