11 Steelman 11 min read
The Case Against This Site
Seven serious objections to everything argued here, stated as well as we can state them — and an honest accounting of what survives.
A site that argues for judgment ought to demonstrate some. The fastest way to find out whether an argument is load-bearing is to attack it properly, so this page collects the seven strongest objections to the rest of the site. They are stated as their best proponents would state them, not as convenient approximations. Three of them we think are wrong. Four we think are at least partly right, and one of those is close to fatal.
1. Comparative advantage does not promise you a living
The objection. Optimists invoke comparative advantage: even if machines are better at everything, trade still makes it sensible for humans to do something. True and irrelevant. The theorem determines allocation, not price. Horses retained a comparative advantage after the combustion engine and their population collapsed by roughly ninety per cent, because their maintenance cost exceeded the value of their remaining tasks. If the cost of keeping a human functioning in a developed economy exceeds the marginal value of the tasks left to them, the theorem offers no protection whatsoever.
Our assessment: correct, and the site agrees with it. The work page states this directly. The distinction from horses is that humans are the consumers whose demand constitutes the economy and the voters who set distributive rules — which makes the outcome political rather than mechanical. That is not reassurance. It relocates the problem from economics to politics, where it is harder, not easier.
2. This is a coordination problem dressed as a self-help problem
The objection. Nearly every practice on this site is individually costly and collectively futile. You cannot preserve your judgment by scheduling unaided work if your employer measures output per hour and your competitor does not bother. Personal discipline against a structural incentive is a losing position held with dignity. The site knows the problem is structural and keeps prescribing individual behaviour, because individual behaviour is what a website can plausibly ask for.
Our assessment: largely correct, and the most serious objection here. Nothing on this site solves a coordination problem. Two partial defences. First, individual practice is what generates the constituency that eventually demands structural change — the people who protect junior training are the people who will argue for it institutionally. Second, some of it genuinely does pay off individually: the unaided-first habit costs minutes and demonstrably reduces commission error. But the honest position is that a personal protocol is not a policy, and this site is much longer on the former.
3. It is a class artefact
The objection. Every recommendation assumes slack. Schedule manual reps. Protect one unassisted domain. Leave the interstitials empty. Take a gap year and build a portfolio. These are instructions for salaried professionals with autonomy over their calendars. A warehouse worker on algorithmic dispatch, a gig driver, a contractor paid per resolved ticket — none of them can decline the efficient path. The people most exposed to automation have the least capacity to follow any of this, and a site addressed to the people with options is doing something other than what it claims.
Our assessment: correct and largely unanswered. The framing should be stated plainly: most of this is written for people with discretion over how they work, which is a minority and a privileged one. The extension to people without that discretion is not a personal protocol at all — it is bargaining power, labour organisation and regulation of algorithmic management. That is a different site, and arguably the more important one.
4. The historical analogies cut against the alarm
The objection. Writing, printing, calculators and search were each met with exactly these arguments about cognitive decline. Each time the predicted collapse failed to materialise and the tool became invisible infrastructure. Doctors do not diagnose from memory; nobody mourns. Pilots hand-fly far less than they did and aviation is by an enormous margin the safest it has ever been — the very example the judgment page leans on shows automation winning on outcomes. The base rate for this genre of worry is poor.
Our assessment: this is the best objection and it is only partly answerable. The base rate really is poor and intellectual humility is warranted. The distinction we would defend is narrow: previous tools automated specific capacities while leaving the general faculty of judgment untouched and, usually, more in demand. This one targets the general faculty. Whether that difference is real or whether it is what every generation believes about its own technology is not something we can currently establish. We think it is real. We would not bet heavily.
5. Preserving difficulty is nostalgia with a rationale
The objection. The desirable-difficulty literature concerns learning a skill, not performing one. Extending it into a general principle that friction is good is a leap the research does not license. Much of the atrophy evidence — the GPS studies, the search-engine memory work — is correlational, modest, and prone to the interpretation that the capacity was reallocated rather than lost. Deliberately doing things the slow way when a better method exists is a hobby, and giving it an epistemic justification does not change what it is.
Our assessment: half right, and it correctly identifies a real risk of puritanism. The claim we would keep is deliberately narrow: for the specific set of skills that are both generative and required to evaluate their own output, atrophy is not a free trade. Outside that set — which is most skills — the objection wins and the correct advice is to automate without guilt. The cognitive fitness page tries to state the test sharply enough to be falsifiable. It could be sharper.
6. Meaning-without-contribution may not survive contact with reality
The objection. The meaning page argues the work-meaning bond is historically contingent and therefore dissolvable. Contingent is not the same as optional. Many deep human arrangements are contingent and load-bearing anyway. Marienthal is cited as evidence for the latent deprivation model, but it is equally evidence that when work disappears, people do not construct alternative sources of meaning — they become apathetic and lose the capacity to use their time at all. The site reads that finding as a design brief. It could just as easily be read as a prognosis.
Our assessment: a fair reading of the same evidence, and unresolved. The difference between the readings is whether the Marienthal villagers lacked meaning or lacked the material and institutional conditions under which meaning can be constructed — they were also, crucially, destitute and stigmatised. We think the conditions mattered, which is why the civilization page insists that distribution and meaning are the same problem. But this is an interpretation, not a finding, and we are not entitled to more confidence than that.
7. Or the whole thing is overblown
The objection. Capability may plateau. Scaling returns may flatten, reliability may prove harder than fluency, deployment may stall for years on integration, liability, regulation and simple organisational inertia — as almost every major technology has. In a decade this site may read like the 1999 literature on how the internet would dissolve the nation state. The safest prediction about transformative-technology commentary is that it overestimates five years and underestimates twenty.
Our assessment: entirely possible, and the site is built to survive it. This is why the practices are framed as things that are good for you regardless: forming your own view before consulting others, keeping a domain deep enough to detect error, staying answerable for what you sign, protecting attention, being needed by someone. If capability plateaus tomorrow, none of that was wasted. That was a deliberate design constraint, and it is the closest thing the site has to a hedge.
What survives
Stripping out everything the objections above successfully damage, this is what we think remains defensible.
Everything else on this site is argument, and argument can be wrong. If you find a better objection than the seven here, it belongs on this page.
Starting with the opposition is a reasonable way to test whether the rest is worth your time. If the objections seem stronger than the responses, trust that judgment — it is the one the site keeps telling you to use.
Objection seven is the comfortable one for people who want to do nothing. Objection three is the comfortable one for people who want to feel critical without changing anything. Comfort is a signal worth interrogating.
The issue tracker is the right venue. Objections that survive get added, with attribution.