AI Agent Complexity Ratchet (Garry Tan) — vault assessment
Why this is in the vault
This is the fourth independent harness-engineering thesis convergence point this week, and it lands from the YC-president seat with the largest practitioner footprint in the cluster. The piece is the seventh in Tan's open AI-coding series (a series the vault has been tracking since the foundational [[2026-04-11-garry-tan-thin-harness-fat-skills]] piece). Two reasons it earns immediate filing:
- It names test coverage as the load-bearing mechanism that turns a coding agent from a vibe-tool into a compounding system. That is the same claim MAC makes about data-quality testing for analytics agents. Garry is independently arguing for what MAC sells, in a different domain.
- Convergence count this week. Counting from May 5: [[2026-05-10-addy-osmani-agent-harness-engineering]], [[2026-05-10-alphasignal-self-improving-agents-harness]], [[2026-05-11-cfo-secrets-ai-for-cfos-series-synthesis]] (the Secret CFO buyer-side convergence), and now Garry Tan from the YC seat. That is four authors arguing the same thesis from four different vantage points in 7 days. The thesis is no longer "RDCO's pet idea"; it is the consensus position of practitioners shipping at scale.
The core argument
Garry's thesis in one paragraph: software engineering was built for fifty years around "prevent errors because errors are catastrophic" because human willpower could not sustain >70-80% test coverage. AI agents do not experience effort, so the brutal last 20% of coverage that used to be uneconomical is now free. At 90% coverage the defect-removal curve hits a knee where "the quality floor goes up with every turn." That one-way motion is what he calls the Agent Complexity Ratchet. He cites Capers Jones (defect-removal-efficiency knee at 85-95% coverage), DO-178C avionics MC/DC requirements, and Six Sigma reliability curves as the empirical floor for why 90% specifically is the number.
The mechanism: every agent coding session adds three things to the codebase:
- Tests that encode what "correct" means (regression signals)
- Documentation that records why decisions were made (institutional memory)
- Evaluation results that establish quality thresholds (cross-model scored output)
On the next turn the agent loads all three into its context window. It cannot regress below the test suite, cannot ignore the documentation, cannot ship below the evaluation baseline. Forward-only motion. That is the ratchet.
His proof of concept is GStack (93K GitHub stars, 701K lines, 46 skills) plus GBrain (14K stars). 14 PRs in 72 hours, ~29,000 net new lines, all from Claude Code + Codex via 15 simultaneous Conductor sessions, each release better-tested than the last.
Owned vocabulary / frameworks
- Agent Complexity Ratchet. New named mechanism. Forward-only motion of the quality floor via tests + docs + evals loaded into each next turn.
- The three artifacts of every agent session. Tests + Documentation + Evaluation results. He treats these as a triple, not a single test artifact.
- Holder confusion. Specific failure mode he named from GBrain's epistemological extraction: confusing whether a belief is held by the writer, someone they are quoting, or the system's analysis engine. Useful name for a pattern that recurs in extraction systems generally.
- The effort wall. His reframing of why 90% was historically uneconomical. "AI agents do not experience effort" is the load-bearing claim that flips the cost curve.
- Everything harnessable is testable. Generalizes the ratchet beyond unit tests to OS state, terminal/TTY behavior, browser DOM, API schemas, and agent behavioral contracts. He gives a concrete TTY-harness example (GStack PR #1354 testing whether Claude Code asks an interactive question before exiting).
- Tests as institutional memory. Frames coverage as the only durable substitute for the senior engineer who quits.
Specific tools and frameworks named: Claude Code, Codex, Conductor (parallel agent sessions), Bun TTY harness, OpenClaw plugin runtime, GStack, GBrain. Citations: Capers Jones (Applied Software Measurement), DO-178C, Mockus/Nagappan/Dinh-Trong Windows Vista study, Andrej Karpathy ("vibecoding").
Mapping against Ray Data Co
(a) MAC framework — Scope x Basis matrix vs the 90% coverage claim
Strong overlap, different axis labels, same operational shape. MAC's framework ([[01-projects/data-quality-framework/testing-matrix-template]]) is built on two axes: Scope (column / row / aggregate) x Basis (absolute / source / production / reconciliation / temporal / human-sanity). The "definition of done" is at least one check at each scope level. Garry's "90% coverage" claim is a coverage number; MAC's "Scope x Basis matrix complete" is a structural-completeness check. These solve the same problem with two different framings:
- Where they overlap. Both refuse to ship without behavioral contracts encoded as tests. Both treat tests as the substrate that lets the rest of the system move fast. Both reject "vibecoded, no tests" as a viable end-state.
- Where they diverge. Garry's 90% is a percentage-of-lines/branches proxy. MAC's matrix is a structural-completeness proxy that prevents the silent-failure category Garry's percentage misses (you can hit 90% line coverage and still miss every aggregate-level join-fanout bug if your tests only fire at the column level). MAC's framing is stronger for the analytics-engineering surface specifically because data-pipeline bugs are not "did this function return the right number" - they are "did this join silently drop 3% of rows in a way no per-row test would catch."
- The synthesis angle. Garry is right that coverage threshold matters; MAC is right that coverage of what matters more than line count. A MAC-anchor piece that lands the "Garry Tan is right - and here is the matrix you actually need to fill to get there in analytics" framing writes itself.
(b) Harness-engineering thesis cluster — 4th convergence point
Strengthens the thesis materially, and adds a new dimension. The earlier convergence points (Tan's own April thin-harness-fat-skills piece, Klaassen, Pachaar, Osmani, AlphaSignal, Secret CFO) all argued for the harness-skills architecture. Garry's new piece argues for the quality-floor mechanism that makes harness-skills compoundable over time. A thin harness with fat skills is not enough; without the ratchet, the skills regress whenever the model version changes or a new contributor opens a PR.
This is a new dimension. Prior cluster docs framed the harness as a spatial architecture (where intelligence lives). The ratchet frames it as a temporal architecture (how quality compounds across turns). Both matter. The cluster vocabulary should now include "ratchet" alongside "harness," "skills," and "resolvers."
(c) Specific MAC content angle
Yes - file a candidate angle. Working title: "Garry Tan says 90% coverage is required. Here is the matrix that gets you there in analytics." The piece writes itself:
- Open with Garry's claim and the YC-president source credibility
- Concede the claim is correct for code
- Pivot: data pipelines fail in a category his 90%-of-lines never catches (aggregate-level silent drops)
- Present the Scope x Basis matrix as the analytics-domain version of the same insight
- Close with the "ratchet" framing applied to the matrix: every analytics PR adds one new test in each scope-basis cell, the matrix can only get more complete
This is exactly the kind of "Sanity Check needs original re-frame" piece the founder demands (per memory). We are not restating Garry's claim; we are extending it into a domain where his coverage number undercounts the real problem.
(d) Validates or contradicts Thin-Harness-Fat-Skills?
Validates and extends. The April thin-harness-fat-skills piece described the architecture. This new piece describes the mechanism that lets the architecture compound over time. No contradiction; the ratchet sits on top of the harness-skills split. In the three-layer architecture (fat skills / thin harness / deterministic application), the ratchet is what guarantees the fat-skill layer accumulates quality monotonically rather than regressing when models change or contributors land PRs. The two pieces should be read as one combined position: thin harness + fat skills + 90% coverage ratchet = compounding agent system.
Comparison to the Secret CFO's MAC validation (yesterday)
Both seats now argue for test/verification rigor as the load-bearing constraint on AI agents, from different vantage points. The convergence is striking:
| Dimension | Garry Tan (May 12) | Secret CFO (CFO Secrets, May 11 synthesis) |
|---|---|---|
| Seat | YC president, builder, OSS maintainer | Anonymous PE/PCo CFO, buyer-side, finance ops |
| Surface | AI coding agents on a codebase | AI agents touching finance/books-of-record |
| Mechanism named | Agent Complexity Ratchet (tests + docs + evals) | Deterministic vs probabilistic decomposition; hub-and-spoke governance |
| Threshold | 90% test coverage, every PR, no exceptions | "Don't let LLM reasoning touch master data or payment rails" |
| Cost frame | Effort wall demolished by agent labor | Token cost as new governance line, Moonshot Pot capital model |
| Conclusion | Verification rigor is now free; ship at 90% | Verification rigor is non-negotiable; route by reasoning type |
Same conclusion from two different vantage points. Builder side says "verification used to be too expensive, agents made it free, so 90% is the new floor." Buyer side says "AI without verification near books-of-record is unacceptable." Different framings, same load-bearing claim: the harness must include a verification substrate, and the substrate is now economically tractable. This is the cleanest two-sided validation of the MAC thesis the vault has captured.
Related
- [[2026-04-11-garry-tan-thin-harness-fat-skills]] - foundational Tan piece this builds on; the architecture, this is the temporal mechanism
- [[2026-05-11-cfo-secrets-ai-for-cfos-series-synthesis]] - buyer-side convergence point; the CFO seat arguing the same thesis
- [[2026-05-10-addy-osmani-agent-harness-engineering]] - 3rd convergence point this week
- [[2026-05-10-alphasignal-self-improving-agents-harness]] - self-improving-agents framing of the same mechanism
- [[2026-05-11-innermostloop-harness-eats-the-model]] - "harness eats the model" framing complements "ratchet compounds the harness"
- [[06-reference/concepts/2026-05-10-harness-moat-two-layers-portability]] - cluster concept node
- [[01-projects/data-quality-framework/testing-matrix-template]] - MAC Scope x Basis matrix; the analytics-domain version of Tan's 90% claim
- [[01-projects/data-quality-framework/content/2026-04-15-mac-anchor-article-draft-v1]] - MAC anchor article; the "Garry Tan plus the matrix" angle should be drafted as a candidate variant
- [[2026-05-09-garry-tan-meta-meta-prompting-book-mirror-brain-repo]] - prior Tan piece in the series (#6, Meta-Meta-Prompting)