Weekly Advisor Update · 2026 · Week 19

Three grammar versions, one knowledge graph runtime, eight production deploys.

From the May 5 grammar review meeting to a fully linted v0.6 regen of the AHA Pressure Test, the week closes with the report studio writing against a real corpus knowledge graph and the brand-power score viewports running on live InfraNodus playgrounds.

Period
2026-05-01 → 2026-05-07
Author
Shur Creative Partners
Audience
Advisor call · Thursday 5 PM ET
Sites live
aha-pressure-v06 · aha-real-viz-v06 · shuriq-report-studio · shuriq-grammar-may05-postcall

By the numbers

One week, one lockdown loop.

Three grammar versions shipped end to end (v0.4 corpus-grounded → v0.5 live runtime → v0.6 review-locked). Eight production deploys on the canonical Cloudflare account. One 73-minute grammar review meeting drove the lockdown loop. Two new pre-publish linters now hard-block every leakage and inversion-rhetoric pattern that escaped into the v0.5 build.

8
production deploys
3
grammar versions in 7 days
30
rules codified in v0.6 contract
28
rules applied in the v0.6 regen
8
v0.5 violations rewritten
5
graph-playground viewports
2
pre-publish linters shipped
3
parallel agents dispatched today

Headline · 2026-05-07

AHA Pressure Test — v0.6 regen.

The first regen of an existing report under the v0.6 grammar contract. Twenty-eight rules applied. Eight previously-shipped violations rewritten in place. Both pre-publish linters sweep clean against the live HTML. The "starting point, not a verdict" inversion, the "10 anchors carry inline citations" leakage, and the prose-form Action Set sequencing are all gone — replaced with affirmative reframes, methodology asides, and an SVG swimlane.

aha-pressure-v06.pages.dev Live · 200 OK · main

What changed under the hood

A paired LLM-Primer artifact ships alongside the markdown — define terms, three depth levels, surface caveats. A reader can paste it straight into their own AI for a guided walk-through.

Headline · 2026-05-07

AHA Viz Hub — graph-playground brand-power score.

The static brand-power pentagon is gone. Each of the five dimensions — Awareness, Trust, Mission, Differentiation, Loyalty — now has its own viewport with a live InfraNodus playground embed focused on the clusters that score it. Scores are equal-weighted at twenty percent each, composite sixty point zero. Every dimension page also carries an "Open in InfraNodus" CTA so the graph stays reachable if a viewer's network blocks the iframe.

aha-real-viz-v06.pages.dev Live · 8 viewports · 200 OK

The five dimension viewports

The two anchor viewports

Lockdown loop · 2026-05-05 → 2026-05-07

Grammar v0.5 → v0.6 — review-driven hard-linter promotion.

Limore, Diana, and Nuri reviewed the v0.5 AHA Pressure Test on 2026-05-05. Seventy-three minutes of meeting time produced thirty new rules and elevated two persistent failure patterns to hard pre-publish linters: inversion rhetoric and agent-scaffolding leakage. Both rules existed before; both broke again in production. Documentation alone proved insufficient.

"This is the thing I asked to be rewritten because it's nonsense. This means nothing. None of this means anything." — Limore Shur, on the v0.5 prelude · 09:33
"This is a leak. It's called leakage. It's leaking the prompt in this section idea. I will fix it." — Jonny Dubowsky, naming the anti-pattern · 31:20

"Leakage" is now the canonical term in the v0.6 contract and code comments. R-LINT.1 blocks any "X, not Y" inversion construction in body copy. R-LINT.2 blocks the full leakage vocabulary — section-routing prose, citation-policy meta-prose, and prompt-leak phrases like "ten anchors" or "load-bearing claim".

shuriq-grammar-may05-postcall.pages.dev Team-facing post-call site · 200 OK

Encoded into the studio runtime today

Twenty-seven of the thirty rules are encoded in the report-studio source. The version constant moves from v0.5 to v0.6 in three places. Both linters land in a new validator module with an appendix-exemption rule and a single override path for direct quotes. A new database migration adds linter-status columns to the reports table; the publish endpoint now returns 409 if the linter rejected the body. Build, migrate, and deploy are queued for a follow-on session — code complete, deploy pending.

Runtime integration · 2026-05-02

v0.5 — the report studio writes against a real knowledge graph.

The headline infrastructure shipped Tuesday. Every claim in the network-analysis, structural-gap, what-breaks, AI-guide, and method-audit sections now cites a corpus node with a real betweenness-centrality value, a named cluster, or a structural gap pulled from the live InfraNodus graph. The pure-fetch HTTP client, the database migration, the production-secret INFRANODUS_API_KEY, the system-prompt grounding helper, and a new D3 topology component all landed in a single commit.

shuriq-report-studio.pages.dev Live · auth-gated · v0.5 in production

The proof point is the AHA seed. Its graph_ref points at shuriq-aha-pressure-real-2026-05, the corpus graph built from the curated 1,959-word AHA briefing. Today's v0.6 regen is the first live test that the runtime actually produces meaningfully different output than the v0.4-without-graph version. It does.

Forward project scaffold · 2026-05-02

System architecture — Lyapunov, Totem Protocol, telemetry.

A starting point for the long-running design pass that follows the grammar lockdown. Twenty-five capabilities and seven team pages cataloged. Three are deep — pressure-test, knowledge-graph creation, and a brand-new Lyapunov stability analysis. The rest are stubs queued for a multi-session design loop alongside the stack-ranking grammar work.

Capabilities

Twenty-five capability pages

Three deep, twenty-two stubs. Pressure-test is the canonical reference; knowledge-graph creation is the load-bearing intelligence-layer capability; Lyapunov stability is a multi-session design pickup.

Teams

Seven team pages

One deep — the bi-report production team. Six stubs for viz hub, cold read, motion graphics, ontology design, negative space, and Lyapunov stability.

Lyapunov layer

Capital efficiency under stress

The control-theoretic stability layer rides on top of the five-dimension stack rank. Body copy reads as plain capital efficiency; the math sits in the appendix.

Totem Protocol

Declarative analysis scripts

Encoding forms and logic circuits scaffolded — JSON-LD, SPARQL, vault frontmatter, declarative scripts. Iterative grammar refinement loop with multi-hour human meetings.

Telemetry

Per-run JSONL log shape

Schema landed; dashboard config queued. Weekly cost and quality synthesis report runs against this once the pipeline ships.

v0.7 schema

Account → project → report hierarchy

Today's commit. Multi-tenant data model lined up for the studio interface integration that closes the v0.5 to MVP gap.

Other shipped this week

Earlier in the week.

2026-05-02 Corpus-grounded Opus 4.7 rebuild

Four AHA review sites produced from the curated corpus before the studio runtime caught up. Aha-real-viz, aha-real-sales, aha-real-pressure-bridge, aha-real-pressure-labor. All four are still live as the comparison set against today's v0.6 work.

2026-05-01 MicroCo competitive intelligence — pipeline audit

Three of four pipeline layers were producing stale or no output despite appearing to run. The Oxigraph semantic layer was unfrozen for weeks ten through sixteen. The weekly scheduler's max-turn limit was raised from fifty to two hundred. Eleven inversion-rhetoric constructions were swept on a user catch. The W17 editorial and viz hub shipped to production.

2026-05-01 V3 prep tabbed reader · pipeline runtime fix

A tabbed reader site for the V2 grammar walkthrough with five inline SVG constellation animations on the topology-caption tab. Two Cloudflare Pages sites deployed: shur-grammar-v2-brief and shur-v3-prep. Same day, the Playfair display Q-tail bug was caught and fixed on the pipeline runtime site — every "ShurIQ" wordmark in serif headings now wraps in the locked sans span.

2026-05-01 Constellation aesthetic refresh

The chord-graph constellation moved from dark canvas to the editorial-tan canvas across every Visual Grammar demo. The seven-color musical-theory palette stays; edges flipped from white-on-dark to ink-on-tan. Constellation work now reads on the same warm background as every other ShurIQ deliverable.

What's next

The forward queue.

This evening
Studio runtime v0.6 deployBuild, migrate, and ship the v0.6 grammar to the live studio. Code complete, deploy queued. Post-meeting follow-on.
Next 1–2 weeks
Stack-Ranking grammar lockdownA second multi-session refinement loop comparable to v0.4 → v0.6 — three review meetings. Resolves the open question about the fifth stack-rank dimension name. Forward project owned by R-METHOD.2.
Multi-session
Lyapunov stability layerPilot the methodology with AHA and Commonwealth Fund as the natural pair. Two brands, one vertical, eight to twelve quarterly snapshots before generalizing.
Multi-session
Totem Protocol scriptingAuthor the encoding-forms, logic-circuits, and worked-examples documents. Iterative refinement with Diana, Limore, Nuri. Ship the protocol versioned like the report grammar.
Stage 1b
Studio interface integrationSingle front-end for both report generation and stack-rank generation against the same knowledge graph. Closes the v0.5 to MVP gap.
After ~10 engagements
Competitor-set recalibration moduleCross-vertical pattern data threshold. Forward project carry-over from 2026-04-30.