ingest: Stanford SWEPR widening-gap study and AI-in-SDLC adoption pitfalls

Add two new sources with summaries, new concepts (developer-as-agent-manager,
review-is-the-new-bottleneck), new entities (SWEPR, Nikolai Sheiko), and a
query on the Stanford source; update related concept pages, overview, index,
and log.

wiki/sources/2026-07-14-gap-between-ai-users-irreversible.md

@@ -37,6 +37,7 @@
 - **Concepts:** [[personal-ai-operating-system]] · [[skills-as-memory]] · [[context-as-scarce-resource]] · [[connections-as-moat]] (human-relationship side effect)
 - **Related sources:** [[2026-07-14-skills-based-on-git]] (skills-as-memory from the engineering side — strong overlap) · [[2026-07-14-everything-we-knew-about-software-has-changed]] (markdown-as-skill ≈ G-brain markdown tier)
 - **Tools mentioned:** [[claude-code]], Claude Cowork, Codex, Flint
+- **External corroboration:** [[2026-07-30-stanford-swepr-widening-gap]] — Stanford SWEPR's difference-in-differences study measured the gap between AI-mastering and lagging teams growing 4.8% → 19% (4×) from April 2023 to July 2025; the title claim's first measured, non-practitioner support. (Traced via [[2026-07-30-stanford-widening-gap-source]].)
 
 ## Open Questions
 

wiki/sources/2026-07-30-rakes-in-ai-sdlc-adoption.md

@@ -0,0 +1,55 @@
+# Грабли во внедрении ИИ в SDLC — Rakes in AI Adoption in the SDLC (Nikolai Sheiko)
+
+#source
+
+## Source Metadata
+
+- **Date:** talk published 2026 (references events through Dec 2025); conclusions doc saved 2026-07-30
+- **Raw path:** `raw/sources/Грабли во внедрении ИИ в SDLC.md`
+- **Source type:** viewer's conclusions from a Russian-language YouTube talk (45:59) — https://www.youtube.com/watch?v=Nm3MsnngCJg — "Грабли во внедрении ИИ в SDLC — почему ИИ есть, а результата нет и как это лечить" ("why the AI is there but the results aren't, and how to treat it"). Not a transcript.
+- **Speaker:** [[nikolai-sheiko]] — AI-adoption consultant/practitioner (works with client companies on SDLC adoption; background otherwise unknown)
+- **Ingestion date:** 2026-07-30
+
+## Core Claims
+
+- **AI in development already delivers real gains, but people, companies and metrics throttle it by an order of magnitude.** The real jump started Dec 2025 (Opus 4.5 / GPT-5.2 + Claude Code / Codex); the SDLC collapsed into days/hours — but *not fully*: two human "red squares" remain — the **reviewer** (tasks queue at review) and the **planner/product person**. Developer time redistributed from "coding in the middle" to "planning on the left + verification on the right." See [[review-is-the-new-bottleneck]].
+- **Universal error #0: a developer is not a manager.** A good developer is 3–5 hours of CPU-bound focus on one feature; a good AI-developer is an IO-bound **manager of an agent-employee**, running several tasks in parallel. "If you launched Claude Code and sit watching it work — you're a bad employee." Not everyone can make the psychological switch, *and that's fine* — don't force everyone. See [[developer-as-agent-manager]].
+- **Measure completed tasks without rework** — never LoC, commit count or PR count (all trivially hacked; the European-outsourcing case shipped more PRs for a +1% gain because rework ate everything). A task counts only if it doesn't come back for fixes; also track task lifetime + rework time.
+- **Review with the agent, not instead of it and not fully by hand.** Manual-only review → queue → burnout → quality collapse; fully delegated review is the opposite error. Treat the model as a smart student: direct it, pose hypotheses, find problems together.
+- **Companies no longer need custom AI development.** Key quote: *"Come in, install Claude Code or Codex, configure everything, attach connectors, think about security — and it works better than any custom build."* Corollary anti-pattern: hiring an external configurator who leaves behind a "magic artifact" nobody owns — teams must configure their own tools; what a company should buy is a **teacher/curator**, not a setup.
+- **Agentic Evolution** — the key concept. Don't ask the expert to explain how they work (you get theory); instead take the new employee (the agent) by the hand through hard tasks, show it the rakes, then say: *"remember all of this and write the manual for the next one."* Verify a skill by launching a **context-free subagent** that must solve the same task from scratch using only the skill; the mentor agent watches what fails and fixes the skill. Without evolution you live on defaults; with it, vertical growth begins.
+- **Best practices matter more with agents, not less.** The "compaction curse": on a huge codebase the agent gathers context → window overflows → compaction → re-gathers → compaction again, and the task barely completes. The cure is locality, isolated modules with interfaces — **the codebase stores the context**. Use AST search instead of grep on colossal projects. And **embeddings/RAG over code do not work** — don't use them unless you understand *very* well why.
+- **Role futures:** a **Product engineer** emerges (answers *why* we build it this way, what to cut, what to ignore); users vs **Agentic Operations** (who tune the SDLC, feedback loops, prompts, skills) gradually split; AI eats **Intelligence** (action sequences requiring intellect) while **Judgment** (taste built over years, or domain expertise — oil & gas, medicine) stays human for now.
+- **Tokens get more expensive near-term, cheaper later. It's the wild west — experiment at full throttle while subscriptions are cheap**; the goal is to land in the top half of the Stanford chart. Don't chase every new tool: what Claude Code / Codex doesn't absorb within a couple of months is probably useless.
+
+## Key Evidence / Details
+
+- **Timeline the talk builds on:** METR study (Jul 2025, measured a *slowdown*, but methodologically contested) → Stanford study (Aug 2025, +20% speed but top-performers pull away — this is [[swepr]]'s research, cited independently) → Karpathy's tweet (Nov–Dec 2025: "80% Claude Code, 20% by hand") → "SDLC is dead" article (stages collapse into days/hours).
+- **Case 1 (frontend migration):** no feedback loop → give the agent a browser (Playwright / Chrome DevTools) to compare old vs new front; crunch → invest in planning (20 min minimum, hours are normal; every 10 min of planning saves hours; target one-shot implementation); Cursor with per-token billing → team economizes instead of experimenting (~30% dearer than subscriptions at the same level); training rollout: record sessions → expert reviews → *writes feedback, team fixes their own agent instructions* → focus on the top 2 performers (an hour with them is worth 10× more).
+- **Case 2 (European outsourcer):** more PRs, +1% — rework was the cause; the metrics table (don't: LoC/commits/PRs; do: completed-tasks-without-rework).
+- **Case 3 (large codebase):** the compaction curse; "agents mean we can drop best practices" is exactly wrong.
+- **Mini-cases:** the middle dev who sped the team up by tens of %, was refused a raise, and left for much more ("if you're that middle — think; if you're the manager — think twice"); the startup doing spec-driven development without knowing what it wants — build the UI first (even with an in-memory browser DB), click around, *then* plan.
+- **Do-tomorrow list:** close the feedback loop; write a skill that analyses your own sessions daily; automate it (Codex schedules / Anthropic routines); voice input (more context, and Russian gives more context than English); ignore tool churn.
+- **Q&A notes:** GLM (good, no vision), Kimi (works, has vision), Xiaomi's agent (interesting memory implementation; ex-DeepSeek Head of AI) as the Chinese-model trend; hardware deficit (H100s unrentable); speaker's prediction that AI will be classed as a cyber-weapon with biotech-style licensing.
+- **One-shot recipe:** feedback loop (must-have) + explicit real *goal* the agent self-checks against + skill verification via the context-free subagent.
+
+## Connections
+
+- **Entities:** [[nikolai-sheiko]], [[swepr]] (its Stanford study is the talk's central chart — "be in the top half"), [[claude-code]]
+- **New concepts:** [[review-is-the-new-bottleneck]], [[developer-as-agent-manager]]
+- **Corroborates:** [[2026-07-30-stanford-swepr-widening-gap]] (independent practitioner citation of the widening-gap result *and* of review-as-downstream-cost); [[solve-first-then-skillify]] (Agentic Evolution is its strongest formulation, plus the missing verification step); [[skills-as-memory]] (a second frontier-practitioner vote *for* the skills/evolution layer, against [[thorsten-ball]]'s dissent — and a vote *against* embeddings/RAG over code, siding with Konstantin in the skills-vs-RAG contradiction); [[context-as-scarce-resource]] (compaction curse; codebase-stores-context converges with Thorsten from the opposite direction); [[async-by-default]] (IO-bound parallel management as the working mode); [[enterprise-ai-reality]] ("no custom AI development needed" names the same managed-harness market; per-token billing shapes behaviour — the metered-vs-subscription split already logged there); [[harness]] (install-and-configure beats custom builds); [[leave-less-room-for-imagination]] (explicit goal + planning discipline).
+- **Complicates:** [[make-more-cheap-code]] (rework-free-completion as the metric is the org-level answer to "generation moved the cost downstream").
+
+## Open Questions
+
+- The cases are anonymous client anecdotes with self-reported numbers (+1%, "tens of %"); none are verifiable. Status: tentative.
+- "Embeddings/RAG over code don't work" is stated flatly with no mechanism given — strong claim, no evidence in the doc. Status: tentative.
+- Does the context-free-subagent verification protocol actually measure skill quality, or only skill *completeness* for one task? Closely related to the corpus's proposed skills falsification test ([[skills-as-memory]]) — this is the first source to describe running one.
+- The Dec-2025 "real jump" periodization is the speaker's own; the corpus's other timeline ([[ai-agent-evolution]]) slices eras differently.
+- Predictions (AI as licensed cyber-weapon; token prices up then down) are speculation. Status: tentative.
+
+## Change Impact on Wiki
+
+- Created [[nikolai-sheiko]] (entity), [[review-is-the-new-bottleneck]] and [[developer-as-agent-manager]] (concepts).
+- Updated [[solve-first-then-skillify]] (Agentic Evolution + skill-verification protocol), [[skills-as-memory]] (second practitioner vote for skills; anti-RAG-for-code), [[context-as-scarce-resource]] (compaction curse, AST search, codebase-stores-context), [[async-by-default]] (IO-bound manager evidence), [[enterprise-ai-reality]] (no-custom-AI-dev quote; external-configurator anti-pattern; token-price prediction), [[make-more-cheap-code]] (related link), [[2026-07-30-stanford-swepr-widening-gap]] (independent citation).
+- Updated [[overview]] (12 → 13 sources), `index.md`, `log.md`.

wiki/sources/2026-07-30-stanford-swepr-widening-gap.md

@@ -0,0 +1,48 @@
+# Stanford SWEPR — AI and the widening productivity gap
+
+#source
+
+## Source Metadata
+
+- **Date:** research presented through 2025 (chart data April 2023 → July 2025; deck Sept 2025); dossier compiled 2026-07-30
+- **Raw path:** `raw/sources/Stanford SWEPR - AI and the widening productivity gap.md` (screenshot: `raw/assets/G6g3O60bkAE05ZW.png`)
+- **Source type:** research dossier — a slide read first-hand plus public coverage of the underlying study; **not** a talk transcript. The corpus's first *quantitative outside study* (every other source is practitioner testimony or synthesis).
+- **Authors:** [[swepr]] (Stanford Software Engineering Productivity Research group), public face Yegor Denisov-Blanch
+- **Ingestion date:** 2026-07-30
+
+## Core Claims
+
+- **The gap between AI-mastering teams and laggards is widening, measurably.** Difference-in-differences analysis of 46 AI-adopting teams vs 46 matched non-AI teams: net productivity difference between top and bottom quartiles grew from **4.8% (April 2023) to 19% (July 2025) — a 4× increase** in ~2.25 years. Early quarters hover near zero or negative; the curve only takes off from mid-2024, then accelerates.
+- **Average net gain from AI is ~15–20%, not 10×.** Gross delivered code volume rises 30–40%, but rework (fixing AI-introduced bugs) eats roughly half.
+- **Gains are strongly context-dependent:** greenfield/low-complexity 30–40%; brownfield/high-complexity 0–10% and can be *negative*. Popular languages gain more than niche ones; gains collapse as codebases grow 10k → 10M lines (context-window limits, signal-to-noise).
+- **The cost moved downstream:** coverage reports +91% PR review time and ~2.6× rework in AI-heavy workflows — writing got cheaper, reviewing got more expensive.
+- **Proposed mechanism for the gap:** quality of AI usage beats volume; teams with clean, modular, well-tested code compound gains, teams with poor hygiene accumulate debt and lose trust in the tools.
+
+## Key Evidence / Details
+
+- Data: private Git repos, 600+ companies, ~100k–120k engineers, since 2022; ML model replicating a panel of expert reviewers, measuring *functionality delivered* (not commits/LOC). Methodology peer-reviewed (arXiv 2409.15152); the DiD result itself is talk-published only.
+- The slide: "Causal Impact of AI on Software Engineering Productivity: Difference-in-Differences Analysis," DID covariate balance < 0.25, 95% CI band, model-release markers on the time axis. Deck: "Will AI Replace Software Engineers?", AI Conference, Sept 2025.
+- Full numbers, links, and per-claim provenance in the raw dossier.
+
+## Connections
+
+- **Entity:** [[swepr]]
+- **Corroborates:** [[2026-07-14-gap-between-ai-users-irreversible]] — Allie Miller's "irreversible gap" prediction is this curve, asserted 18 months earlier from advisory experience; Stanford supplies the measurement. Also the team-level twin of [[levels-of-ai-usage]] (the gap grows between *rungs of mastery*, not between license-holders and others).
+- **Corroborates:** [[context-as-scarce-resource]] — the codebase-size finding (gains collapse toward 10M LOC, attributed to context-window limits and signal-to-noise) is the corpus's first outside quantitative support for context as the binding constraint.
+- **Complicates:** [[make-more-cheap-code]] / [[code-as-throwaway]] — +91% PR review time and 2.6× rework externally confirm that generation moved the cost to review/verification, which is Theo's premise; but the *negative* gains in complex brownfield code sharpen the honest caveat that "code is cheap" holds least where most code lives.
+- **Independently cited by a practitioner source:** [[2026-07-30-rakes-in-ai-sdlc-adoption]] builds its stakes on this study ("the Stanford study, Aug 2025: +20% speed, but the top performers pull away — the goal is to be in the top half of the chart") and anecdotally mirrors its downstream-cost finding (the +1%-despite-more-PRs rework case; review as the bottleneck — see [[review-is-the-new-bottleneck]]).
+- **Queries:** [[2026-07-30-stanford-widening-gap-source]] (how this source was traced), [[2026-07-28-webinar-theses]] (thesis T7 gains its number)
+
+## Open Questions
+
+- Does the 46-team DiD analysis ever get a peer-reviewed publication? (Claim status upgrades when it does.)
+- What exactly distinguishes the mastering teams — tooling, codebase hygiene, or skills/context practices? The proposed mechanism ("quality of usage") is asserted in talks, not decomposed. Directly relevant to whether the webinar's skills rung is *the* differentiator. Status: tentative.
+- The study measures teams of engineers; how far do team-level results transfer to the webinar's non-engineer audience? Status: tentative.
+
+## Change Impact on Wiki
+
+- Created [[swepr]] entity.
+- Added measured external evidence to [[levels-of-ai-usage]], [[context-as-scarce-resource]] and [[make-more-cheap-code]].
+- Upgraded the corroboration pointer on [[2026-07-14-gap-between-ai-users-irreversible]] to cite this page.
+- Added the "Stanford measured it: 4× in two years" note to thesis T7 in [[2026-07-28-webinar-theses]].
+- Updated [[overview]] (11 → 12 sources; first quantitative outside study), `index.md`, `log.md`.