Process and Planning
5 entriesHow the work was planned, attacked, and corrected: the spines published before their chapters were drafted, and the register of claims that did not survive pressure.
- The Spine: What the Methodology Must Do, and the Part That Is Not Solved Published on: Published unfinished, on purpose. The review's last line is a commission: whether the missing composition can be built. That leaves one job, and the review's own finding sets the trap, since every field it surveyed ends by handing judgement to a human who can ask. A discipline validated by its designer's judgement puts a human back at the boundary the review said must be mechanised. The artefact's typing is inherited from the review rather than invented; failure is declared with two faces, since over-conservatism is a failure mode and not a safe harbour; the criteria are inherited from the ten fields' own stated limits, authored by people who never heard of this work. The discrimination problem, feared unanswerable without the designer vouching, has a route: the coverage and risk trade-off from the selective-prediction literature, which evaluates systems that decline without labelling any case as one that ought to have been refused.
- The Literature Review Plan Published on: How the gap gets discharged rather than asserted: the near-neighbour mechanisms the review confronts from their own specifications, the mechanism it must warrant, and the open-access sources it works from.
- Claims Under Pressure: A Register Published on: An open, living register of the review claims found to overreach or to rest on unexamined prior machinery, triaged by whether each could change the shape of the artefact. Tier 1: load-bearing premises corrected now (the novelty claim, policy classification, learning-to-defer, weakest-ancestor inheritance). Tier 2: six adjacent fields owed a targeted design-threatening review before the basis model is frozen. Tier 3: positioning debt carried forward, eyes open.
- The Tier 2 Reading: What Each Neighbouring Formalism Is Asked, Before Any Is Opened Published on: The methodology incurred a debt in public: six neighbouring formalisms may already hold semantics the basis model would otherwise reinvent worse, and the reading of them was owed before the model is fixed. Now the model is fixed enough to say what it actually needs, this spine schedules that reading as a targeted reading rather than a survey. It names, for each of the six fields, the one load-bearing part of the model it touches and the single adopt-or-reinvent question it is asked, and it fixes the stopping rule that keeps a targeted reading from becoming six chapters. It opens no source. The targeted reading is the one stone on the road with an unbounded input, and a masters that reads three fields deep for a week has turned into a doctorate without deciding to, so the reading is bounded here, in the open, before a single formalism is opened.
- Superseded: The Scenario Protocol as First Published Published on: The methodology committed the work to fixing its evaluation scenarios before the admissibility rules harden, and to making their expected verdicts answerable to something other than the designer. This spine sets the protocol that keeps those commitments, and it is published as an attackable skeleton before a single case is written, because a scenario set is the easiest place in the whole dissertation to cheat without noticing. It fixes where cases come from, how a verdict earns authority independent of the author, how the distribution is kept honest, how many cases and of what kinds, and what an expected failure concretely means. It names the method. It lists no cases, on purpose, because the moment it does the attack window closes.