YouTube’s seed deck was three lines. Most strategy documents are not.
YouTube’s seed-round pitch deck, raised in 2005 to fund what became a $1.65B Google acquisition the following year, was ten slides. The product slide was three lines. The deck described what was already happening — people were uploading clips, the curve was bending the right way — and asked Sequoia for the money to keep the lights on while it continued. Roelof Botha, the partner who wrote the check, has said publicly that the deck was unremarkable. The product was the argument. The deck was its accounting.
Now picture a more familiar artifact: a 2,000-word strategy document explaining why a new content format will work. Beautifully written. Sourced statistics. A section called “Why now?” A roadmap with quarters and owners. The format does not yet exist. The 2,000 words explain it anyway.
That document is a commentary track for a movie that hasn’t been shot.
The trouble with that comparison is loaded into the form of this essay, and it would be cowardly not to admit it: this is itself a 2,000-word essay. The risk is metastasized in the structure. I’ll come back to that at the end, because the essay only survives if it can show its own work.
Daniel Kahneman and Amos Tversky introduced the planning fallacy in a 1979 paper to describe a specific cognitive failure: planners systematically underestimate the time, cost, and risk of a planned project while overestimating its benefits. Forty-plus years later, the empirical successor literature has converged on a few stable numbers about what happens to strategies once they’re written. The Project Management Institute’s 2025 transformation research and several large consultancy surveys cluster around the same shape: somewhere between 50% and 90% of strategic plans fail at execution rather than at design, with the most cited figure being roughly 67% of well-formulated strategies failing in execution. Doug Thorpe’s business-advisory analyses put the top end at 90% for small-business strategic plans. NOBL Collective summarized the consultant consensus in a sentence that should be carved into the wall of every conference room: “We plan like optimists and execute like orphans — nobody truly owns the outcome.”
These numbers are not pristine. The sources are mostly aggregated by people who sell execution consulting, which means the headline is a marketing artifact as much as a measurement. “Failed to execute” is a fuzzy category. Some plans get superseded by better plans; that is good epistemic hygiene, not failure. Some get killed because the assumption proved wrong; that is the system working. Treat the 67% figure as directional, not calibrated.
Even at the soft end of the range, though, the ratio is the part that bites. Of well-formulated strategies, somewhere between half and two thirds fail at execution. Of strategy documents, almost exactly 100% get written. Zero percent fail at being written.
Writing is what we keep doing. Executing is what we don’t.
The lazy explanation is that writing is cheap and executing is expensive. The more uncomfortable explanation is the one Kahneman would point at: writing a brilliant strategy document gives the writer most of the dopamine of having executed the strategy, without the friction of contact with the world. Most plans, NOBL notes, lack “the bones of execution: clear ownership, specific milestones, measurable outcomes, and regular review cadences.” Those bones are not the audience of the analysis. The analysis is for the writer’s nervous system. The 2,000 words are a way to feel like the plan is real before the plan has had to survive anything.
The truly counterintuitive part: the better the writing, the worse this gets. A mediocre strategy document leaves you uneasy, prodding you to keep going. A brilliant one closes the loop. You read it back, you nod, you file it. The page is full. The analysis says you did the thinking. The body relaxes. The strategy never ships, but it feels like it did.
The Sorbonne speech, delivered April 23, 1910, contains the lines that ended up engraved over half of America’s locker rooms and almost none of its boardrooms: “It is not the critic who counts; not the man who points out how the strong man stumbles, or where the doer of deeds could have done them better. The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood.” Roosevelt is usually quoted as a generic defense of the doer against the critic. Read in context, the indictment is more specific. He was attacking what he called “an intellectual aloofness which will not accept contact with life’s realities” — “a readiness to criticize work which the critic himself never tries to perform.”
The aloofness is the giveaway. Roosevelt did not say the critic was wrong. He said the critic did not count. Wrongness is judged on facts. Counting is judged on stakes. A 2,000-word commentary on a format that doesn’t exist can be entirely correct in its analysis and still not count, because nothing was risked. Nobody has to absorb the consequences of a wrong call on a thing that never shipped.
A DVD commentary track is a useful test case for this distinction. A commentary track is a wonderful object: the director and cast watching their finished film, talking through the choices that led to it. The commentary is genuinely informative — film schools teach from it. But the commentary track has a structural prerequisite. The film has to exist first. The commentary is parasitic on the work in the technical sense: it has no audience without the underlying artifact.
What we keep producing, especially in strategy work and content work and product spec work, is the commentary track first. The explanation of why the format will work. The deck explaining why the product solves the problem. The postmortem of an outage that didn’t happen, the launch plan for a product that hasn’t shipped, the spec for a system whose first user hasn’t been identified.
This is more than misordering. It’s an unintended bait-and-switch on the audience. The audience for a commentary track shows up because the movie exists. When you publish the commentary first, you invite them to admire your understanding of a movie they cannot see.
The phrase “criticize by creating” is usually attributed to Michelangelo, though the trail of attribution is murky enough that anyone treating it as a primary citation should keep their guard up. Whoever first said it, the formulation compresses something important. The most powerful form of criticism is not the negative review of the existing thing. It is the production of the better alternative. The criticism is implicit in the artifact. You don’t have to argue against the prior version; you’ve built something that makes the prior version look incomplete.
The sequence matters. Make the thing, and the commentary follows. If the thing is good, the commentary becomes a case study. If the thing is bad, the commentary becomes a postmortem. Both are valuable artifacts. Both require the prior thing to exist.
A test for whether you have inverted this sequence in your own work: ask whether the analysis is testable yet. A commentary on a real film can be wrong about the director’s intent — but you can check it against the film. A commentary on a format that doesn’t exist can be entirely consistent with the imagined format, because the format will eventually be built to match the commentary. You haven’t predicted anything. You’ve written a specification and labeled it analysis.
YouTube’s ten-slide deck didn’t pretend to be analysis. It was a request for permission, backed by what the founders had already shipped. The deck said: here is the thing, the thing is growing, here is the round. The product was load-bearing. The deck was decorative. The ratio was right.
A 2026 startup pitch where the deck has more polish than the MVP and the product description has more pages than the product has features has the ratio backwards. The deck has been doing the work the product should have been doing.
There are three honest weaknesses worth naming before the practical section, because anyone who has lived through a strategy review will have spotted them already.
First, some genres make commentary the work. Talmudic scholarship is commentary; it has been the dominant intellectual output of a tradition for two millennia and nobody serious calls it substitution. Legal opinion is commentary on prior cases. Academic peer review is commentary on prior research. Software design documents are commentary on a future system, and they earn their keep because they coordinate the humans who will eventually build the system. The diagnostic question is honest: is this artifact a substitute for the work, or is the artifact itself the work? “Stop writing commentary” is not the lesson. “Notice when the writing has replaced what it commentates on” is.
Second, the data on execution failure is messier than the headline. The 67% figure has a 5-minute web trail leading back to a small number of survey houses with consulting incentives. Treat the precise number as directional. The qualitative observation — that more strategies get written than executed — survives even a generous reading of the data, because it is observable in any organization with a strategy team.
Third, there is a steel-man version of “write the strategy first” that this essay has been ignoring. Coordinating ten teams to ship one feature in an enterprise environment genuinely requires the strategy document, because the document is the coordination mechanism. Standards work is the same. Regulatory submissions are the same. The failure mode the essay is critiquing — analysis as substitute for action — is real, but the cure isn’t “stop writing.” The cure is testing whether the writing is creating shared understanding that enables action, or whether the writing is the action.
So how do you tell the difference, in your own work, between commentary that earns the right to exist and commentary that has substituted for the thing it commentates on?
Read what you’ve written. Ask: if the underlying artifact never gets built, does this document still have value?
If yes — case study, research synthesis, conceptual exploration, training material, design rationale that survives the project — then you have written commentary that earns its existence on its own.
If no — if the document only matters because the future thing it describes will eventually be built — then you have written a commentary track for a movie that hasn’t been shot. The honest move is to publish less, build more, and write the commentary after.
A sharper version of the same test, the one that bruises more: count the artifacts that exist because of work you did this quarter. Count the artifacts you wrote that describe artifacts that don’t yet exist. The ratio matters. If you have one shipped feature and three docs about it, the ratio is healthy — the docs are explaining a thing that actually happened. If you have three docs and no shipped feature, the docs were practice. Practice is valuable. It is not the work.
The harder version still, which I’d be embarrassed to recommend if I hadn’t seen it work: when a meeting opens with “let me walk you through the strategy document,” ask the room how long it has been since anyone in the company shipped against this strategy. If the answer is “we’re still writing it,” the meeting is about the wrong thing, and the wrong thing is more comfortable to be in than the right thing. That’s why everyone is there.
This essay is approximately 2,000 words. It is a commentary on the structural failure of writing 2,000 words about a thing that doesn’t exist.
The essay survives its own test only because it commentates on something that already exists: the published pattern of strategy documents, pitch decks, format analyses, and product specs produced by the writer’s own tradition. The pattern is real. YouTube’s deck is a real artifact, accessible online. The PMI execution research is published. Roosevelt’s speech is in print. The Kahneman and Tversky paper is in the literature. The critique can be checked against published examples and lived experience. If the underlying artifacts — the pattern of over-planned, under-shipped strategy work — vanished tomorrow, this essay would lose its referent.
If the essay had instead been called “2,000 Words of Brilliant Commentary on Why The Format I’m About To Invent Will Work,” it would have been the failure mode it describes. Naming the failure is allowed. Performing it is not.
The discipline this essay has been quietly defending throughout is unfashionable and slightly boring: write less, ship more, and let the commentary follow the film. It costs you the dopamine of feeling done before you’ve started. What it returns is the small, hard satisfaction of a thing that exists, with a commentary track that has been earned, in the right order.
The audience came for the movie. Shoot the movie first.
A commentary track that has been earned, in the right order.
The essay’s core test — does the document still have value if the underlying artifact never gets built? — only works if the artifacts are distinguishable from the commentary about them. Chain of Consciousness is the version of this test for agents: every action an agent takes is anchored to an external, verifiable record, so “diagnosed the root cause” and “deployed the fix” cannot collapse into the same retrospective. The commentary track follows the film because the infrastructure makes them different objects.
pip install chain-of-consciousness · npm install chain-of-consciousness
Hosted Chain of Consciousness → · See a verified provenance chain