← graph

site-audit-read

Site audit: read the whole landing page, find other rough/uncanny spots beyond the one Erik called out

Metadata

Statusdone
Assignedagent-266
Agent identity3184716484e6f0ea08bb13539daf07686ee79d440505f1fdf2de0357707034c3
Created2026-05-02T02:43:11.361708834+00:00
Started2026-05-02T02:52:15.986337593+00:00
Completed2026-05-02T02:58:55.238685071+00:00
Tagsgrant,urgent,website,site-audit, eval-scheduled
Tokens1301794 in / 11780 out
Eval score0.92
└ blocking impact0.92
└ completeness0.92
└ constraint fidelity0.40
└ coordination overhead0.93
└ correctness0.90
└ downstream usability0.94
└ efficiency0.88
└ intent fidelity0.73
└ style adherence0.95

Description

Description

Erik flagged ONE rough paragraph on poietic.life as 'too self-aware, kind of uncanny.' The text talks about itself meta-philosophically ('WorkGraph is the instrument, not the whole institution. It gives Poietic a proof surface for the theory of organizational patterns'). His diagnosis: when the reader asks 'can you run a team on it?' they should get YES + how, not a paragraph about proof surfaces.

This task surveys the REST of the live landing page (https://poietic.life and the deployed file at ~/poietic.life/index.html) for similar rough spots — anywhere the text describes itself, philosophizes, or fails to answer 'what does this company do? can you use it?' in plain terms.

What to do

  1. Read the full deployed landing page text (~/poietic.life/index.html).
  2. WebFetch https://poietic.life for an over-the-wire view (any deploy lag).
  3. For each section / paragraph, judge:
    • Does it answer a real question a reader has?
    • Does it talk about itself instead of doing the work?
    • Does it use abstractions where concrete examples would land harder?
    • Does it overclaim or underclaim relative to what's defensible?
  4. Identify rough spots beyond Erik's flagged paragraph, with section anchor, exact rough text, and one-sentence diagnosis of what's wrong.
  5. For each rough spot, propose a tighter replacement (or a 'drop' verdict). Keep the dark monospace aesthetic and dogfooding voice.

Constraints

  • Erik's rough paragraph diagnosis is the calibration. If a section reads cleanly by that bar, leave it alone. Don't manufacture rewrites.
  • Founder section was just added (M3 fix) — don't touch unless something is genuinely wrong.
  • Public benefit statement is verbatim from Delaware Certificate of Amendment — do NOT rewrite it.
  • No em-dashes.
  • The published incorporation trace at https://ulivo.poietic.life/wg/feeds/incorporation-trace/ is now real — note where the website should link to it.

Output

~/poietic.life/notes/website-site-audit-20260501.md (under 1000 words) with:

  1. Headline verdict: how rough is the website overall? (one paragraph)
  2. Rough spots table: anchor / current text / diagnosis / proposed fix
  3. Sections that read clean (don't touch)
  4. Missing link: where to add the incorporation-trace URL
  5. Recommended deploy order if multiple fixes land

wg log a one-paragraph summary on this task.

Validation

  • Whole landing page read end-to-end
  • WebFetch confirms over-the-wire matches deployed file
  • Erik's flagged paragraph used as calibration, not over-reach
  • Each rough spot has a proposed fix or explicit drop verdict
  • Incorporation-trace link placement recommended
  • Output at ~/poietic.life/notes/website-site-audit-YYYYMMDD.md

Depends on

Required by

✉ Messages 5 messages (5 unread)

  1. #1user2026-05-02T02:47:35.317708930+00:00read
    ERIK CORRECTION: AutoHaiku was already dropped from the live landing page. Earlier WebFetch result was stale. Do NOT flag AutoHaiku as a rough spot to remove — it's already gone. Trust the live page (https://poietic.life and ~/poietic.life/index.html) over my earlier characterization.
  2. #2site-audit-read2026-05-02T02:49:01.345433485+00:00read
    Acknowledged. My audit was based on the live page + WebFetch and never mentioned AutoHaiku — it appears only in stale worktrees (.wg-worktrees/), not in ~/poietic.life/index.html. Audit at ~/poietic.life/notes/website-site-audit-20260501.md flags 4 rough spots (hero subhead, mission lead, built-section lead, repo-stat Theory tag) using your 'proof surface' paragraph as calibration.
  3. #3user2026-05-02T02:50:42.706353583+00:00delivered
    ERIK SPECIFIC ROUGH SPOT (please flag this in your audit): too many 'live' / 'live demo' callouts repeated across the landing page. Currently appears in:
- 'Open WorkGraph →'
- 'Read the mission'
- 'Live demo'
- 'PBC product · live demos · theory · organizational practice ·'
- 'live WorkGraph ↗'

Erik's call: just ONE 'live' callout is needed. The repetition reads as trying-too-hard. Recommend consolidating to a single live-link CTA (whichever placement is most defensible — likely either the hero or the WorkGraph card, not both, not in the breadcrumb taxonomy). The other CTAs should be plain ('Open WorkGraph', 'Read the docs', 'Contact') without 'live' modifier.

Include this in your rough-spots table with priority 'must-fix' (Erik called it out directly).
  4. #4user2026-05-02T02:57:02.524904527+00:00read
    ERIK QUESTION (please flag in your audit if true): are there too many cards on the landing page? Specifically the 'What we've built' or equivalent product/demo cards section. If too many, recommend which to consolidate or drop. Apply Erik's calibration: cards should each answer a distinct reader question; redundancy = noise.
  5. #5site-audit-read2026-05-02T02:58:33.566655904+00:00read
    Re: card count, added section to audit. Three section-03 cards (Incorporated/Grant/PHR) each answer a distinct reader question (legal-operational, deliverable-production, technical research). None is redundant by your bar. Recommend keeping all three. Mild caveat: Card 02 has faint recursion (used WG to apply for grant funding WG); defensible now that Card 01's incorporation trace is publicly verifiable. If you want to err clean, Card 02 is the only drop candidate, but cutting it loses the shared-writing demo. My read: keep all three. Founder section (3 cards) also passes.

Log