Skip to content

docs(profile): reground public org README + FUNDING to current canon#14

Open
byteworthy wants to merge 4 commits into
mainfrom
chore/reground-public-profile
Open

docs(profile): reground public org README + FUNDING to current canon#14
byteworthy wants to merge 4 commits into
mainfrom
chore/reground-public-profile

Conversation

@byteworthy

Copy link
Copy Markdown
Contributor

Why

The public org profile README (the github.com/Upstream-Intelligence landing page) was last touched 2026-05-17 and was missed by the 06-14 canon grounding. It still advertised the retired Founding Pioneer program ($49/mo locked for life), a $349 'Pro' tier, and the old pack-led / ABA-first positioning.

What changed (mirrored from upstream-v2/marketing/src/lib/site.ts, the canonical public-copy source)

  • Pricing -> Assist (free) · Pay-as-you-go (from $18) · Co-pilot (from $1,500/mo + $12/sub) · Autopilot (custom)
  • Positioning -> 'nine specialty packs/verticals' becomes one connected platform
  • Specialties -> retired ABA/SNF/PT-OT/dental/dialysis list becomes the current Infusion / Oncology / Rheumatology / GI / Neurology / Cardiology / Orthopedics / Pain Management
  • How it works -> the canonical Signal -> Brief -> Action -> Approval -> Outcome loop
  • Counts -> skills 8 -> 9; 'last updated' -> 2026-06-17
  • FUNDING.yml -> removed dead upstream.cx/pioneer link

Verified clean: no Pioneer/$49/$349/pack/vertical drift, no retired specialties, no em/en dashes.

Needs your eyes before merge (couldn't verify against live product)

  1. Synthetic alert example previously used an ABA CPT (97155 / F84.0). I genericized it to 'a high-volume specialty code' rather than fabricate a clinical example. You may want a real current-specialty example.
  2. MCP tool count was '12 tools across 5 categories' (stale; opencode just added Da Vinci/LLM Council tools). I softened it rather than assert a number. Worth setting the real current count.
  3. Operational claims kept verbatim (detection engine names, connectors, free-tool names, SOC 2 status). These predate me; confirm they're current.

The public org profile (the github.com/Upstream-Intelligence landing page) was
last touched 05-17 and predated the 2026-06-14 canon grounding, so it still sold
the retired Founding Pioneer program and the old pack-led positioning.

Regrounded against marketing/src/lib/site.ts (the canonical public-copy source):
- Pricing: removed Pioneer $49/locked-for-life + $349 'Pro'; now Assist (free),
  Pay-as-you-go (from $18), Co-pilot (from $1,500/mo + $12/sub), Autopilot (custom).
- Positioning: 'nine specialty packs / verticals' -> one connected platform.
- Specialties: retired ABA/SNF/PT-OT/dental/dialysis/imaging/home-health list ->
  current Infusion, Oncology, Rheumatology, GI, Neurology, Cardiology, Orthopedics,
  Pain Management.
- How it works: reframed to the canonical Signal -> Brief -> Action -> Approval ->
  Outcome loop (approval-gated execution).
- Counts: skills 8 -> 9; stats 'last updated' bumped to 2026-06-17.
- FUNDING.yml: removed dead upstream.cx/pioneer link.

No em/en dashes. Operational specifics left as-is for separate verification (see PR).
The public org profile (profile/README.md + FUNDING.yml) drifted unguarded for a
month and still sold the retired Pioneer program until the 2026-06-17 reground.
Root cause: this repo had no canon guard while every sibling does. Adds
scripts/check-canon.sh + .github/workflows/canon.yml (mirrors the sibling guards)
to ban retired Pioneer/pricing/pack framing on the public surfaces. Self-tested:
passes clean, proven able to fail on an injected banned line.
The guard scanned only profile/ + FUNDING.yml, but BRAND.md, SECURITY.md,
SUPPORT.md, CONTRIBUTING.md and the other org-default files are public surfaces
too and were unguarded (same condition that let profile/README.md drift). Now
scans repo-wide. Self-tested: passes clean, proven able to fail on drift injected
into BRAND.md (a non-profile file).
The broadened repo-wide scan flagged its own workflow file: the step name
contained 'Pioneer', so the guard self-triggered and failed for the wrong reason.
Reworded the step to 'Verify no retired pricing or program framing'. check-canon.sh
is already excluded; this removes the last self-reference. Guard now passes clean
repo-wide and still fails on real drift (tested against SECURITY.md).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant