Human × AI capability lens + design technique extraction

Verdict: Two design books — Refactoring UI and Universal Principles of Design — were mined into a two-track catalog (machine-executable rules vs. human-judgment calls), and that split was generalized into a durable model, Human vs AI Capability Lens: every capability is scored on two independent axes (Human, AI), each decomposed into five facets, each item landing in one of four zones — Own, Augment, Delegate, Low-leverage. The facet sets are locked and traced to the vault’s own Higher-Order Generativity vs. Judgment plus Naval; the ~250-row master scorecard and the per-doc score badge are specced and pending build.

What shipped

• Design Expansion — Reading & Resources — 19 sources to push the Norman-only Design notes multi-source.
• Design Two-Track Extraction and Human Track (35 judgment entries), extracted from Refactoring UI (3,690 lines of text) and Universal Principles of Design (200 principles, 14,503 lines). Both PDFs and their extracted text are staged in raw/sources/design/.
• Human vs AI Capability Lens — the model below.

The model

Two independent axes, not a balance bar: a capability can be Human 5 and AI 4, which is its own zone (Augment).

• Human — Discernment: Taste, Judgment. Origination: Originality, Specific Knowledge, Accountability.
• AI — Production: Fluency, Knowledge, Scale. Reliability: Verifiability, Autonomy.
• Mirror pairs: Originality↔Fluency (extrapolation vs. interpolation), Specific Knowledge↔Knowledge (private vs. public/average), Judgment+Accountability↔Verifiability (the uncheckable call vs. the checkable spec), Agency↔Autonomy (will vs. execution). Unpaired: Taste (human moat with no machine counterpart) and Scale (machine advantage with no human counterpart).

Decision ledger

Decided	Why	Ruled out
Human facets: Taste · Judgment · Originality · Specific Knowledge · Accountability	Traced to the Higher-Order note (“what stays human is OOD origination + accountable discernment”) and Naval	Perception, Creativity-as-such, Meaning — the vault never held them up as durable edges
Keep Accountability over Agency	The one facet categorically impossible for a model; Agency is contested in the sources and already captured elsewhere	—
Agency → AI axis as Autonomy, plus the meta-will above the human penta	It was doing double duty; on the AI side it is concrete and gradable	Agency as a sixth human facet
AI facets: Fluency · Knowledge · Scale · Verifiability · Autonomy	Durable dimensions of machine capability; mirror the human five plus Scale	—
“Order” (low/high) is a derived label, not a penta axis	It is the composite (polygon size / model tier), not a component; an axis would double-count and hide which facet makes a model higher-order	Low/High Order as a sixth AI axis
Badge: two independent value-colored bars (Human, AI) + Build/Learning relevance; Human expands to the penta on the published site	Honest about independence; degrades to static bars in Obsidian	The diverging Human⇄AI skew bar (falsely treated the axes as a tradeoff)
Master scorecard stays flat numeric columns (Human/AI/Build/Learning + Zone)	A penta per row across 250 rows is noise	Per-row radars
Graduation: Human-5 standouts (e.g. Wabi-Sabi) get their own page	High-signal only; the rest are rows	One page per principle

Flip condition: if the lens is read as a personal-growth compass rather than a durability map, Agency replaces Accountability in the human five — Accountability is automatic for a solo operator, so initiative is the scarcer edge.

Price

The lens is a dated snapshot, not a law. AI scores drift upward as models improve, so per-task AI grades are made against the current frontier and carry a last-audited date. The Reliability pillar (Verifiability, Autonomy) is closing fastest; the model is load-bearing only while that gap holds. A persistent cheap/fast tier (Scale traded against frontier reasoning) keeps “Order” a moving band, not a category that collapses.

Open questions

• Are the two axes truly independent, or does rising AI verifiability quietly lower the Human score over time — today’s Augment becoming tomorrow’s Delegate?
• Does Accountability belong on the human penta, or one level up with Agency as a precondition?
• Build the ~250-row scorecard against the current frontier — on what cadence does it get re-audited before the scores rot?

Provenance

This Cowork session. Grounded in the two PDFs read directly, the vault’s Higher-Order Generativity vs. Judgment and The Almanack of Naval Ravikant, plus a web check of Naval’s current taste / judgment / specific-knowledge framing.