History

Add the Corrections page: seven errors this repository published and fixed The profile claims every published project carries a corrections ledger. This one did not, which made the claim false for the repository it points at first. Seven entries, all found by checking rather than by a reader: a spine that named nine frameworks and defined three, a front page advertising five tools where two other pages said seven, a wrong layer number, an Agent Card missing its three highest-governance patterns, a YOUR-USERNAME placeholder shipped in a public clone command, a framework name that outlived the framework, and an earlier definition of the third proof as Credibility rather than Competence. Six of the seven are the same failure: a fact written in more than one place, one copy changed, nothing checking the others.

canonaihq-dot committed Aug 24, 2026

Make the pages client-facing: caveats and confidence move to one page Every page carried a confidence line at the top and a "what is not verified" section at the bottom. On a wiki someone is reading to decide something, that reads as hedging, and a caveat repeated on twenty-five pages stops being read on the second one. - The confidence lines and per-page caveats are consolidated into one new page, What Is Not Verified, which keeps all of it: what each page rests on, the confidence scale, and the two rules that hold everywhere (no named-company figure appears anywhere, and thresholds are deliberately unstated). - Home, the sidebar and Published-Tools each promised a per-page confidence score. That promise is replaced with a link to the one page that delivers it. - The Evidence Standard described a per-page section that no longer exists; it now points at the consolidated page instead. - Drafting-process asides removed from the framework spine, the glossary, the three drifts and Published-Tools. They were written for a reviewer, not a reader. Nothing was deleted. The pages argue; one page says what the arguments rest on.

canonaihq-dot committed Aug 24, 2026

Build out the wiki: 21 new pages, four concept drawings, five pages revised The wiki was five pages and named nine frameworks while defining three. It now defines every framework it names. New (21): the agent factory, the talent stack, agent patterns, the seven-layer stack, the seniority ladder, the three drifts, governance as code, the agent registry, adaptation as change control, the three proof gates, the deployment ladder, finding the work, the first 90 days, migrating an RPA estate, value and the CFO, NORTH and the trust ladder, people power and politics, choosing what to build, the glossary, the evidence standard, the reading list. Revised (5): - Home said five published tools where the README and Published-Tools said seven. Now seven, plus a map of the whole wiki by group. - Framework-Spine named five frameworks it never defined. Every name now resolves, and risk tiers are defined inline from the two shipped tools. - Published-Tools now says what "seven" counts, since the repo holds five files in tools/ and seven in cheatsheets/. - Why-Programs-Stall gains one outbound link per numbered failure, nothing else. - _Sidebar replaces a three-entry nav that would have left new pages unreachable. Four concept drawings in the house style, replacing the mermaid flowchart on Home: the spine, seniority overshoot, the three drifts, and the registry gate. Every page carries a last-reviewed date and a confidence line, which Home and the sidebar have promised for some time and no page delivered.

canonaihq-dot committed Aug 24, 2026

Sweep: rules to a single masthead Separation by whitespace and heading weight, matching the rest of the estate.

MarioLazo committed Aug 23, 2026

Rebreathe the wiki: shorter blocks, more air, one idea at a time Kleon's actual discipline is restraint rather than whimsy. Minimal palette, generous whitespace, short chunks, italics over bold, one idea per block, and horizontal rules used as rhythm rather than as dividers. That translates well to dense enterprise material; the doodles would not, and this audience is CIOs. Home now averages under seven words per block, down from paragraphs. Same content, a fraction of the cognitive load per unit. Two pages the sidebar promised and did not have. Why Programs Stall walks the five failures in the order they actually happen, each one an operating model failure and each predictable months ahead, closing on what a CoE is actually for: the function that owns the questions nobody else is paid to ask. Published Tools gives each of the seven its own block with the detail that matters, plus the publishing rule stated as a rule. All internal links resolve.

MarioLazo committed Aug 23, 2026

Build the wiki: the operating-model problem, and the framework spine Home opens on the misconception that enterprise AI programs fail on technology. They fail because nothing around it is built to fund, govern or scale it, which is an operating model problem and the one nobody assigns an owner. Framework Spine records two corrections the estate made in August and which this repository had not yet absorbed. Three Proofs is Technology, Value and Competence, not Technical, Business and Governance. And BXT is explicitly not the same framework: BXT names the dimension, Three Proofs names the evidence you bring to it. Trust is the outcome; Competence is what earns it. They were once described as equivalent and are companions. Application Modes and Five Modes were both called Four Modes until 2026-08-20, and the collision produced a long-running fork. Application Modes classifies what the AI does and is what an Agent Card records; Five Modes classifies what the human does while it runs. Design follows the house style used across the estate: name as H1, one bold sentence, section headings that make a claim, a mermaid spine with the Operate-to-Diagnose loop drawn because it is the one most programs never build, and a sidebar grouped by reader intent.

MarioLazo committed Aug 23, 2026

Initial Home page

MarioLazo committed Aug 23, 2026