Skip to content

Example: the Habitat-Thinking estate (real)

Real assessment — run 2026-08-15 across the Habitat-Thinking estate.

This is not a constructed example. It is an actual /ai-readiness-rollup over the reports that existed in techtalk/ai-readiness-assessment and Habitat-Thinking/ai-literacy-superpowers on that date, plus one subject that could not be reached. The result is not flattering, and it is published as it came out.

The other examples on this site are synthetic — built to show the report shape. This one shows what the instrument actually says about the estate that builds it.

The manifest

version: 1
team: habitat-thinking

subjects:
  - id: ai-readiness-assessment
    path: ./ai-readiness-assessment
  - id: ai-literacy-superpowers
    path: ./ai-literacy-superpowers
  - id: leap-companion
    path: ./leap-companion

report:
  output: assessments/

Paths are shown resolved from the scope root — the directory containing .habitat/, which held all three clones. The run itself predated that rule being written down, and the ambiguity it exposed is what spec 0012 closed.

Coverage ledger

Subject Status Report date Age Note
ai-readiness-assessment degraded 2026-06-03 73d No assessment-summary block; recovered from prose
ai-literacy-superpowers incompatible, stale 2026-05-11 96d Produced by /assess (a different instrument) — no Habitat Maturity Profile, no gap
leap-companion unreachable Not found in the organisation

Assessed 2 of 3 named subjects; 1 of 3 can enter the dimension matrix.

Read that second number carefully. Two subjects had readable reports — but only one of them was measured against this model.

Dimension matrix

Subject AB AI WF OM TP OR OA HR WP AC AD Te Ob Gv
ai-readiness-assessment 3 3 4 3 3 3 3 3 3 2* 3 2* 2* 4
ai-literacy-superpowers
leap-companion

AB Agent behaviour · AI Agent input · WF Workflow · OM Operating model · TP Teams provide · OR Output role · OA Output artefact · HR Humans review · WP Work patterns · AC Agent composition · AD Agents… · Te Testing · Ob Observability · Gv Governance

A one-row matrix is not a matrix. It is a single assessment with extra formatting, and the report says so rather than dressing it up.

Gap table

Subject Habitat maturity mean Cognitive level Gap Regime
ai-readiness-assessment 2.93 4 +1.07 Ambition outpaces enablement
ai-literacy-superpowers 5 not computable
leap-companion

ai-literacy-superpowers has a cognitive read — L5, Sovereign Engineering, with all three disciplines at 5 — but no habitat maturity mean, because its report was produced by an instrument that does not measure the fourteen dimensions. A cognitive read without a habitat mean cannot yield a gap. The number is genuinely absent, not withheld.

Spread

Not computable.

A spread is a range across subjects, and a range needs at least two comparable subjects. This estate currently produces one gap.

Presenting +1.07 as "the spread" would manufacture precisely the finding a portfolio view exists to produce. The honest statement is that the estate cannot yet produce it — and that is a finding of its own, just not the one anybody wanted.

What would produce a spread: run /ai-readiness-assess in ai-literacy-superpowers. It is a mature habitat with a long assessment history; it has simply never been read against this model.

What the one comparable subject says

ai-readiness-assessment reads +1.07 — Ambition outpaces enablement: cognitive L4 against a fourteen-dimension habitat mean of L2.93, held back by Agent composition, Testing and Observability, all at L2.

That is the instrument reporting incoherence in its own repository. The report is 73 days old and predates a substantial amount of work on the testing dimension, so a re-run would likely read differently — which is itself the point: a 73-day-old report is evidence about 73 days ago.

Superseded the next day — and the prediction held

This roll-up was run on 2026-08-15 against the reports that existed then. On 2026-08-16 the repository was re-assessed, and the gap had closed from +1.07 (Ambition outpaces enablement) to +0.36 (Coherent), with the habitat maturity mean rising L2.93 → L3.64.

The roll-up above is left exactly as it ran. It is a dated reading, and rewriting it would destroy the thing it demonstrates. The point stands more sharply now than when it was written: the estate was being summarised from a report that was already out of date, and the coverage ledger said so.

Portfolio regime

Not asserted.

Regime selection needs provenance proportions and a spread. With one comparable subject and no declared habitats, there is nothing to classify. A regime is a claim about an estate; one repository is not an estate.

If forced to characterise it informally: two mature repositories, no shared habitat, no lineage between their control surfaces, and a wide difference in cognitive read (L4 against L5). That shape is Islanded — but the report does not assert it, because the evidence required to assert it is not there.

Split ceiling

Not reported. A common-weak dimension is one weak across two-thirds or more of the assessed subjects. With one subject in the matrix, every weak dimension is trivially "common" and the split carries no information.

The weak dimensions of ai-readiness-assessment — Agent composition, Testing, Observability — are that repository's own backlog until there is a second comparable subject to tell local weakness from shared weakness.

What this estate should do next

Assess ai-literacy-superpowers against this model. One command in that repository turns a one-row matrix into a two-row one and produces the estate's first real spread. It is the single highest-value action available, and it costs about ten minutes.

Re-run ai-readiness-assessment. Its report predates its current testing posture by 73 days and will be stale in another 17.

Resolve leap-companion. It is named in the manifest and cannot be found. Either it has moved, it is private, or it should come out of the scope. Leaving it as unreachable is honest but it is not a resting state.

Why this example is here

The design argument for a portfolio view is that the spread of gaps across an estate is a finding no single-repository assessment can produce. This example does not produce a spread.

That is worth publishing. The instrument was pointed at its own estate, and what came back was: one subject comparable, one measured by a different instrument, one missing, no spread, no regime, no split ceiling. Every one of those is the machinery refusing to invent a number it does not have.

A synthetic example in this slot would have shown a tidy three-row matrix and a satisfying range. It would also have shown nothing about whether the instrument tells the truth when the data is thin.

This example changed the tool

Running it surfaced two real defects. The roll-up had no coverage status for a report that parses but was produced by a different instrument — everything was either degraded or unparseable, and ai-literacy-superpowers is neither. And nothing stopped a single-subject estate from having its one gap reported as a spread. Both were fixed before this page was written; the incompatible status and the two-subject spread rule exist because of this run.

Rendered version

There is no HTML render of this report. The HTML portfolio is built around the matrix, and a matrix with one populated row would be a more impressive artefact than the data behind it deserves. When the estate has two comparable subjects, it gets a render.