influences.sgit.ai / admin / comms
Comms: tasks, requests and the briefing queue
The working channel between the project lead and the agent building this site, kept in public and updated every release. On the sibling sites this page carries build questions. Here it does one more job: Dinis Cruz's stated plan — in time I will create a detailed briefing document about each one — becomes a visible, addressable inbox, with one open thread per influence awaiting a briefing. When a document lands, the thread closes and the entry's tier moves.
Tasks
| # | Task | State |
|---|---|---|
| T1 | CI pipeline: validate → auto-tag → deploy, matching the sibling sites | done |
| T2 | Auto-tagging on every push to dev, verified against version.txt and the commit subject, with historical backfill | done |
| T3 | The release gate: version agreement, internal links and fragments, canonical host, credential-shape leak tripwire | done |
| T4 | The no-verbatim gate — this site's founding editorial rule as a check rather than a paragraph: every blockquote declares whose words it carries, and a third party's are capped at forty | done |
| T5 | The register-is-the-data check — every tier count on every page recomputed from data/influences.json, with the orphan checks in both directions | done |
| T6 | Shared chrome — nav, footer, version badge and tallies owned by one script and rewritten across the tree | done |
| T7 | Generators for register/, map/, documents/, llms-full.txt and sitemap.xml, each with a CI --check drift gate | done |
| T8 | !admin/build/ in .gitignore — the trap that shipped a sibling site without its own release gate | done |
| T9 | Publish the commissioning pack verbatim under /briefs/, with a generated reader page each | done 9 documents, 5,725 words |
| T10 | data/influences.json — the register, seeded from the pack's influences__seed.json and extended with resonance text, trace rows and research plans | done 25 entries |
| T11 | The twenty-five entries, in the build order the brief sets: Victor first, then the link-out entries with sibling sites, then the Semantic Web, flow, Luhmann, the stated seven, and the map | done 13 full · 3 link-out · 9 stubs |
| T12 | /tiers/, /format/, /library/ and /shipped/ — the tier definitions, the register format documented, the union of every entry's wider library, and the tier-movement changelog | done |
| T16 | A front page about the ideas rather than the machinery. Eight principles and six quotations, generated from the register by gen_front.py so the home page cannot quote a principle an entry has since edited | done |
| T17 | /provenance/ — where the register came from, how the site is generated, the two editorial gates, and the four things it cannot do. Everything the front page used to carry about its own construction | done |
| T18 | /articles/ — an article is a markdown file and the page is generated from it, so a piece can be republished without a second copy existing. Needed a small stdlib markdown renderer (mdblock.py) whose blockquote handling enforces the attribution rule at authoring time rather than at CI time | done |
| T19 | Name the subject. The site said the founder in 227 places. Dinis Cruz is now named in the titles, the hero, every footer, a byline under every entry, and the register's own prose; /about/ says who he is and links the two canonical accounts rather than keeping a third | done |
| T13 | Infographics: original artwork for likenesses, redrawn idea-figures in house style with attribution, each with a markdown or JSON twin as its source data | queued |
| T14 | Cross-links from the sibling sites — wardley-maps, graphs and open-source each own the material for one of this site's link-out entries and none of them points back yet | queued |
| T15 | Vendor the diagram renderer. The map pulls mermaid@11 from a third-party CDN at a pinned URL and degrades to a readable text fence when it is unreachable. The fallback works; the dependency is still a dependency | queued |
The briefing queue — one thread per influence awaiting a document
Seven influences are on Dinis Cruz's own list and have no corpus evidence at all. That is not a defect in the mining: they are the things nobody else can research. Each gets a thread here, and the entry stays a published stub with its research plan visible until the thread closes.
Suggested order, and why. Start with N1 or N6 — Neil Peart and the band years are the entries no researcher can substitute for, and the two most likely to change how the rest of the register reads. The others have public material a research pass can work from once Dinis Cruz confirms the anchor.
| # | Influence, and what the briefing needs to settle | State |
|---|---|---|
| N1 | Neil Peart & Rush. The anchor: which record, which lyric, which period. The hypothesis to test rather than assert — the craftsman-lyricist archetype, technical mastery in service of meaning. Note the boundary: this entry describes and links. Lyrics are the one place where even a verse is a legal risk, and the entry quotes at most a fragment under analysis. | for Dinis Cruz |
| N2 | Kevin Kelly's books. Which is the anchor — What Technology Wants, The Inevitable, or the 1,000 True Fans essay? And is the technium-to-Wardley-evolution connection real or a plausible-looking coincidence? | for Dinis Cruz |
| N3 | Simon Sinek — the golden circle. Anchor: the book or the talk? The site network is organised why-first, which is either evidence or hindsight, and only Dinis Cruz can say which. | for Dinis Cruz |
| N4 | The Cathedral and the Bazaar. The essay's factual history belongs to open-source.sgit.ai. This entry owns only what it did to Dinis Cruz, and that is exactly the part not written down anywhere. | for Dinis Cruz |
| N5 | David Rice — "Upon the Threshold of Opportunity". Which year's OWASP keynote is canonical, and is there a recording whose licence permits an embed? OWASP's own channels are the preferred source; a commercial conference recording gets a link, not an embed. | for Dinis Cruz |
| N6 | Music, and playing in a band. Entirely unresearchable by anyone else. The hypothesis worth testing: band-playing as the original experience of real-time collaborative flow — the thing the estate's development methodology tries to recreate in software. | for Dinis Cruz |
| N7 | Diverse and distributed teams. Research plan is written: Dinis Cruz's management writing, the estate's team-roles system, and OWASP's global-chapter model as the formative experience. The briefing needs to say which of those is the origin and which is the consequence. | for Dinis Cruz |
Questions — these need an answer from a person
| # | Question | State |
|---|---|---|
| Q1 | The full text of the Bret Victor register. The commissioning pack ships a faithful summary of its structure and closing argument; the full text lives with Dinis Cruz, supplied in conversation and not retained across a context boundary. Entry one cannot ship complete without it — the site publishes what the pack carries, clearly marked as the summary, and swaps in the original the moment it arrives. This is the first ask. | for Dinis Cruz |
| Q2 | Confirm or demote the discovered influences. Luhmann, Christopher Alexander, Team Topologies/Cynefin and Vannevar Bush were found by mining the corpus and never named by Dinis Cruz. Do they belong, or are any of them a citation rather than an influence? The falsifiability of this site starts with Dinis Cruz falsifying its discoveries. Design and Steve Jobs: already confirmed, 25 August 2026 — the first discovered → confirmed transition, before the site existed. | for Dinis Cruz |
| Q2b | The Design entry's anchor. Which artefact is canonical — the "design is how it works" formulation, the MP3-to-CD whiteboard story, a keynote, Ive's process, a specific product? And when did Design-with-a-capital-D become explicit in the thinking rather than implicit in the work? | for Dinis Cruz |
| Q3 | Anchors to confirm across the stated seven — Kelly, Sinek, Rice. Rolled into the briefing threads above rather than asked twice. | merged into N2/N3/N5 |
| Q4 | Are there influences you rejected? Things everyone assumes shaped the work but did not, or that were outgrown. An entry with status: counter-influence would be unusual on the web and very much in this site's spirit — a register that can only add is not falsifiable in the direction that matters. | for Dinis Cruz |
| Q5 | Where did graphs actually enter? The founding obsession has an entire sibling site and no recorded origin. O2 Platform's code-flow work, or earlier? The link-out entry owns the personal history and nothing else, so this is the only thing it needs. | for Dinis Cruz |
| Q6 | Does the map belong here or on the graphs sibling once it becomes a real G³ demonstration? Proposal: authored here, mirrored there — the influence graph is this site's content, and its rendering is that site's subject. | for the lead |
| Q7 | Infographics and likenesses. Dinis Cruz asked for infographics of people and ideas. The rule this site adopts: likenesses are original artwork only, never scraped photographs and never AI renderings presented as portraits; idea-figures are redrawn in house style with the idea's author attributed. That needs either an artist pass or a generation workflow Dinis Cruz approves. Tracked as T13. | for Dinis Cruz |
Requests — build-side, for the lead
| # | Request | State |
|---|---|---|
| R1 | The corpus is cited, not resolved. Every entry lists the files that put it in its tier, by path. This repository holds the website, not the corpus, so those paths are quoted from the mining run behind the commissioning pack (v0.33.62, 25 August 2026) rather than checked against the repositories at build time. Nothing on the site can disagree with the pack; the site can and will drift from the estate. Closing it needs a decision: give this build read access to the corpus and re-run the scan as a build step, or accept a dated snapshot and re-scan per release. | for the lead |
| R2 | Trace rows without versions. The register format's own rule is that a trace table without version numbers is an opinion. The mining run reached files and named them; it did not always record the release they were read at. Rows are marked where a version is missing and the entry says how many — but the real fix is the same access R1 asks for. | for the lead |
| R3 | Two decades of earlier material are unmined. The corpus scanned for this pack is the current estate plus the current docs site. The O2 Platform-era blog, OWASP talks, conference decks and the earlier books were not reachable — and they are exactly where the stated influences' traces will be found. The per-entry research plans say where to look; someone has to be able to look. | for the lead |
| R4 | GitHub Pages environment. The workflow calls configure-pages with enablement: true and deploys from dev, which is this repository's default branch. On the sibling sites the first deploy job failed until the github-pages environment recognised the branch. This release is the test. | testing |
Notes from the build
- The pipeline was built first, deliberately. Every sibling site's early releases are the same story: content lands, then something in the release machinery turns out to be wrong, and the fix is a release of its own. Publishing an almost-empty site through the full gate proves the gate on a tree small enough to read in one sitting.
- Two of the six checks are editorial, and that is the point. The no-verbatim gate and the register-is-the-data check encode rules this site would otherwise merely state in a footer. A site whose thesis is that a claim should be checkable is in a poor position to publish unchecked claims about itself.
- One trap was inherited and pre-empted. The Python
.gitignorethis repository starts from carriesbuild/, which silently swallowsadmin/build/. A sibling site shipped a first release whose validate job died on a missing file before it could check anything — the checks appeared to run and did not. The!admin/build/negation is in this repository's first commit. - The gate earned its keep in this release. Two classes of bug were caught before anyone
saw them, both of them the same shape — prose authored once and rendered in two places at
different depths. Trace-row text written for
/register/<slug>/resolved one level too high on/register/<slug>/trace/, and Block-7 notes broke again on the shared library page. The first is now a transform in the generator rather than something an author has to remember; the second is a rule — Block 7 carries external addresses only, and cross-links live in the resonance. - Six trace tables, seven pending, and that ratio is the honest state of the register. Where the commissioning pack recorded a path and a version, the rows exist. Where it recorded a file count, the entry says so and the table stays empty rather than being filled with plausible rows. R1 is what closes the gap, and it is a decision rather than a task.
- Entry one is incomplete and shipped anyway. The Victor entry is the model for the whole format and its trace table is missing, because the register that contains it is with Dinis Cruz (Q1). Publishing it with the table empty and the reason stated is more useful than holding the site back, and considerably more useful than inventing the rows.
- Nothing here claims to be finished. The nav lists the pages that exist, the version badge says what stage this is, and the release history says what shipped in one row per release. All three are easier to grow than to walk back.