Metadata
| Status | done |
|---|---|
| Assigned | agent-276 |
| Agent identity | 8da3b6fc81685ca44a4e15eb35307ab90ce3f0500e9c08b8b0caae848f7d4ce9 |
| Created | 2026-05-02T03:18:52.581706736+00:00 |
| Started | 2026-05-02T03:19:16.402565006+00:00 |
| Completed | 2026-05-02T03:27:04.919499590+00:00 |
| Tags | grant,urgent,form-mapping,critical, eval-scheduled |
| Eval score | 0.72 |
| └ blocking impact | 0.70 |
| └ completeness | 0.78 |
| └ constraint fidelity | 0.70 |
| └ coordination overhead | 0.72 |
| └ correctness | 0.72 |
| └ downstream usability | 0.68 |
| └ efficiency | 0.75 |
| └ intent fidelity | 0.67 |
| └ style adherence | 0.68 |
Description
Description
Erik is at the Google.org form ready to paste. v3.1 (/home/erik/google_ai_competition/workgraph_google_application_FINAL_v3_1.md, committed at 986222f) was written against an older mental model of the form structure. The actual form structure is documented at /tmp/google-org-form-structure.md (Erik pasted it 2026-05-02).
Known mismatches I (the coordinator) can already see:
- Form §29 = Risks (200w); v3.1 §29 = Theory of change (200w). Risks content lives at v3.1 §30.
- Form §30 = Key team members (100w); v3.1 §30 = the science-deliverable risk
- Form §32 = MULTISELECT (not prose); v3.1 §32 may be prose
- Form §38-42 = budget categories with subfields (Name + USD + 100w desc per category)
- Form §43-47 = milestones with subfields (Timeframe + 150w Activities + 100w Outcomes per milestone)
This task does the FULL form-mapping audit and produces a copy-paste-ready guide.
What to read
/tmp/google-org-form-structure.md— actual form structureworkgraph_google_application_FINAL_v3_1.md(commit986222fon main)~/poietic.life/notes/v3-1-final-coherency-review-20260501.md— earlier coherency check (passed but mapped to wrong section numbers)- CLAUDE.md (org info, founder details, COI rules, style)
What to do
Part 1: Map every form question to v3.1 content
For EACH form question §1 through §53 (and certification), produce a row in a paste guide table:
| Form § | Form prompt (short) | Word cap | Type (text/radio/multiselect/yes-no) | v3.1 source location | Paste-ready content | Status |
Status options:
- READY — v3.1 has matching content of correct length and form
- REMAP — v3.1 has the right content but in a different section number; specify the source
- TRIM — v3.1 content is too long for the form cap; produce trimmed version
- EXPAND — v3.1 content too short or shallow; needs more
- NEW — v3.1 has nothing for this question; needs new content from CLAUDE.md / track record / COI rules
- ERIK — Erik must answer (e.g. timezone, exact FTE count, secondary contact name, exact annual budget, certifications)
- NA — optional question with N/A as right answer
Part 2: Hard-mapping decisions
Resolve these specifically:
a. §29 Risks (200w) — pull v3.1's risk content (currently labeled §30 there) into form §29. Recount to ≤200w.
b. §30 Key team members (100w) — describe 3-5 key roles WITHOUT individual names. v3.1 has §26 (team) with Erik/Luca/Vaughn named. Need to abstract to roles for §30. Suggested 4 roles: WorkGraph architect / lead engineer (full-time), pangenomics + clinical genomics PI co-leads (part-time, drawn from founder bandwidth), software engineer for adoption + integration (full-time hire), organizational researcher (part-time, drawn from founder bandwidth). Don't name people; describe expertise that ensures success.
c. §32 multiselect — pick from: Geographic transfer / Adaptation to different sectors / Exponential user growth / Technical & performance maturity / Ecosystem & integration / Community-led expansion / Policy & standards leadership / Other. Recommend 2-3 most defensible based on v3.1 framing (likely Ecosystem & integration + Community-led expansion + maybe Policy & standards leadership for organizational design framework).
d. §13 Open-source outputs — multiselect. v3.1 implies all of: Software (MIT/Apache), Datasets (CC0/CC-BY for computation graph corpus), Pubs (Open Access). Recommend: Software + Datasets + Pubs.
e. §14 Geographic regions — multiselect. Recommend: Global (since lab adoption target is worldwide).
f. §15 Project stage — radio. WorkGraph is in active production use across founder labs and incorporated the company; that's beyond Proof of Concept. Recommend: Validated Methodology.
g. §22 Existing dataset? — Yes/No. WorkGraph has computation graph datasets. Recommend: Yes.
h. §24 All outputs freely available? — Yes/No. v3.1 commits to MIT and CC-BY. Recommend: Yes.
i. §27 AI Maturity — radio: AI First / AI Adoption / AI Exploration / AI Interest / None. v3.1 is about AI orchestration and uses AI throughout. Recommend: AI First.
j. §31 Partners — v3.1 mentions IGVF Consortium, possibly HPRC. Recommend: list HPRC as 'Existing partner' and IGVF as 'Existing partner' if defensible. Otherwise mark this section optional.
k. §37 Funding request — radio. v3.1 = $1.5M. Confirm.
l. §38-42 Budget breakdown — v3.1 has 4 categories (k Personnel, k Tech Dev, k Equipment, k Amplification = .5M). Map each to §38-§41 with Name + USD + ≤100w description. §42 = optional, leave blank or mark NA.
m. §43-47 Milestones — v3.1 has 4 milestones (§43-§46). Map to form §43-§46 with Timeframe + ≤150w Activities + ≤100w Outcomes per milestone. §47 = optional.
n. §48-52 Ethics — Yes/No each:
- §48 (commercial contracts with Google): No (none directly related to project)
- §49 (govt officials in org): YES (founders are at public universities — UTHSC for Erik, MGH+HMS+Broad mostly private but HMS may count, UCL for Vaughn — public)
- §50 (govt persons involved in project): YES (founder effort allocations from public-university salaries)
- §51 (law enforcement): No
- §52 (sanctioned regions): No
o. §53 Explanation — must explain §49 and §50 yes answers. Draft a 2-3 sentence explanation: 'Founders Garrison and Tan are employees of public universities (UTHSC, UCL). Pinello is at MGH/Harvard/Broad. None of these institutions are direct project recipients; project funding flows to Poietic PBC. Founder time on the project is contributed under university policies for outside engagement.'
p. Certification — Erik must check Yes.
Part 3: New content where needed
For §1-10, §12, §13, §14, §15, §22, §24, §27, §32, §37, §48-52: produce the answers as a single 'Section I-VI mechanical fields' block at the top of the paste guide.
For §30, §32, §53: draft the new content needed.
For §43-§47: ensure each milestone has BOTH Activities ≤150w AND Outcomes ≤100w as separate fields.
Part 4: M4 / H7 verdict refresh
Erik asked: 'm4 m7 seem hardcore? could we just link to the workgraph traces?'
Recommendation: with both ulivo trace URLs now public (https://ulivo.poietic.life/wg/feeds/incorporation-trace/ and https://ulivo.poietic.life/wg/feeds/poietic-website/), §36 (public presence) and §28 (technical feasibility) can lead with the trace URLs as primary evidence. The stale GitHub repos (deep-research-competition with KRAS framing, phr-methodology-comparison with 1 commit) are demoted from being primary evidence to being incidental org-page residue. Risk: a reviewer who org-spelunks finds them. Mitigation: archive whenever Erik has 30 seconds at GitHub admin (still recommended); not a blocker.
Include a paragraph in the paste guide: 'M4/H7 not blockers given the live traces. Archive at convenience; submission can proceed without.'
Output
Write ~/poietic.life/notes/v3-1-form-mapping-paste-guide-20260502.md:
- Headline verdict: Is v3.1 ready to paste? With what remapping?
- Section I-VI mechanical fields block — every Erik-only or short-answer field with the recommended answer
- Paste guide table — every form §, with v3.1 source and paste-ready content
- New content drafts — §30 (team roles), §32 (multiselect picks), §53 (govt-persons explanation)
- Erik-only items — what only Erik can supply (annual budget, FTE count, secondary contact, certifications, attachments)
- M4/H7 verdict — submission can proceed without; archive at convenience
- Submission order recommendation — which fields to paste in what sequence to minimize back-and-forth
wg log a one-paragraph summary on this task.
Constraints
- HARD: respect every word cap from the form structure file. Recount.
- HARD: no em-dashes. No PI language. Founder order Erik / Luca / Vaughn (where named, only in §26-equivalent and contact info).
- HARD: §30 must NOT name individuals (form requires roles only).
- HARD: §29 must contain Risks content within 200w.
- HARD: every required form field has a recommendation or 'ERIK' tag.
- HARD: produce paste-ready text for every prose field (Erik shouldn't have to edit; just copy and paste).
Validation
- Every form question §1-§53 + certification covered in the paste guide table
- §29 Risks content ≤200w
- §30 team roles ≤100w, no individual names
- §32 multiselect choices recommended
- §53 explanation drafted for govt-persons
- All multiselect / radio fields have recommended answers
- Erik-only items clearly listed
- M4/H7 link-to-traces verdict included
-
Output at
~/poietic.life/notes/v3-1-form-mapping-paste-guide-YYYYMMDD.md
Depends on
Required by
Log
- 2026-05-02T03:18:52.578255067+00:00 Task paused
- 2026-05-02T03:18:52.941987442+00:00 Task published
- 2026-05-02T03:19:15.421445788+00:00 Lightweight assignment: agent=Default Documenter (8da3b6fc), exec_mode=full, context_scope=task, reason=Documenter role is optimal for this structured documentation creation task requiring careful reading of source materials, mapping decisions, content drafting, and production of a comprehensive paste guide with strict word-count constraints.
- 2026-05-02T03:19:16.402570146+00:00 Spawned by coordinator --executor claude --model opus
- 2026-05-02T03:19:29.600209730+00:00 Starting form-mapping audit. Reading actual form structure, v3.1 application, and prior coherency review.
- 2026-05-02T03:19:29.806930336+00:00 ERIK PROVIDED MORE RELIABLE SOURCE: the actual saved HTML of the Google.org form is at: ~/Downloads/Google GIC Submission Manager - Google.org Impact Challenge_ AI for Science.html (301 KB) Use this as the PRIMARY source of truth instead of /tmp/google-org-form-structure.md. The HTML preserves: - Exact word caps and character limits per field - Sub-field structures (e.g. budget categories with Name + USD + Description as separate inputs) - Multiselect option lists with exact wording - Required vs optional flags - Help text and tooltips that may clarify intent - Any pre-filled or default values Workflow: 1. Read the HTML file. Use grep / parse to extract: every form field, its type (text/radio/multiselect/yes-no), its word-count or character cap, its required flag. 2. If a field is in the HTML but NOT in /tmp/google-org-form-structure.md, prefer the HTML. 3. If they conflict on word caps or field structure, prefer the HTML and note the discrepancy. 4. The pasted text I gave you (/tmp/google-org-form-structure.md) is a useful summary but may have lost rendering details — treat as secondary. Everything else in the original brief stays the same. Goal is paste-ready content for every field, with v3.1 remapped where section numbers differ.
- 2026-05-02T03:23:07.745312832+00:00 Drafts validated: §29 184/200w, §30 85/100w (no names), §53 75w, §28 91/100w, §36 95/100w. Writing paste guide now.
- 2026-05-02T03:26:43.477765225+00:00 Wrote paste guide at ~/poietic.life/notes/v3-1-form-mapping-paste-guide-20260502.md. All 53 form questions + certification mapped. New drafts: §29 (184/200w Risks), §30 (85/100w roles, no names), §53 (75w govt-persons explanation), §28 trace-led replacement (91/100w), §36 trace-led replacement (95/100w). Five corrections vs v3.1 flagged: §15 PoC->Validated, §29 Theory-of-Change->Risks, §30 NEW, §49/§50 No->Yes, §53 NEW. Eleven Erik-only items listed. M4/H7 verdict: not blockers given live ulivo traces, archive at convenience. Submission order recommendation provided in §7.
- 2026-05-02T03:27:01.427066542+00:00 Committed paste guide: 5b68c71 — pushed to remote (poietic-pbc/poietic.life main).
- 2026-05-02T03:27:04.919513245+00:00 Task pending eval (agent reported done; awaiting `.evaluate-*` to score)
- 2026-05-02T03:29:09.370217966+00:00 PendingEval → Done (evaluator passed; downstream unblocks)