Upgrading Your Operating System

Twenty-two slides on the Framework Trap: why teams that follow every process perfectly still ship work nobody needed. The argument is not that frameworks are bad. It is that filling one in feels like thinking, and the feeling is much cheaper to obtain than the thing itself.

1. Part 1 The symptom

   Perfect execution, a velocity target hit every sprint, every artifact in place — OKRs, JTBD, UAC — and work that lands flat. I followed the recipe, measured the ingredients perfectly, but the dish came out bland. The talk rules out the easy explanation first: if the cause were talent or effort, the pattern would not survive good teams. So it has to be structural.

2. Part 2 The mechanism

   Kahneman’s two systems, applied to product decisions. System 1 is automatic, pattern-matching and cheap; System 2 is analytical, evidence- seeking and expensive. The point is not that System 1 is bad — it is that cheap wins by default unless something forces the expensive mode. That is what makes the trap silent: nothing breaks, nothing is late, and the artifact looks finished, so no signal ever arrives to say the thinking was skipped.

3. Part 3 The case

   CloudMetrics, anonymised. Eighteen consecutive sprints averaging 87 story points with a standard deviation under 5 — and over the same period MRR growth fell from 25% to 8%, with engagement and satisfaction going the same way. High velocity is not high value. Then the question that made the PM pause: of the 47 items on the roadmap, can you connect each one to a customer problem or a business outcome? The answer named a source and a scoring method for all 47, and a reason for none. Framework Trap × Velocity Trap = feature factory.

4. Part 4 The trap in reverse

   The symmetry that makes this more than an anti-process talk. A framework-averse team at Mekari — “agile with a lowercase a” — worked incredibly hard and never accumulated coherent value. They were drowning in their own flexibility. Introducing frameworks did not make them less creative; it freed them, because they stopped spending cognitive energy on coordination. Frameworks work beautifully when you are learning a domain, coordinating a team, scaling decisions, or working somewhere cause and effect are already known.

5. Part 5 The upgrade

   A decision rule, read top to bottom, stop at the first yes. Well-understood problem, still learning the domain, several people moving in step → follow. Genuine uncertainty, high stakes, a framework optimising the wrong thing → transcend. The masters are not the ones who know the most frameworks, nor the ones who reject them all. They know which to apply in which context — and the talk closes on why that is hard: your brain is 2% of your body weight and 20% of its energy budget, so it will hand work to the cheaper system every chance it gets. Cognitive ease and cognitive quality are not the same thing.

Are we in a feature factory?

Six questions from the deck, all of them ones the talk already asks — of the roadmap, of the users, and of the number the team is proudest of. Every one you cannot answer is the trap doing its work quietly.

1. Can you connect every item on your roadmap to a specific customer problem or business outcome?
2. Can you name the job your users are hiring your product to do, rather than the features they requested?
3. Do you know the single biggest barrier stopping them from succeeding at that job, not just what’s next on the backlog?
4. If you could ship only one thing this quarter, do you know what would create disproportionate value, rather than what you could ship fastest?
5. Is there a number your team is prouder of than velocity?
6. Over the last six months your velocity held steady. Did revenue, engagement and satisfaction hold with it?

A talk given 17 April 2026, adapted for the web in August 2026. The CloudMetrics case is anonymised. The two-systems material draws on Daniel Kahneman’s Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011); the framing of the trap, the case and the follow-or-transcend rule are this talk’s own. Part one of two — the diagnosis is complete, the practice is not yet given.