AI Industrial Revolution ingest (2026-06-15)

Ingested Naval’s “The AI Industrial Revolution” (nav.al/industrial) into the agentic cluster. Anchor page created at The AI Industrial Revolution; fused into Agentic Engineering (+ Condensed), Vibe Coding, Software 3.0, Thinking Models, with light touches on Nonlinear Returns and Higher-Order Generativity. Full operation list in log.md (2026-06-15 ingest).

Decisions

Scope: agentic core + light adjacent, folded into existing pages. The article is rich but half of it is regulatory/healthcare/FDA politics that sits outside the vault’s domains. Kept the agentic-engineering and leverage material; left the regulatory and healthcare threads out. No new standalone concept pages beyond the anchor — the ideas had owners already.

Fused direct to wiki, skipping the workbench L2 draft. Per the request to fuse directly. Reversible via git; the trade is no staged review surface before the edits landed.

The planning question — surfaced, not settled

The request floated “foregoing learning about plans and PRDs because the models are becoming increasingly capable.” That directly contradicts the operating doc (“PRD first. Always.”). Surfaced it; Wedge clarified he is writing about the possibility of relaxing planning in the near future, not changing the rule now, and that the PRD stays important.

Resolution encoded, not as a rule change but as a distinction: planning-as-spec vs planning-as-ritual. The article supports it from both sides — Naval ignored “always use plan mode” on the bet the model improves faster than he can learn the tricks; Rauch reports models now plan unprompted and return trade-offs (“principal engineer”). What is rotting is the human hand-drafting the route. What is not is the spec: the problem, success criteria, scope, and which trade-off you actually want. The Condensed page already holds “specs are source code,” so the prediction is narrow: the authorship of the plan migrates to the model; the ownership of intent and acceptance criteria stays human. Logged as a dated tactic in Agentic Engineering, Condensed and as the lead open question in generated-questions.md. The operating doc is untouched until Wedge decides to move it.

The n=1 about-line — flagged, pending

Wedge likes reframing the about page around “I write about n=1 learning + encoding solutions.” The phrase traces to the article’s N-of-1 medicine thread (Sid from GitLab self-directing his own treatment ladder) — single-subject self-experimentation, ported to learning.

Conflict flagged: the High-Signal Front-Facing standard (set 2026-06-12, after ~a dozen struck hero lines) bans workshop words on the door — “topics in reader words, not workshop words… internal vocabulary belongs inside the pages.” “n=1” and “encoding solutions” are exactly that internal vocabulary, and the current about line already passed the bar. Decision: translate to reader words rather than ship the jargon or leave it. Candidate lines pending Wedge’s pick; about.md untouched until then.

Outstanding

• Pick the translated about-line (or decide to leave about.md as-is).
• Decide whether the planning-as-spec/ritual split ever earns a move in the operating doc, or stays a watched open question.
• The four ingest open questions in generated-questions.md (cheap-verifier tasks, generativity-vs-judgment handoffs, train-the-agent skills extraction).