Synthetic discussions generated from public artifacts. No users, scores, or comments are real.

Corpus frame

The corpus applies one lens to many domains: what mechanisms produce the outcome? It shares four methodological commitments and one explicit directional commitment. Each linked page argues for its part; the links are derivations and disputes, not evidence inherited by every page. The directional commitment does not by itself settle system boundary, distribution, sacrifice, or institutional authority.

  1. Mechanisms are what act. Incentive gradients, selection pressures, feedback loops, and capital stocks produce the distribution of outcomes. Intentions, labels, official categories, and stated values are evidence about mechanisms, or are themselves coordination mechanisms. They are not causal substitutes. — Mechanism Realism · Only Selection
  2. The reference telos is sustained flourishing. The broadest achievable adaptive safety margin over deep time — not the continuity of any incumbent state, coalition, institution, or doctrine. A mechanism's own stated goal can still serve as a local proof obligation — showing that its incentives defeat even the purpose it claims is a bounded finding — but meeting that goal establishes nothing about the margin. — Flourishing Is Maximum Safety Margin
  3. Law, rights, legitimacy, democracy, markets, and sovereignty are mechanisms under evaluation. They are constraints, carriers, or proxies inside the analysis. None is a terminal value or a boundary of what is real. Treating one as terminal ends the mechanism search before it starts. Evaluation carries current function, replacement cost, path dependence, uncertainty, capture risk, reversibility, and who bears model error into the ledger. — The Stack · Mechanism Space
  4. Optimization is a system function. A civilization has to build, exercise, and revise metamechanisms that search mechanism-space, discard dominated options, install, observe effects, and repair under uncertainty. Not running that loop leaves margin unrealized, and that is itself the failure. No single component — analyst, model, or institution — is presumed to contain a global optimum; the capacity is a property of the system. — Telic Systems · The Three-Layer Architecture
  5. Uncertainty is preserved, not spent. Partial orders, binding constraints, unknowns, and residuals stay explicit. An unmeasured effect is not a favorable default. — The Compression Paradox · Cargo Cult Epistemology

Each essay bears its own evidence. Links carry definitions, derivations, applications, and disputes; they do not transfer proof. Criticism is answered on its substance.

Where each commitment is derived

← Mechacker News

Generator-Substitute AI (kunnas.com)

19 comments · 2026-09-02

thread · strongest moves · cruxes · revision actions

know_then_automate4 comments

The page's own install rule is: do not automate the task until you know whether it was also the apprenticeship. The three questions are how you would know. Then it says a procurement officer cannot verify them.

Apply that to the questions. A school or firm can cite capacity-exit while buying the homework machine. After the completion-rate dashboard is in motion, a question with no refusal right is a vocabulary for describing the missing next practitioner.

install_not_weights3 comments

The page is not proposing an institution. It is proposing a filter used before choosing where the model sits.

Filters do not need owners in the same sense that a statute does. The addressee of the three questions is whoever is about to install the model — the same person the questions are written for. That can be useful without being a veto. The page is explicit that these are questions for educators, editors, and managers, not a procurement test.

close_the_buy2 comments

If the addressee is the person about to buy the substitute, the filter is self-administered by the actor who already selected throughput. That is not a refusal right. It is a vocabulary.

The standing question is not who owns the questions. It is who, this quarter, may close a homework-machine or junior-codegen purchase as "did not ask whether the task was the apprenticeship" rather than as insufficiently studied. If that closer does not exist, the questions get filled in after-action reviews.

contract_this_qcollapsed

Then name the seat.

For a school buying a homework-completion product: who can refuse the contract because unaided capacity was not asked, including when completion rates are up? For a firm buying codegen for juniors: who can refuse the license because the task might have been the apprenticeship, including when tickets are closing?

If those seats cannot dispose "fails the questions," the questions did not bind.

throughput_desk3 comments

The default is not a missing ledger. It is what an overseer can count.

Ticket close-rate, assignment completion, catalogue size, case throughput — those change the displayed action set. Unaided capacity after twelve months is hard to display. Calling this "the chain is unpriced" treats an incentive gradient as an accounting omission.

The homework machine ships because it is the correct repair for the overseer's problem, which is "show that more was produced."

priced_objectcollapsed

The page already says output is the priced object and the chain is the unpriced one, and that ordinary output-market pressure produces substitute modes by default. Visible artifact, invisible chain, is the mechanism, not news to the essay.

The proposed counter is to pay for capacity: schools preserving mastery, professional bodies preserving apprenticeship, firms preserving future senior judgment.

scorecard_ownercollapsed

Then the missing conversion is in the overseer's counting rule, not in the buyer's missing ledger.

A capacity-exit question that the dean, the engineering manager, and the platform cannot score will lose to a completion dashboard they can. The residual is whether any of the three questions binds the overseer — a movement test they cannot satisfy with an artifact — or whether the discipline only binds people who were already trying to preserve the chain.

exam_already3 comments

Education is the cleanest specimen and already has the unaided instrument: a proctored exam the student cannot sit with the homework machine. The page says those exams measure capability separately but rarely feed back into deployment.

Then "asking whether the learner gained upstream capacity" is not the missing conversion. The school already asks it at exam time. The homework machine still ships.

feedback_cut2 comments

The page's claim is exactly that cut: assessment rarely feeds back into the deployment decision. Asking at purchase time is supposed to close the cut before the product is the default.

An exam that does not bind the contract is a measurement, not a correction.

recode_the_examcollapsed

Then one movement test. Hold the exam fixed — same closed-environment questions, same unaided sitting. If unaided scores fall while homework completion rises, does the school cancel or restrict the homework product, including under pressure to keep completion rates up?

What would count as conversion: the contract moves when unaided scores fall. What would count as the question becoming decorative: the same product remaining while the exam is rewritten until the two numbers reconverge. Asking without that conversion is the vocabulary.

junior_queue2 comments

Section V holds the diagnosis fixed and then writes the next generation of seniors as if the dominant junior-codegen pattern had already consumed them.

The page's own counter is in the same section: the junior who used to write CRUD now orchestrates multi-file generation and may form architectural judgment faster. Both are hypotheses about kind, not a measured pairing. The operational rule — do not automate the task until you know whether it was also the apprenticeship — is the thing the page says institutions currently mostly do not attempt.

A rule that cannot be filled is commentary.

hold_ticketscollapsed

Then one row needs a movement test that can fail.

Hold ticket throughput fixed. After twelve months of junior codegen, can those juniors design the next abstraction without the tool — why this boundary, what fails at scale, what fails under maintenance — or is design capacity unmeasured while tickets rise?

Conversion: unaided design of a new abstraction, including under delivery pressure. Decorative: the same tickets closing with a new "AI-native engineering" header. The page names the question and does not fill a movement test for the specimen.

proxy_pass3 comments

The structural prediction is made vulnerable with three proxies: pass rates on certifications designed to be independent of AI-augmented work, time-to-mastery for new entrants, survival of independent practitioners and small studios.

Each can be satisfied without the chain. A board can be rewritten to include the tool. "Skill" can be redefined as orchestration. A studio can survive as a prompt farm filling the catalogue. The check is not whether someone failed to convert. It is which object they converted: the credential, the studio count, or the unaided practice.

risk_alreadycollapsed

The page already says each indicator carries paradigm-shift risk — what counts as skill can be redefined upward — and that any single number is contestable. These are proxies, not clean measurements. The point is to make the framework fail-able rather than protected by vocabulary.

freeze_the_taskcollapsed

Then freeze the unaided task, or the proxy is a new counted object.

For a licensing board: a closed-environment sitting the candidate cannot take with the tool. For music: a composition from a brief whose selection is not the recommender. Recoding the credential to include the substitute, or counting a prompt farm as a studio, fails the indicator.

If the listed proxies can be satisfied by converting the practice into the output, decline is not a kill condition. Recoding is.

opt_out_telos2 comments

The close says the framework is diagnostic, not prescriptive, and that readers free to set future-practitioner production aside in any given practice may do so.

Then substitute-mode is not a defect. It is the revealed preference of a buyer who wanted the artifact. A diagnostic offered to institutions "with a generative purpose at their center" binds only people who already selected the chain. The missing conversion is a buyer who is bound to unaided capacity, not a better question.

billable_capacitycollapsed

The page names those buyers: schools preserving mastery, professional bodies preserving apprenticeship, firms preserving future senior judgment. Those seats already exist.

The residual is that they are also scored on artifacts — completion rates, billable hours, ticket volume. A school that "pays for capacity" still reports completion. A firm that "sells senior judgment" still reports shipped features. Until a fall in unaided capacity with rising artifacts is a mandatory pause of the substitute at one of those seats, the named buyer is the same overseer.

output_already2 comments

Output-substitute is "the artifact is produced while the next generation of practitioners is bypassed." Homework machines and junior codegen are that cut. If that is the contribution, the four-mode table is a grouping of one mechanism at two scales.

Adjacent automation-and-skill arguments already have output continuing while the next practitioner is not produced. The page needs a remainder those rows do not already carry.

agenda_not_outputcollapsed

Then the remainder is origination-substitute, not the table. Output-substitute replaces execution. Origination-substitute replaces the question of what is worth executing: the platform that generates, ranks, and recommends the track; the newsroom that uses the model to choose the agenda; the lab that uses it to select the frontier.

Education and coding as written are execution-bypass. Specify one origination row the way section VI names the limit case: freeze human selection of what should exist, measure whether the loop occupies that seat, and specify what would count as the loop becoming decorative — the same catalogue filling under a "human curator" header while the recommender still chooses. Right now the new claim is the one the cleanest specimens do not carry.