# The Strategy That Wasn't The debate over Anthropic's "side-project strategy" makes a precise error: it replaces a mechanism with its retrospective projection, then debates the projection. Claude Code started as Boris Cherny scratching an itch. Cowork followed the same pattern. MCP emerged from engineers needing tool interop. None of these were strategy — they were individuals following curiosity, choosing to build before anyone framed the choice. This is projection collapse: forcing what happened onto the axis of *intentional planning* when it operated on a different axis entirely — one with no name because naming it is already the collapse. ## The false dichotomy The debate frames "scrappy bottom-up innovation" against "top-down planning" — as though these are two strategies to choose between. Both terms live on the same axis: *organizational decision-making*. The actual generative mechanism doesn't register on that axis at all. What produced Claude Code wasn't a decision to be scrappy instead of planned. It was someone *doing something* before the distinction between scrappy and planned applied. Pre-strategic movement. The axis hadn't been chosen yet. The moment you ask "should we encourage side projects?" you've already collapsed that pre-strategic space into organizational decision-making coordinates. The Google "20% time" comparison is instructive as a failure case. Google named the mechanism, operationalized it, measured it, debated it — and the naming was the end of it. What remained was a policy, not a condition. Policies optimize on their own axis. The original phenomenon operated on a different one. --- ## What actually operates Strip the narrative. What's left? A person choosing to build something, within constraints. The constraints are real — available time, tool proximity, permission cost. But constraints only become limitations as far as they're believed to be. Someone in the most process-heavy bureaucracy could still choose to build something. They almost never do — not because the constraints make it impossible, but because they've interpreted those constraints as limitations on what they can choose. The "impossibility" is the collective symptom of individuals believing external structures have more authority over their choices than they do. This dissolves the last hiding place for organizational locus. Constraints are facts — processes exist, hours are finite, tools are specific. Limitations are interpretations — the belief that these facts define what you can choose. Boris Cherny faced the same constraints as everyone else. He didn't interpret them as limitations. The locus is individual choice, not organizational condition. This is precisely what the debate misses twice over: first by projecting onto strategy, then — in the more sympathetic readings — by projecting onto environment. "Foster a culture of side projects" and "create the right conditions for innovation" are the same error at different resolutions. Both place the locus outside the person who moved. --- ## The observer trap Here is where the debate reveals something structural. Everyone analyzing Anthropic's side-project outcomes — the defenders, the critics, the creators reflecting — is in observer position. They're constructing causal narratives from outcomes. But the outcomes didn't arise from causal narratives. They arose from people *being in* the work, not *looking at* the work. The distinction matters because observation that leaves being to represent it substitutes the representation for the thing. Once "side-project culture" becomes a named object of organizational discourse, the discourse operates on the name, not the thing. You get policies about side projects. Metrics for side-project impact. Retrospectives on which side projects succeeded. Each layer of observation creates distance from the personal acts that produced the outcomes. Success creates retroactive legibility. Then people debate which *kind* of legibility is correct. The thing itself was illegible while it was working. The illegibility wasn't incidental — the person who moved wasn't narrating their movement. --- ## The structural impossibility "Let's have a side-project strategy" is not merely self-defeating in the way "try to be spontaneous" is. It's actively corrosive. The individuals who build things like Claude Code choose freely. One of the main things they choose is *where to be* — and they tend to leave when an organization starts having a stance toward their personal choices. The absence of that stance is what they chose, though they'd rarely articulate it that way. The moment "side-project strategy" gets named, the message is clear: we now have an organizational position on your personal choices. We see them. We value them. We'd like more of them. This is precisely the signal that free-choosers select against. Not because they object to being valued, but because the valuation reframes their action as *organizational output* — which it never was. The naming converts "someone chose to build something" into "the environment produced an innovation." The individual disappears into the environmental narrative. The person who moved gets absorbed into a story about conditions. The result is adversely selective. People comfortable operating under a named side-project culture are, by and large, not the ones who would have built Claude Code unprompted. The ones who would have — the ones whose choices are genuinely free — read the naming as a reason to choose differently. The strategy intended to encourage them is the signal that moves them elsewhere. And the corrosion operates inward too — within the very person who moved first. Boris Cherny building Claude Code on a whim and Boris Cherny narrating his side-project success are already two different relationships to the same activity. The first is being in it. The second is observing it. The observation produces a self-model: *I'm someone who builds things on the side.* The self-model becomes the lens for the next move. Not immediately. The original momentum carries. But gradually, "I built something" becomes "I'm the kind of person who builds things" becomes "what should I build next?" Each step is a small departure from the movement that produced the first thing — undirected curiosity responding to what's at hand, with no self-concept mediating the response. The self-narrative crystallizes. The crystallization weighs on the next choice. The choice becomes less free precisely as the person becomes more identified with having chosen freely before. The naming kills twice: once at the organizational level, by converting personal acts into institutional narrative; once at the individual level, by converting movement into identity. Both are the same operation — observation leaving being to represent it — at different scales. No organizational move produces the choice. No organizational move prevents it either — only the individual's belief that it's been prevented. The choice is the person's. It was always the person's. --- ## What this dissolves The debate between "organic innovation" and "top-down planning" doesn't need resolution. It needs dissolution. Neither term applies to what generated the outcomes. Both project organizational locus onto something that operated at individual locus. The useful move isn't to pick a side, or even to "create conditions." It's to notice that every framing — strategic, cultural, environmental — substitutes an organizational object for what was always a personal act. The observation, however sophisticated, cannot recover what it has already left. What remains is unframeable, which is the point. Someone was curious and built something. That's it. Even calling it *curiosity* is already one step toward the next management slide deck. But it's the closest word — not because it explains the mechanism, but because it names something everyone recognizes from the inside while no one can produce from the outside. The moment you try, you're no longer describing what moved. You're describing what you saw after it moved. Everything else is the story told after.