Build Report · 2026-05-12

What we shipped: the ShurIQ v0.7 data model, the AHA composite knowledge-graph viewer, and the AHA Business Model Canvas.

Three days of work landed in one Tuesday afternoon. The studio runtime now runs grammar v0.6 on production. AHA has four live knowledge-graph layers visible inside the report studio. AHA has its first Business Model Canvas overlay, applying the method we shipped for Fiserv.

Audience: Shur team (Limore, Diana, Nuri, Jonny) Distribution: internal only
01The arc of the work

From eight parallel sessions to one Tuesday plan.

The week opened with eight open Claude Code sessions in parallel and a synthesis pass that pulled them into one Monday plan. Thirteen decisions came out of that pass. The session log on 2026-05-12 worked through them in order.

The first decision was diagnostic: did the v0.6 grammar checkpoint on master actually run on the deployed ShurIQ Report Studio, or was the production site still on v0.5? We verified by redeploying from master and discovered the v0.6 checkpoint commit shipped uncompilable code, one line of unescaped backticks inside the system-prompt template literal. We fixed it, committed, redeployed. The studio is now provably on grammar v0.6.

The second decision merged the v0.7 schema work into master. The feat/account-project-hierarchy branch had carried five new D1 tables (accounts, projects, kg_layers, business_model_canvas, kg_deltas) for five days while the v0.6 grammar work landed on master in parallel. We merged with the predicted conflict in functions/lib/types.ts resolved by keeping both regions. Typecheck and build clean. Production main now serves the merged tree.

From there, the work moved into D5 of the two-week data-model plan: the composite knowledge-graph viewer. We built two new API endpoints, one new React screen, and extended the TopologyViz component with three modes (Concepts, Gaps, Clusters). The studio now has a new route at /projects/:id/composite where any project's knowledge-graph layers render in a single view with a mode toggle. AHA was the first project seeded with real layers.

The final move applied the Fiserv Business Model Canvas method to AHA. The Fiserv overlay (shipped 2026-04-24) was a 512-line BMC for a pre-launch initiative: 77% projection, 23% Observation. The AHA overlay is the inverse: 71% Observation, 28% projection. AHA executes its legacy mission at scale; the projection portion is concentrated in the strategic differentiation moves (AI guru, cardiovascular underwriter of employer-workforce health economics, youth pipeline, male-ally activation, consumer technology product). The BMC names where those moves stand today and what advancing each one requires.

02What shipped

Six concrete artifacts, all production, all verified.

Studio runtime
ShurIQ Report Studio on v0.6 grammar
Master commit b1013e9. Production main at jonny@weareshur.com. GRAMMAR_VERSION = v0.6 across wrangler.toml and both grammar.ts files. Auth gate intact.
Schema
v0.7 account → project → kg_layer hierarchy
Migration 0006 applied to local + remote D1. Five new tables. Four backfilled accounts: AHA, Hasbro, AGD, TruData. Master carries both the v0.6 linter work and the v0.7 schema in one merged tree.
commit 22b06a5a on master
Studio surface
Composite KG viewer route
New screen at /projects/:id/composite. Layer list on the left, force-directed graph on the right. Three modes: Concepts, Gaps, Clusters. Live InfraNodus fetch per layer, side panels per mode.
AHA data layer
Four AHA knowledge-graph layers seeded
domain_ontologyaha-business-intelligence (7 clusters, modularity 0.659). discourse_grammarshuriq-aha-pressure-real-2026-05. sbpi_scoreaha-brand-score. bmc_overlayaha-bmc-overlay (16 clusters, 150 nodes, 202 edges).
kg_layers WHERE project_id='default-aha'
Strategy artifact
AHA Business Model Canvas v0.1
413-line BMC overlay. 9 blocks, 123 mapped entities. 71% Observation, 9% Plan, 17% Recipe, 2% Missing + 1 inert. Five Chain Break narratives. Seven Strategic Insights including the Behavior-Change-Science Paradox.
projects/AHA/bmc/aha-bmc-overlay.md
Synthesis
2026-05-11 weekly plan with 13 locked decisions
Single source of truth for the week. The Decisions section, Execution log, Two-week data-model plan reminder with REDIRECT cascade, Thursday team deck reframe (subscription pitch folds into 3-month plan).
projects/totem-orchestration/synthesis/2026-05-11-weekly-plan.md
03The 13 weekly decisions Jonny locked

Each decision carries a downstream consequence visible elsewhere in this report.

01
ServiceTitan v05 is canonical.
The 17 surgical zinger upgrades from the 2026-05-08 zinger pass supersede the v04 theses update. The v04 14-statement Theses section ports into v05 as a new section. Live at servicetitan-millionaire-v05.pages.dev.
02
Pitch Support enters the studio archetype enum as the fifth value.
Reffer Pitch Support proved the archetype distinction from Pressure Test. Same subject, two registers (founder-facing warm vs investor-facing skeptical). The enum now reads: Editorial Brief Sales, Editorial Brief Delivery, Pressure Test, Cold Read, Pitch Support.
03
W19 MicroCo ships today under Stack Rank Grammar v0.2.
The cadence is the product. v0.3 formalization carries into W20. No drift on the weekly publication.
04
LLM Primer sidecar ships both forms.
Inline collapsed HTML under §18 for the human reader, plus the raw markdown link for AI ingestion. Closes the UX gap flagged on the v06 call.
05
Brand Power Score: v1 for NGOs, v2 for everyone else.
AHA stays on v2 by audience fit (equal-weight 20% × 50/50 present/opportunity). Future NGO clients default to v1 (NGO-weighted 15/35/30/10/10) unless an industry justification flips them.
06
The data-model plan REDIRECTS — drop D2, keep D5-D14 dates.
D2 was the cold-read kickoff. Four blocking inputs (business plan source path, GTM session location, Arc Capital portco name, executive-summary merge strategy) never arrived. D3 and D4 roll into this week as compressed work. D8-D14 dates hold.
07
Subscription pitch folds into the Thursday team deck.
The $9k commitment deadline (2026-05-14) is dropped. The pitch logic now lives as Section 4 of a Thursday "3-month plan to total self-sufficiency" deck for the Shur team. Candidate retainer set locked: AHA + Hasbro + Fiserv + MicroCo at $3k/mo.
08
Studio runtime: redeploy from master HEAD, verify in DOM.
Discovered uncompilable backticks in the v0.6 grammar checkpoint. Fixed (commit c7060c47). Redeployed (commit b1013e9c). Production now provably on v0.6.
09
Merge feat into master, do not rebase.
Merge commit 115f94a9 brings v0.7 schema into the trunk. The functions/lib/types.ts conflict resolved mechanically (keep both v0.6 linter types and v0.7 Account/Project rows).
10
AHA stack-rank surface: build this week against v0.2 grammar.
R-METHOD.1 in the v0.6 grammar requires every report bundle its vertical's stack rank. AHA constellation does not have one yet. Building once v0.3 formalization lands.
11
Codify both patterns as new skills.
aha-zinger-pattern and multi-voice-synthesis graduate from mem0 entries to named skills under ~/.dotfiles/ai/skills/. Next client brief inherits the workflow without re-invention.
12
Skills Catalog Google Sheet becomes a generated mirror.
Add --export-sheet mode to system/skills-catalog/.generate.py. The Sheet ID stays stable, the vault catalog is canonical, the Sheet stays fresh.
13
Anti-restatement encoded as R-LINT.3.
Each number used once per report, indirect reference after. The discipline that made the ServiceTitan v05 zingers land becomes a hard linter rule in app/functions/lib/linters.ts.
04The AHA Business Model Canvas — what it shows

71% Observation, 28% projection — the inverse of the Fiserv pattern.

The Fiserv BMC overlay from April was the worked example: 9 blocks, per-entity rows, temporal classification (Observation, Plan, Recipe, Missing), block health summary, chain-break narratives, strategic insights. Applied to AHA today.

AHA executes its legacy mission at scale. The strategic vulnerability lives in the 28% projection portion of the BMC, and that 28% is the entire competitive space against Apple Health, Google Health, Whoop, and the Commonwealth Fund. From the AHA BMC v0.1 strategic insights

The numbers tell the story:

PatternFiserv (2026-04-24)AHA (2026-05-12)
Observation23%71%
Plan62%9%
Recipe13%17%
Missing1%2%
StatePre-launch initiativeEstablished nonprofit since 1924
Strategic riskRuns out of time if it cannot startRuns out of relevance if it cannot evolve

Five Chain Breaks named in the canvas

Each chain break is a place where a Value Proposition or Resource depends on infrastructure that does not yet exist. The BMC marks the breaks so they can be staffed.

Chain Break 1 — $5.7B Research → Public Behavior Change → Population Outcome

AHA funded behavior change science but does not use it.

BCTO and the Health Belief Model are the methodologies that turn knowledge into behavior. AHA funded the research that produced them. The Research Funding cluster and the Behavior Change cluster sit on opposite sides of the aha-business-intelligence knowledge graph with no bridge between them. This is the structural finding behind the Mission Alignment score of 3.2/5.0.

Chain Break 2 — Authority → Loyalty → Lifetime Engagement

AHA Guidelines are the strongest moat in the sector. They do not produce loyalty.

Apple Health, Google Health, and Whoop are capturing the daily health relationship through wearables. The Tech Gap cluster in the aha-brand-score ontology is the most isolated cluster in the entire brand. Authority sits at one pole; loyalty at the other; the bridge that converts one to the other is the missing consumer technology product.

Chain Break 3 — Kids Heart Challenge → Teen → Young Adult → Go Red

AHA loses every generation at age 13 and does not reconnect until the mid-40s.

The Kids Heart Challenge runs through K-12. Go Red for Women lands in mid-40s. The teen and young-adult years are silent. During exactly those years, consumer-tech companies are building continuous health relationships with the same demographic.

Chain Break 4 — Women's Mission → Employer-Cost Vocabulary → Policy Impact

The cardiovascular underwriter of employer-workforce health economics is an unstaked institutional category.

The Commonwealth Fund holds the labor-cost-to-health vocabulary. AHA holds the cardiovascular data. Nobody fuses them. The May 2 intelligence package surfaced this as the transcendent strategic move available. It currently sits at Recipe across multiple BMC blocks (Value Proposition, Partnership, Resource, Activity, Customer Segment).

Chain Break 5 — Local Chapter Network → National Trust → Scaled Local Delivery

The chapter network is the most underleveraged asset on the BMC.

Health trust is local. AHA owns a national chapter network. The combination is a position no commercial competitor can replicate. Today the chapters fundraise; tomorrow they could be the human layer in an AI-mediated cardiovascular guru product. National platform plus local delivery is the position consumer-tech competitors cannot occupy.

Six strategic activities have "Unassigned" ownership

Inside the Key Activities block, six entries appear with the same owner: Unassigned. Behavior change integration. Youth pipeline build. Go Red refresh as prevention ecosystem. Male-ally activation. AI guru build. Cross-sector vocabulary research with Commonwealth Fund-tier peers.

These are the activities most critical to the differentiating value propositions, and they have no internal home at AHA. The Brand Power Score recommendations and the v06 intelligence package surfaced them as priorities; the BMC reveals they have not been staffed.

05What this means for our work

The studio is now multi-graph. AHA has a structural blueprint. The BMC method is portable.

For the ShurIQ product

The composite knowledge-graph viewer is the first surface inside the studio that shows multiple knowledge graphs for a single client in one frame, with mode switching that mirrors the InfraNodus app's analyst patterns. Concepts mode shows the network structure. Gaps mode overlays the structural bridges. Clusters mode reorganizes the layout around topical regions. Each layer in kg_layers represents a distinct analytical lens; the viewer makes them composable inside the studio rather than scattered across InfraNodus URLs.

D6 of the data-model plan ships tomorrow: the weekly stack-rank-to-BMC overlay delta. That work depends on the composite viewer being demonstrable end-to-end, which is what AHA's four seeded layers give us today.

For the AHA engagement

The BMC overlay is now the structural blueprint for AHA's next decade of strategic decisions. It names where AHA executes (the 71% Observation tier), where AHA has stated intent (the 9% Plan), and where the differentiation moves sit unowned (the 17% Recipe and 2% Missing). Every chain break is testable. Every "Unassigned" activity is a staffing decision.

For Limore, Diana, and Nuri: this BMC is the artifact to reference when AHA conversations turn to "what should we propose." The structural moves the v06 pressure test surfaced as recommendations now live in a single canvas, with their dependency structure visible.

For Shur Creative Partners as a whole

The BMC method is now portable. Fiserv worked. AHA worked. The same template applies to MicroCo's Brand Intel Cards expansion, to Hasbro's engagement, and to any future client engagement where we need to map the operating mission, the stated strategy, and the unstaked moves into one frame.

The Thursday team deck will fold this together: where ShurIQ stands as a product, how the methodology ships every week, how the production engine logs every claim, what 3-4 retainers at $3k/mo close the gap to operating self-sufficiency.

06What's next on the queue

From the locked decision set plus the data-model plan, ranked by what unlocks the most downstream work.

DayItemStatus
2026-05-12 PMW19 MicroCo issue ships under Grammar v0.2queued
2026-05-13 (D6)Weekly stack-rank → BMC overlay delta wired. First ShurIQ-self stack rank deployed.queued
2026-05-14 AM (Thu)Team state-of-the-union deck. 3-month plan to self-sufficiency. Section 4 names AHA + Hasbro + Fiserv + MicroCo as the candidate retainer set.queued
2026-05-14 (D7)Nightly InfraNodus delta scan via scheduler. "What changed last night" card on the account dashboard.queued
This weekR-LINT.1 inversion regex tune for sentence-terminal form. R-LINT.3 anti-restatement encoded.queued
This weekaha-zinger-pattern + multi-voice-synthesis skills codified under ~/.dotfiles/ai/skills/queued
This week (end)AHA stack-rank surface deployed as aha-stack-rank-v06.pages.dev against locked v0.3 grammar.queued
2026-05-17 (D10)ShurIQ Business Intelligence Report deploys at shuriq-self-bi.pages.dev. Self-applied competitive analysis + fair-dimensions stack rank.queued
2026-05-18 (D11)Investor editorial: "What ontology engineering needs to become, and what we're already shipping."queued
2026-05-19 to 21 (D12-14)Telegram Totem revival round 2: Memory (Letta three-tier), Knowledge (unified ontology slash command), Gap Minder (multi-platform scheduled scan).queued

Inputs Jonny still needs to land

The data-model REDIRECT dropped D2 because four inputs never arrived. They remain blocking on the parallel cold-read substreams.