FB-26
Live secondary pages "thin, text only, and somewhat vague"
HERS by her own motion (10:57a offer); ADIP + agents drafts are the pattern; build→zip→we stage; team keeps links/staging/production.
Updated with edition #015 · Monday, August 31, 2026
This page is the full record behind the short update emails. The email tells you only what needs you; this page holds everything — every request, who has it, and where it stands. It refreshes with each update. Reply to any update email and your item lands here as a numbered row. The page is unlisted and marked not-for-search-engines.
Straight from the record, one block per item. The star marks the items that unblock the most other work.
Live secondary pages "thin, text only, and somewhat vague"
HERS by her own motion (10:57a offer); ADIP + agents drafts are the pattern; build→zip→we stage; team keeps links/staging/production.
Two held figures — one published set settled, one pair needs a home
The large Year-1 figure cleared earlier by phone now publishes in the wording of Sunday's slides. The remaining inventory and cash-conversion pair appears in neither slide, so it comes off at the swap unless it gets a home — asked above.
3-5 links per section (homepage + /adip sections), "for maximizing search results"
Link passes where REAL targets exist (~14 public pages — no filler links; quality over count, explanation goes in #007).
The full site is LIVE — published Monday afternoon
The Sunday build went live across the whole site Monday afternoon: the new homepage, the Company page, and thirteen new pages (Platform, The ontology, Worked decisions, Integrations, Agents, Why agentic, Industries with four verticals, Results, Trust). One customer is now named with the approved figures; the other stays anonymous; the agent roster is the current library. The page-by-page review copy stays at the preview address, and the site map now marks every row live or hidden. The re-test checklist runs against the live site as the follow-up pass.
The official customer numbers — answered Sunday morning
Mike answered all six decisions on Sunday: one customer publishes under its name, the other stays anonymous, every result figure now comes from the two slides he attached, benefits state at the customer level rather than per pillar, and the agent roster comes from his Agentic Wheel deck under the name "our most frequently used agents." The port applies all of it. One legacy figure pair appears in neither slide and comes off at the swap unless it gets a home — asked above.
"Should we create a secondary page specifically for the, what I call, mini demos? Or should we put it on the page with the Results? … The clickable content is something you don't see everywhere and it should grab attention. They can also be used in calls with prospects or even as sales collateral. Let me know thoughts on this before I drag everyone else into it." (links /learn/decisions/clear-to-build)
Stephen answered directly 11:55a ("Maybe look up a level for ideas? https://a2go.ai/learn/decisions") — the index already holds BOTH interactive decisions + 5 teasers. Logged as her open placement decision: dedicated index page vs fold-into-Results. Context: the mini-demo family is NINE strong (2 interactive decisions at /learn/decisions + 7 SCOR caselets at /scor/<caselet>) — argues for an index of its own linked from /supply-chain-ai-agents, with Results carrying real-engagement numbers and the index carrying demonstrations (matches the full-site audit's use-case-index finding). Her call as copy owner; Artemis builds either way. Sales-collateral use noted. MOVEMENT 8/28 8:44a (same thread): "I saw that before. I do want to use this but we need to figure out where and then use it to push people to fill out the survey and schedule calls" — intent CONFIRMED, purpose stated (drive survey fills + booked calls; dovetails with FB-46/FB-60 wiring); placement still her open call.
Eight cover images for the Library blog posts, one per title (Decision intelligence / Seven S&OP challenges / Ontology of Execution / Agentic AI Shift / AI Orchestration / Operations Break Down / Governance / Optimizing Supply Chains guide), labeled "first/second/… image above"
Stephen 11:50a on-thread: attachments don't preserve sequence — indicate by FILE NAME; also image filenames are public on the web and need compliance-grade names + written descriptions for accessibility ("I bet Claude will do descriptions" = Artemis writes the alt text + renames at staging). Received filenames incl. shutterstock_2614009279 copy.jpg, supplychainai2/3.jpg, governance image copy.jpg, bRcETgD8KbDrAMmOEKXtp.png, Abstract_artistic_visualization…png — the title→filename list from her unblocks; staging is same-day after that. NOTE: Outlook connector can't fetch binary attachments — pull via OWA at staging time (a2go-web-roles memory).
"There are currently 3 different header styles depending on the page, it is not uniform, I will show all 3" (three screenshots in his report)
Matches the known chrome fragmentation (static arms + /about grafted mast vs Astro chrome; the full-site audit's footer/chrome finding is the same defect family). One-chrome-everywhere is standing law (FB-29/36 lineage). Unification pass queued as ONE build with the FB-69 mobile-menu fix — both uniformity findings, one deploy, one retest ask for Jason.
"Can you use pattern below to sweep entire site as an 'adversarial human review' role like in the Claude training? Please log to Artemis attn." The pattern = his 4:53p team note on the preview line "Disconnect ADIP and your business keeps running exactly as it did" ("raises my eyebrows ;-)") — copy that passes grammar QA but fails a hostile read: as written it says the product's absence changes nothing; the intended claim is no-lock-in. 6:16p amendment (his note says update this entry, not a new one): "and trap British spelling equivalents like 'colour' please."
STANDING SWEEP LENS established: sounds-right-means-wrong claims (self-defeating statements, accidental negations, unintended literal readings, in-sentence contradictions) + UK-spelling traps + the standing law checks (claim strength, naming, SCOR DS terms), swept in an adversarial-human-reviewer role. First full pass over all 64 live pages + the staged full-site preview run 8/28 (results itemized in this row after the run). Triage rule: live-page findings → build/fix items; preview findings → the FB-70 port fix list (the seeding line itself is a port-time reword). Companion human motion: Stephen's 4:53p asked Jason to start checking copy too — same lens, human side. LENS SHARPENED (Stephen 8/28): every sentence must survive being QUOTED ALONE — a reader meeting it in a search-result summary or an LLM answer with zero surrounding context (the site is llms.txt-indexed; contextless quotation is now the normal read, not the edge case). And rewords must aim at each claim's TRUE TARGET: the seed line was trying to defuse IT-policy defensiveness about ERP write-back (as-is environments are treated as fragile; a day ERP simply ran counts as a win) — the honest claim is bounded, approved, auditable, REVERSIBLE write-back through official APIs, plus no rip-out, plus everything already written back stays in the customer's systems if ADIP is ever switched off. Never "our absence changes nothing." FIRST PASS DONE 8/28 (workflow: 8 sweepers over all 64 live pages + the 28-page preview file, 2 independent refuters, consolidation — 84 raw → 51 verified → 44 deduped entries: 12 fix-now / 29 fix-at-next-touch / 3 notes; 19 meaning / 14 howlers / 7 UK spellings / 4 law; 33 live / 11 preview). Full triage memo with Stephen (session copy-sweep-2026-08-28.md). Notable: the console's Plan/Source/MAKE/DELIVER taxonomy is a SCOR DS law breach (folds into FB-51's alignment); an MDX tilde renders half a caselet paragraph as strikethrough; one caselet sentence states the inverse of its own recommendation; the governance paper's benefit list sits under a copy-pasted "immaturity" heading. ROUTING (Stephen 8/28 ~2p, supersedes the lane split): ALL live-site items — house content, archive posts, homepage arms alike — go into ONE approval email (#012), each item as link-paragraph + description-paragraph (no tables in email), numbered 1–32 so Betsy can approve wholesale or strike by number; ON HER APPROVAL Artemis ships THE HYGIENE BUILD with standard logging (Stephen's name for it; pirate voice declined). Item 3 (the IBM Distinguished Engineer credential) is a confirm-the-fact item, and item 4 (one survey name + duration) is her pick — both flagged as such in the email. Preview items (11, incl. the seed line's replacement — "disconnection as proof of retention": everything written back stays, only the next recommendation stops) = the FB-70 port fix list, nothing changes without her. Standing lens continues; future passes ride the toolsite health-check when built. FULL APPROVAL 8/28 2:13p (her reply pasted the whole list back with per-item verdicts): all 31 approved — "Approved"/"OK" ×29, item 3 = "Use \"Pain-point survey, 10 minutes\" throughout" (her pick, applied everywhere incl. the SiteFooter chrome and the survey page's own duration), item 10 = "A2go everywhere" (text corrected; uppercase look kept via CSS). HYGIENE BUILD SHIPPED same day via scripts/hygiene-build-2026-08-28.py (asserted, idempotent transforms; ~85 edits across 29 files; two same-class bonuses logged: the notebook-concept "Analogue" and the menu-preview survey label; the two import-defect editorialNotes updated to record the fix). Row stays open as the STANDING sweep. SECOND PASS 8/31 (the v20 delta): the four NEW pages in full + the copy delta on 11 changed shared pages + two full-file passes (5 lenses → consolidate → 2 refuters; 74 sustained / 7 refuted, then 4 more corrected by orchestrator verification — the sweep brief had mislabeled the Databricks-story published-class figures (28h→<1h, 65/17/150k) as uncleared, "32-person" is on the approved live /about, and the BOM "off-by-one" is really v20's ontology-page COMPRESSION dropping the live example's second self-clearing short). HEADLINES: the held ~30%/25% pair is PUBLISHED in v20 (staging re-ambered it; strips at swap unless homed — FB-28); the ontology page's "an agent has no such instinct… a wrong answer looks exactly as well-formed" fails quoted-alone unqualified; the synthetic worked-example figures ($438k etc.) spread to ontology/index/library WITHOUT the live pages' synthetic disclaimer; a new Brazil/SAP delivery claim on the agribusiness page needs business confirm; Matt's bio appears in full despite the own-subject gate (FB-71 holds it at port); "8–12 weeks" got unbounded to "live in weeks"; one classic SCOR code ("P2") on Library; the site-map card's 13 library links are mechanically broken (bare hrefs); About lost its booking CTA in the trim (flag — FB-60/46 expects it); plus a ~12-item UK-spelling batch and wording-alignment set for the port. Full triage: ~/Developer/fb70-inputs-2026-08-31/v20-sweep-2026-08-31.md (+ raw JSON). Triage rule stands: v20 findings ride the FB-70 port fix list; nothing changes without Betsy — the 1:00 session's shared screen is the natural venue.
"'No Decision trace panel' — LIKELY FAIL, high priority. The full Decision trace panel is present in the page content… may be hidden via CSS rather than actually removed… file this as a row and flag for visual confirmation: shipping hidden-but-present content is a common source of accidental leaks (e.g., if CSS fails to load)."
CONFIRMED, and by design at the staging layer: stage-full-site.py hides the panel via CSS per the 8/26 "no product surface yet" ruling — the as-received file stays faithful, the concept stays on file. VISUAL CONFIRMATION done 8/28: real-browser screenshot of the preview shows the [PANEL HELD] note and NO rendered panel (the hide works). His leak-risk point is right and stands: at PORT the panel is REMOVED FROM THE MARKUP entirely, not merely hidden — logged on the FB-70 port pass. Interim exposure accepted for a noindex staged page.
Apollo.io notifications → a SHARED mailbox or distribution list (never an individual's inbox — same shared-surface philosophy as the calendar rule)
Stephen configures in Apollo; pairs with FB-72 (Pat is now set up in Apollo).
Configure and/or TEST that the Apollo.io scripts are working on the website (trigger: Pat set up in Apollo.io)
Standing law from the 8/14 Apollo brief: the visitor snippet rides ZARAZ consent-gated ("Visitor identification" purpose), never pasted into templates; verify with a REAL browser only (Zaraz is curl-blind); the RB2B / ZoomInfo / Apollo three-tracker stacking decision is still open and this work surfaces it.
Betsy, on #008: "I will be adding Matt's bio to the Company page and I'd like this part of that page removed as I think it is redundant and a bit confusing" — the FB-70 build's Leadership intro block ("LEADERSHIP / The people running the work. / A2go was founded by Mike Romeri and Stephen Hutson. The leaders below run the disciplines the platform depends on every day.")
Removal executes at the FB-70 port (or in her next build version — whichever lands first). THE BIO (Stephen ~8p): publishes when she provides it; until then it is logged as DUE FROM the copy owner (the placeholder card holds meanwhile). Side effect: removing the block removes the "A2go was founded by…" public founder-set line, which was itself a confirm-first fact. SIGN-OFF OBTAINED 8/31 midday (Stephen: "Betsy got sign off on Matt's bio"): the own-subject gate clears — the full bio that arrived in v20 publishes at port as-received (the 8/31 sweep's bio hold LIFTS; its employers/figures/name-drops are now cleared-class as received from the subject via the sign-off). The Leadership intro-block removal still executes at port. Row closes when the port lands both.
Jason QA (run 3, same reply as the FB-34 confirm): mobile menu options "not being uniform on pages" — Home and Company render WITHOUT the menu button, "pretty much every other page" WITH it; "I think it looks cleaner with the menu button personally"
Likely root cause: the homepage arms + /about are static pass-throughs with GRAFTED chrome (add_mast_chrome) — the graft lacks the Astro chrome's mobile collapse button. Fix = uniform mobile menu on the statics (his preference noted: button everywhere). NOTE: this reply arrived 11:34a and was missed by the 8/26 session AND the 3:36p sweep (its window started 1p) — caught 4:15p inside Betsy's quote-chain. Sweep lesson: window from the last PROCESSED mail, not the last session. RUN-4 UPDATE (8/27 3:04p): the home page now renders the menu button properly; the Company page is the one page left with the manual list — fix rides one build with FB-78.
/pain-point-survey collects ten steps and returns nothing back — "lead capture, not a diagnostic"
DIRECTION SET by Stephen 8/26 (~9p, close-unless-corrected in the edition): the survey ENDS WITH A MESSAGE + A LINK TO BOOK A MEETING on the SHARED calendar — "customers only compete for calendar time with other customers"; never an individual's calendar ("a recipe for frequent changes"). Supersedes the 8/25 route-to-a-team-member proposal (recorded faithfully as hers) and the instant-scoring fallback. Betsy writes the closing message copy; the wiring rides the same shared-calendar build as FB-60. DIRECTION STOOD — no correction through #010 (8/27): the close-unless-corrected window passed quietly; the closing-message copy is the open deliverable. HER SPEC 8/28 2:17p (on #012): "When completed, we need to ask for the person's work email and also suggest signing up for a 30 minutes call with the A2go team" — work-email capture ADDED to the build spec; the 30-minute-call suggestion IS the decided direction. Her "Use Matt Calamusa's calendly link" rides FB-60 (see there).
Taxonomy alignment: FB-45's four-pillar canon vs the six signal families (/signals) and the five-domain 35+ agent console (/a2go-signal-examples/) — Betsy: "I don't know where it is describes as six signal families or five process domains so I'm confused"
Align both surfaces to "supply chain specific agents" within FOUR PILLARS (no 35+ count in copy); next edition links her to both pages so she can see what changes. SKU/packaging context rides Stephen's Frank/Cesar/Mike thread.
Results page + homepage proof band (audit fix 1, CRITICAL; "one thing to decide this week")
DRAFT at /homepage-drafts/results.html. Live now: Databricks-published numbers + the 18h→15m anonymized stat. Two amber slots await NAMED-customer clearances — Betsy obtains + logs the sign-offs and delivers the figures with them; the page ships the day the slots fill (or ships with slots removed, her call).
Traditional → Generative → Agentic reframe page (audit fix 6 — "highest-value SEO page")
DRAFT at /homepage-drafts/traditional-generative-agentic.html; slot for her sequence-tracker framing so site echoes outbound word-for-word.
Four vertical pages (audit fix 7): food & protein, industrial manufacturing, distribution, life sciences
DRAFTS at /homepage-drafts/vertical-*.html; agent-role slots to confirm against the Agent Selection Guide (the Guide document is not in team hands — she supplies or points at it). HER 8/28 4:19p: "Don't know what Agent names on the Agents page means, or what the Agent Selection Guide means" — ANSWERED same day by email with the agent-names PDF (the preview Agents page's amber note; the ~22 named agents under the pillars; the vertical drafts' Guide slot): the Guide is the audit document's own name for wherever the agents-by-industry master list lives; if no such document exists, nothing is blocked — the names ride the same sign-off sitting as the figures (FB-28/74), and the vertical pages follow that list. Ask softened accordingly: send it if it exists; otherwise the sitting covers it. 8/30 RESOLUTION (Mike, via FB-74): the master list now EXISTS — the A2go Agentic Wheel deck (his "third attachment"). The vertical drafts' agent-role slots fill from the Wheel's four quadrants; the Guide question retires. Nomenclature on every surface: "most frequently used agents", never "standard agents".
Two persona pages (audit fix 11): CFO and COO/VP Supply Chain
DRAFTS at /homepage-drafts/persona-*.html; buyer-research-doc slots.
Trust & security page (audit fix 9)
DRAFT at /homepage-drafts/trust.html stating ONLY claims already live. CORRECTED 8/26 (Stephen): the certification/posture/subprocessor FACTS come from CESAR + ALEXANDRE (Alexandre owns security) — not Stephen; Betsy obtains + logs them per the clearance rule. NO certification claim (SOC 2 etc.) publishes unless confirmed true. Same sources gate the FB-70 build's trust-page architecture statements.
Single sitewide demo CTA — "Book a 30-minute working session" (audit fix 3)
RE-OWNED to Betsy 8/26 (~8p, Stephen): she takes her own 5:23p proposal (a connected calendar for 30-minute initial conversations, survey sent ahead when useful) to Mike. Stephen: any booking link will be a SHARED calendar anyway — mechanics settled in principle; the sitewide CTA standardization pass fires on the Betsy+Mike word. Survey stays the secondary path. Every draft + the full-site build carry wired slots. RATIFIED FORWARDABLE WORDING (Stephen 8/26 ~10:30p, drafted to travel well on the sales side — never mention admin visibility): "How the booking link works: the site's 'Book a working session' button points at one team booking calendar. That keeps the public link permanent — it never changes as the team grows or roles shift — and customers always see real open slots instead of any one person's busy week, so a change on somebody's calendar never ripples back to the website. Every booked session lands in the calendar of whoever takes it: individual calendars connect to the team calendar in Outlook, and booked sessions show up there like any other meeting. Who takes which session stays entirely a sales-side call — the link books the time; it doesn't decide who gets it, so any way the sales team wants to share sessions out works behind it, and can change without touching the site." HER 8/28 2:17p (on #012) re-proposes the individual calendar ("Use Matt Calamusa's calendly link") — the standing shared-calendar rule holds (one permanent link; customers compete only with customers; sessions land with whoever takes them), and the ratified forwardable wording above is the answer; still fires on the Betsy+Mike word. STEPHEN'S DIRECTION 8/28 evening: CALENDLY REJECTED as an approach — senior buyers "are not gonna open someone's calendly and then compare it to their calendar"; target = an EMPTY calendar where the buyer just picks any day + time. FUTURE VISION (he asks Betsy + Matt): a simple picker — pick day, pick time, free-form text, plus a small box demonstrating our AI. INTERIM (his plan): his Outlook via a separate empty-calendar account. CHECK DONE (his ask): Microsoft's Calendly equivalent — "Bookings with me" (the personal page he tried early, since matured: Outlook-integrated, custom questions, Teams links, public meeting types) and, the better fit, the SHARED Bookings page: one permanent team page, staff availability from Outlook, auto Teams meetings, custom + required intake questions (the free-form box exists today), bookers need NO Microsoft account, included in Business Basic/Standard/Premium + E1/E3/E5 — the a2go.ai tenant is almost certainly already licensed. The empty-calendar want maps exactly: attach the page to a dedicated bookings mailbox with an empty calendar → every business-hours slot always shows open. RECOMMENDED WIRING: the site button points at OUR route (a2go.ai/book → redirect to the Bookings page) so the published link NEVER changes — the future custom picker replaces the redirect target, not the link. BUILT LIVE 8/29 (Stephen watching, Artemis driving his Chrome): (1) tenant Bookings org toggle verified ON + booking-page search indexing DISABLED (link-only, saved); (2) root cause of "not enabled for you" found and fixed — the Microsoft Bookings SERVICE PLAN was individually unchecked inside his Business Standard license (the only unchecked app in its stretch); ticked, saved, provisioned in ~15 min; (3) SHARED Bookings page "A2go" created (auto-created scheduling mailbox A2go@analytics2go.com = the always-empty calendar); staff = Stephen (Administrator) with "Events on Office calendar affect availability" UNCHECKED — every business-hours slot always open, his real calendar never consulted; (4) service = "30-minute working session," 30 min, TEAMS MEETING ON, Mon–Fri 9:00–5:00, description "Pick any time that works for you — 30 minutes with the A2go team, online via Teams."; intake = customer email REQUIRED + Company (custom, required) + Customer notes (free-form); (5) access = ANYONE (public self-service, no Microsoft account needed) — cold-tested from a logged-out browser: loads, slots, form; (6) page time zone corrected UTC → EASTERN (the one real catch: the created page defaulted to UTC; buyers see their own local times); (7) a2go.ai/book SHIPPED (arethusa 2bc41d4): middleware 302 → https://outlook.office.com/book/A2go@analytics2go.com/ — the permanent public address; the future picker swaps the target, never the link; site-map row added, not yet linked from any page. HARDENED 8/29 afternoon (Stephen, on seeing the URL: "folks are paranoid now about anything that is not the official company domain due to all the phishing and scams"): (8) booking mailbox primary SMTP RENAMED A2go@analytics2go.com → A2go@a2go.ai (legacy address kept as invisible secondary) — the buyer-visible URL now reads bookings.cloud.microsoft/book/A2go@a2go.ai; legacy tenant domain gone from sight. HOW (for the record): the Exchange admin center hides scheduling mailboxes entirely (searching "A2go" surfaced only the Artemis user mailbox), so Exchange Online PowerShell on his Mac (module already present); device-code sign-in BLOCKED by tenant Conditional Access ("sign-in was successful but does not meet the criteria… authentication flow restricted by your admin" — the managed anti-phishing policy doing its job; left fully intact, no weakening suggested); interactive browser sign-in sailed through on his live session with zero prompts; guarded script acted only after finding exactly one mailbox at the target address — the tenant holds FOUR scheduling mailboxes (old experiments "Meet With Us!", "Analytics2Go", one of Betsy's) and only ours was touched, before/after printed. (9) STAFF SELECTOR REMOVED (Stephen: "we do not need them to choose Staff. They would not recognize the names anyways. If we keep names they also need to show title" — TITLES RULE noted for any future named-staff change): "Allow customers to choose a specific person" UNCHECKED + "Send a meeting invite to the customer" CHECKED (bookers now get a real calendar invite) at the PAGE level — but the selector persisted on the public page, and the cause was NOT cache: Bookings GOTCHA (logged for reuse): each service snapshots its OWN copy of the scheduling policy at creation (the service's "Default scheduling policy" toggle sat OFF), so page-level policy edits never flow to existing services. Fixed AT THE SERVICE: "Allow customers to choose a particular staff for booking" toggled OFF (service-level notifications checked too — the customer meeting invite was already ON there); SELECT STAFF vanished from the public page IMMEDIATELY on the service save, verified live. Also confirmed en route: the service has NO staff assigned — bookings land straight on the empty scheduling-mailbox calendar, which IS the always-open design. (10) /book TARGET SWAPPED to the new address (arethusa b3615b0; site-map row freshened 0f574f6) — a2go.ai/book → outlook.office.com/book/A2go@a2go.ai/ VERIFIED LIVE end-to-end incl. www chain; the permanent-link design proved itself same-week: the engine moved domains and no published link changed. STILL GATED: the sitewide CTA pass fires on the Betsy+Mike word; staff additions (with titles, per the rule) + shared-mailbox notifications (FB-73 pattern) at that time.
New testimonials — "we used to have some, but they are all old" (email)
Customer outreach; each testimonial lands with its permission logged (same clearance rule as figures). Owner re-labeled 8/26 (Stephen: Mike is not on the team — non-team people are reached through the copy owner, never listed as owners).
Per-pillar ranges — closed by Sunday's answers
Benefits no longer attach to individual pillars; results state at the customer level, the way Sunday's slides model them. The port removes the per-pillar range slots. This row makes its farewell appearance with this edition.
Delete the smartphone-analogy section from the architecture arm (audit fix 2, CRITICAL — her own section, her own verdict)
CONFIRMED in her own words 8/26 5:23p: "I did want the iphone analogy taken out, yes." Her FB-70 full site already omits the analogy — the swap port executes the deletion; closes when live. Question row retired.
Vocabulary consolidation on the homepage arms (audit fix 10): ADIP + four pillars up top, branded terms reserved for deep pages
Her copy edit on her arms; team ports.
Monthly practitioner POV publishing cadence (audit fix 12)
Process motion, not a page; blog byline law stands (Betsy Romeri); migrated blog archive is the base. Owner re-labeled 8/26 (same Mike-not-on-team rule).
/about company-facts strip: founding year, team scale, LinkedIn footer link (audit fix 10 part)
Small build. FOUNDING YEAR SETTLED 8/28 (Stephen): 2017, per public records — supersedes "2018 per the Databricks story"; if the story says 2018, the story is not the source for this fact. The port corrects the build's "Founded 2018" ×4 → 2017; the facts strip uses 2017.
Stephen's own bio trim on the Company page ("just log and do"): the nomination line + the orchestration-layer line out; Watson APAC stays
Closed-shipped same day (farewell #014); both surfaces edited (live static + dormant shadow); spawned the bios-signed-off-by-their-own-subject convention.
"I want to create a third homepage/secondary page option for our alternating site"
Closed-superseded (#009), by her own 3:49p words: the build arrived same day as the FB-70 full site and the intent changed from third-arm to SWAP ("I'd like to swap this out for that version"). The arm-C scaffold served as reference (8 of its 9 blocks survive in her homepage); the alternation stays two-arm. Work continues under FB-70.
Jason QA: mobile sideways overflow — "a white column will appear on the right side of the screen"; header CTA off-screen on phone
Closed-verified (#009, Jason 8/26 11:34a): "The site on mobile now has no errant side scrolling, all looks well in that regard." Guard shipped 8/25 (890dd3b) site-wide; his phone retest confirms. The NEW finding in the same reply (menu-button uniformity) = FB-69.
Matthew, on the audit thread: "Here is a tech website I really love" — atlassian.com, cursor.com, domino.ai
Closed-logged (#008, reference): three reference sites into the design lane (now the FB-70 full-site work); no action beyond the handoff to the copy owner.
THE WEBSITE MESSAGING AUDIT ("website ideas" email + A2go_Website_Messaging_Audit docx, addressed Mike+Cesar, cc team): 12 findings, 12-fix prioritized list, 30/60/90 roadmap, 9-block homepage structure
Closed-decomposed (#008) into FB-54..66 per Stephen's 8/26 governance review: ONE tracker — the audit's fix numbering and 90-day windows collapse into this ledger, never a parallel roadmap; CLEARANCES ARE THE COPY OWNER'S DELIVERABLES — she obtains AND logs each sign-off and ships figures with it (naming Mike/Cesar = provenance, not ownership transfer). Scope note: the audit reviewed home (mostly arm B), /supply-chain-ai-agents, /about + the Databricks story — the already-built list (worked examples ARE use cases, Library brief IS the judgment-layer brief, glossary + LLM Edition, blog archive) went to the team in #008. Audit fix 4 (links) folds into FB-31 + the new pages.
"Can Artemis now create links for page sections based on hidden content on our site map? To keep things interesting?"
Closed-answered (#008), yes-with-boundaries (close-unless-corrected): Artemis proposes the FB-31 per-section links from the FULL public inventory incl. quiet pages (/learn, decision examples, /scor library, /glossary, /judgment-layer). Boundaries: staged/noindex/legal-hold targets NEVER linked; links stay STABLE once placed (churn hurts the search goal — variety comes from breadth, not rotation); team reviews before ship. Work rides FB-31 (counts as resolved-by-absorption from 8/25; formal close #008).