Argues there is a 'room' — workshop, studio, garage, practice space — where actual disciplinary capability is built, but the economy only sees finished outputs that walk out of it. The catalogue spans musicians' practice spaces, novelists' unpublished drafts, mechanics' home garages, and kids soldering at the kitchen table — unmeasured work that every measured output silently depends on.
The top comments translated the metaphor directly into software: after-hours debugging that taught you the kernel, Saturday-built toy compilers nobody used, Discord arguments that taught distributed systems, and 200 abandoned GitHub repos that taught what not to do. None of this appears in perf reviews, JIRA velocity charts, or quarterly OKRs, yet it's where engineers actually learn their craft.
Contends that when a senior engineer's calendar is 90% meetings and 10% 'deep work,' the workshop has already been closed from the outside — only a corridor remains. Identifies the 2023-2025 layoff cycle as the moment side-project time was eliminated and never restored, marking an acceleration in how fast the room is closing.
Acknowledges the lineage — Jane Jacobs on 'uncatalogued work,' David Graeber on those who keep things running while the org chart points elsewhere, Stewart Brand on how buildings learn through informal modifications. What's new isn't the diagnosis but the velocity at which economic pressure is shutting the workshop down.
A post titled "The room the economy can't see" went up on wilsoniumite.com on June 19 and climbed to 161 points on Hacker News inside a day. The piece is short, more essay than analysis, and its core move is a metaphor: there is a *room* — a workshop, a studio, a garage, a practice space — where the actual capability of a discipline is built. The economy can't see into that room. It only sees the things that walk out of it.
The author's catalogue of rooms is deliberately broad. The musician's practice space. The novelist's drawer of unpublished drafts. The mechanic's home garage. The kid soldering on the kitchen table. The point isn't that these rooms produce *unpaid* work — it's that they produce *unmeasured* work, the kind that no spreadsheet anywhere will ever account for, and the kind that every measured output silently depends on.
The HN comment thread did what HN does: dragged the metaphor into software. Within a few hours the top comments were about the after-hours debugging that taught you the kernel. The Saturday spent building a toy compiler nobody used. The Discord server where you learned distributed systems by arguing with strangers. The 200 abandoned GitHub repos that taught you what *not* to do — none of which appear in a perf review, a JIRA velocity chart, or a quarterly OKR.
This isn't a new observation. Jane Jacobs wrote about "uncatalogued work." David Graeber wrote about the people who keep everything running while the org chart points elsewhere. Stewart Brand wrote about how buildings learn — the slow informal modifications that the original architect never sees. What's new is the *velocity* at which the room is being closed.
The immediate pressure is economic. When a senior engineer's calendar is 90% meetings and 10% "deep work," the room has already been padlocked from the outside; what's left is the corridor. Side-project time was the first casualty of the 2023-2025 layoff cycle, and it never came back. The unspoken bargain — "we'll pay you a salary, you'll secretly do half your learning on weekends" — got renegotiated downward without anyone signing a new contract.
The slower pressure is AI. Frontier coding agents are, by 2026, genuinely good at the things the economy *can* see — the PR, the test suite, the boilerplate refactor, the bug fix with a clear repro. They are not good — and arguably can't be good — at the thing the room produces, which is *taste*. Taste is the residue of ten thousand small choices you made while nobody was watching and nobody was paying. Every leader currently asking "can the AI do the junior's job" is asking the wrong question; the right question is "if we stop hiring juniors, where does the next batch of senior taste come from?"
The wilsoniumite post doesn't quite say this part out loud, but the HN thread does, repeatedly. The replacement for the apprenticeship room isn't another room — it's a leaderboard. And leaderboards optimize for the legible. They cannot, by construction, reward the work whose value is that it taught you something you can't yet name.
There's a related thread in the responses about open source. Maintainers have been describing this dynamic for a decade: the visible artifact (the package, the binary, the GitHub star count) is a tiny fraction of the actual work (the triage, the security advisories, the long arguments in issues that taught five contributors how to think about a problem). The xz backdoor in 2024 was, among other things, an attack on the room — exploiting the fact that nobody is paid to keep the lights on in there.
If you run an engineering org, the actionable read is uncomfortable. The KPIs you use to justify headcount — tickets closed, PRs merged, story points burned — measure the corridor, not the room. They will, given enough quarters, optimize the room out of existence. This was already true; AI just compresses the timeline from decades to fiscal years.
Practically: protect unstructured time on calendars like you protect production access. "Friday is for the thing you've been curious about" is not a perk, it's a capability investment, and treating it as a perk is how you lose it the next time the CFO sharpens a pencil. Re-read your perf criteria for "senior" and ask whether anything in there could be earned without ever having been bored, stuck, or wrong about something for a long weekend. If not, you are hiring on outputs your AI tools already simulate, and you will be surprised when the simulation costs less.
For individual engineers, the read is the inverse. The market for your visible output is collapsing toward the cost of a GPU-hour. The market for your *judgment* — the thing the room produced in you — is not. Keep the room. Buy the parts on your own dime if you have to. The economy will continue to fail to see it. That has always been the deal.
The quietly subversive claim in the post is that the unmeasured economy isn't a *bug* in capitalism — it's a load-bearing wall. Every model of productivity growth from Solow forward has been smuggling in unaccounted-for labor from the room and calling it "residual." As AI eats more of the corridor, the residual is going to start showing up in the books as missing supply — of judgment, of taste, of the people who can tell the model when it's confidently wrong. The orgs that figure out how to fund the room, even though they'll never be able to defend the line item, are the ones that will still have senior engineers in 2030.
In my experience, the key element is slack.For a while, I did not have to worry about money, so I could afford to be generous with my time, and to work on things that are not financially viable. It did a lot of good. I've built so many useful things and helped so many people individually.Now, A
Here is where we’ve found those rooms: 1. In scouts gatherings at churches or campgrounds (zero religious connotation - just use of unused space during the week). 2. In play dates with other families. 3. In sports clubs that have regular practice. 4. In local libraries which are a fantastic resource
> You cannot sell “a place for lonely teenagers to feel less lonely.” The value is real, but it spills out sidewaysThe economic notion of value is wealth-weighted. This is very, uhh, unique -- other notions of value are generally not. Whenever the economic notion of value is saying that (obviousl
We shouldn’t expect markets to solve all problems. That’s why there are public institutions and government regulations, to take care of the issues that the markets can’t. That the room only exists due to public grants isn’t a flaw, it’s what a functioning society should be doing. What the economy sh
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
Very good and well-written. I wish we would also acknowledge that the market, by disincentivizing spend on stuff like this, is performing well. It is optimizing. The reason it matters to acknowledge this up front is so that we can, as the article says, get to the rule below all this which is that th