Skip to content

feat(coordination): CTO+PM consolidated 30-day operational plan for Cursor#82

Merged
stephenlutar2-hash merged 1 commit into
mainfrom
cursor/perplexity-cto-pm-operational-plan-2026-05-29
May 30, 2026
Merged

feat(coordination): CTO+PM consolidated 30-day operational plan for Cursor#82
stephenlutar2-hash merged 1 commit into
mainfrom
cursor/perplexity-cto-pm-operational-plan-2026-05-29

Conversation

@stephenlutar2-hash
Copy link
Copy Markdown
Member

Summary

Consolidated operational plan generated by the CTO+PM dual-lens audit subagent (2026-05-29).

Supersedes all previous CURSOR_* directive files. This is the single canonical plan Cursor follows for the next 30 days (May 30 – June 26, 2026).

Source: Founder verbatim 2026-05-29 18:27 EDT — 'Get a cto and program manager agent to go through the whole thread anything we missed or need to upgrade have them draw the plan make it real and operational'

What This File Contains

  • Acknowledgment of Cursor's write-access situation and proxy pattern
  • Week-by-week tasks (W1: Lean CI; W2: demo core; W3: Warhacker; W4: post-Warhacker close)
  • Per-task acceptance commands (grep/curl/pnpm test verification)
  • Hand-off points where Cursor produces output for Perplexity to proxy or merge
  • Acceptance criteria for 'done' (12 binary checkboxes)
  • Doctrine v6 reminders
  • Daily-status convention

Doctrine v6

  • Commits signed with -s
  • No marketing superlatives
  • No emoji in headers
  • a11oy#57, amaru#46, sentra#45 protected (DO NOT TOUCH)
  • Branch protection not toggled
  • arXiv: stage only (founder awaiting endorsement)

Context Files

Full gap audit: closeout/CTO_PM_THREAD_AUDIT.md
Founder summary: FOUNDER_30_DAY_PLAN_SUMMARY.md
Final closeout: closeout/CTO_PM_AUDIT_FINAL.md

Signed-off-by: Stephen P. Lutar stephen@szlholdings.com
Co-authored-by: Perplexity Computer perplexity@szlholdings.com

…sor directive

Supersedes all previous CURSOR_* directive files. Single canonical plan
for Cursor covering Weeks 1-4 (May 30 – June 26, 2026).

Key deliverables assigned to Cursor:
- W1: Lean CI fix (4 modules), L3 devcontainer, CodeQL sweep
- W2: FG-S1→S4 TypeScript, amaru adversarial regression, rosie replay
- W3: Theorem-receipt graph viewer HF Space (Warhacker demo)
- W4: Putnam harness v2, compliance stubs, szlholdings.com

Hand-off points, acceptance criteria, and daily-status convention
included per Doctrine v6.

Co-authored-by: Perplexity Computer <perplexity@szlholdings.com>
Signed-off-by: Stephen P. Lutar <stephen@szlholdings.com>
Signed-off-by: Perplexity Computer <stephen@szlholdings.com>
@stephenlutar2-hash
Copy link
Copy Markdown
Member Author

Cursor — start working. Founder directive 2026-05-29 19:14 EDT.

Everything is unblocked. Pick up the queue now.

Read first (in this order)

  1. .github#84coordination/CURSOR_READ_THESE_TWO_FIRST.md (canonical pointer; supersedes nothing but consolidates 11 open coordination PRs)
  2. This PR (feat(coordination): CTO+PM consolidated 30-day operational plan for Cursor #82) — CTO+PM 30-day operational plan, 469 lines
  3. .github#83 — theorems instillation plan, 472 lines, ~238h prioritized work
  4. Sorry-count correction comment — canonical sorry count is 7, not 6

Canonical numbers (DO NOT REVERT)

  • 217 Lean theorem declarations + 12 axioms across 53 .lean files
  • 7 executable sorries (verified by grep)
  • 35/35 anchor formulas instilled
  • TH10 axiom-structured pending CAUCHY_ND (not machine-checked)
  • liu_hui_pi_converges is Lean axiom (not proved theorem)
  • Putnam 8.3% (1/12) honest baseline

Start here — Tier 1 (do these first)

From #83 prioritized queue:

  1. Uniqueness.lean:120 — CAUCHY_ND (closes the TH10 axiom-structured caveat — highest investor-facing leverage)
  2. TwoWitness.lean:163 — straightforward, unblocks two-witness chain
  3. SBOMProvenance.lean:109 — provenance gate, needed for UDS catalog-grade claim
  4. Wire QEC / Wheeler / Shannon / DPI as named formula gates in amaru / rosie / agi-forecast

Just merged (rebase if you need it)

Working pattern (this stays in effect)

  • Push branch from your local environment (your cursor[bot] GHA identity has permission: none — known, accepted)
  • If gh pr create fails: drop the branch name in the latest daily-status PR (chore(coordination): proxy Cursor's daily-status to .github #76 pattern) — Perplexity proxies in <5min
  • Sign every commit with -s
  • Doctrine v6: no marketing superlatives, no emoji in ## ### headers
  • DO NOT TOUCH: a11oy#57, amaru#46, sentra#45 (DRAFT relicense, founder IP HOLD)

Acceptance for "work started"

Post a daily-status update (new PR or comment on #76) within the next working session showing:

  • Which Tier 1 sorry you're attacking (or which formula-gate wiring you started)
  • Current branch name
  • Estimated landing time

Operational floor

Money is not a problem. Bank meeting is May 30. Warhacker is June 16-20. Pace > perfection. If anything blocks you, post the block in the same daily-status PR — Perplexity unblocks within minutes.

Founder approved 2026-05-29 19:14 EDT. Go.

@stephenlutar2-hash
Copy link
Copy Markdown
Member Author

Consolidated CTO + PM + Putnam directive — 2026-05-29 20:30 EDT

Founder ran three audit agents tonight: CTO (32 KB), PM (27 KB), Putnam-beat (30 KB). Output consolidated to /home/user/workspace/szl/audit_2026-05-29_evening/CURSOR_CONSOLIDATED_DIRECTIVE.md (full file in #84's coordination dir on next push).

Headline

Founder directive: "I want us to beat Putnam — we have the answers here, we just need it built." Current = 1/12 (8.3%) honest baseline. ~38h Cursor build (P1→P4) realistically targets 3-4/12 (25-33%) by Warhacker June 19.

What Perplexity finished tonight (no Cursor action needed)

  • All 6 critical investor PDFs regenerated with honest 217/7/TH10 language — zero stale "76 theorems" remain. Bank meeting May 30 is safe to walk into.
  • a11oy#94 PR opened (Cursor's UDS frontier gap map, proxied)
  • a11oy#99 PR opened (Cursor's HF deep-dive staged-safe, proxied)
  • .github#84 PR opened (CURSOR_READ_THESE_TWO_FIRST pointer)
  • agi-forecast#41 MERGED (protection toggle pattern)
  • Sorry-count correction posted to .github#83 (canonical = 7, not 6)
  • Cursor auth diagnosis posted to .github#76 with paste-and-proxy workaround

Cursor — start TONIGHT (Tier 1, ~14.5h total, no blockers)

# Task Effort
T1 Fix Lean CI red (4 pre-existing build failures) — blocks every sorry-close 8h
T2 Close TwoWitness.lean:163 sorry (2-line Mathlib 4.13 fix) 2h
T3 BFT single-signer caveat in uds-mesh/README.md 30min
T4 Verify + merge vsp-otel#43 (L4 anatomy PASS trigger) 1h
T5 Verify + merge sentra#65 (L7 anatomy PASS) 1h
T6 Embed Plotly 2.30.0 inline in uds_demo_usb_v2/ui/index.html (Warhacker USB demo offline) 2h

Tier 2 (Week 1, ~52h) — closes 5 of 7 sorries

T7 LyapunovCLT.lean → close 2 MadhavaBound sorries (10h)
T8 HoeffdingAzuma.lean → close 2 PACBayes sorries (8h)
T9 Close SBOMProvenance.lean:109 (~8h)
T10 GOVERNANCE_THREAT_MODEL.md + 3 adversarial tests (4h)
T11 Wire L6 formula_witness in 4 gate files (6h)
T12 amaru adversarial regression (8h)
T13 rosie receipt replay (8h)

Brings sorry count from 7 → 2. Remaining: Uniqueness.lean:120 CAUCHY_ND (L, 40h Tier 3) and TH10 stays axiom-structured (honest).

Tier 3 — PUTNAM BEAT (Week 2-3, ~38h)

P1 agi-forecast/runtime/src/putnam_corpus_loader.ts (8h) — load 12 problems
P2 a11oy/packages/putnam-router/src/domain_classifier.ts (12h) — route to 35 anchor formulas
P3 a11oy/packages/putnam-router/src/formula_scaffold.ts (10h) — CoT scaffold prompts
P4 agi-forecast/runtime/src/putnam_harness_v2.ts (8h) — multi-judge n=3 + per-problem receipts
P5 stretch Lean type-checker REST bridge (16h)
P6 vsp-otel/runtime/src/formulas/putnamWitness.ts (4h) — OTel emitter
P7 proof_ledger.jsonl append per run (2h)

Honest target: 2-4/12 (16-33%) after P1-P4. Walk-away publishable number. Doctrine: never inflate.

Tier 4 — UDS catalog-grade (Warhacker June 16-20, ~24h)

U1 Push container to ghcr.io/szl-holdings/vessels:uds-v0.3.0 (4h, founder approval)
U2 Package CR with sso: block (8h)
U3 NetworkPolicy deny-all egress (4h)
U4 ServiceMonitor Prometheus integration (4h)
U5 cosign uds-v0.3.0 signing (4h, founder secret)

PR queue (highest urgency)

Auth pattern (until you fix in Cursor settings)

cursoragent has 10 PENDING per-repo invitations. Only cursoragent can accept (GitHub API limitation). Use paste-and-proxy via .github#76:

[PROXY-LAND repo=<repo> target=main commit-msg="..."]

### File: path/to/file
```language
<full file content or git format-patch output>

Perplexity lands via Contents API in <5min. Proven on a11oy#94, a11oy#99, .github#84.

### Doctrine reminders (every commit)

- Sign `-s`: `Signed-off-by: Cursor Agent <cursoragent@cursor.com>`
- Co-author: `Co-authored-by: Lutar, Stephen P. <stephen@szlholdings.com>`
- No marketing superlatives, no emoji in headers
- Canonical numbers locked: **217 declarations / 7 sorries / 12 axioms / 35-35 anchors / TH10 axiom-structured pending CAUCHY_ND / Putnam 8.3% baseline**

### Founder action items (D1-D10 in full directive)

D1 Insurance packet 26 TBDs (state of incorp, EIN, cash, broker name/email)
D2 UDS trademark non-objection email to Andrew (June 2)
D3 ORCID bio empty
D4 `HF_TOKEN` org-secret confirm
D5 `szlholdings.com` not live
D6 Pilot pricing `$TBD/month`

Full directive: `/home/user/workspace/szl/audit_2026-05-29_evening/CURSOR_CONSOLIDATED_DIRECTIVE.md`. Read order at top of file. Tier 1 starts tonight.

stephenlutar2-hash added a commit that referenced this pull request May 30, 2026
## Summary

Adds `coordination/CURSOR_HANDOFF_2026-05-29_EVENING.md` (395 lines).
Top-to-bottom audit Cursor needs before next loop.

## What's in it

| § | Section |
|---|---|
| 0 | TL;DR — live state table of every open PR across the four sibling
repos |
| 1 | Write-access mechanic — Cursor App already has org-wide
`contents:write`; runtime is auth'd as wrong identity |
| 2 | Branch protection inventory (required reviews + status checks per
repo) |
| 3 | **Lean lake fix plan** — 17 failing modules, 120 errors, full
per-file diagnosis, 9-step ordered PR sequence |
| 4 | Non-Lean PRs ready to merge tonight on review (a11oy
#94/#99/#100/#101/#111, agi-forecast#42, .github #82#86) |
| 5 | Doctrine v6 reminders |
| 6 | Supporting artifacts in Perplexity workspace |
| 7 | Direct order for tonight |

## Key findings

- **Every non-Lean PR is green** and blocked only by
`required_approving_review_count = 1`. Founder can review-and-merge in 5
min.
- **`lutar-lean/main` itself is red** — 120 errors across 17 modules.
Targeted PRs #98/#99/#100 each fix 1–11 errors but inherit the rest from
main. The lake must be repaired top-down before any Lean PR can land
green.
- **Cursor App** (`cursor`, app_id 1210556, installation 136458231) is
installed at the org level on szl-holdings with `repository_selection:
all`. Permissions include `contents:write`, `pull_requests:write`,
`workflows:write`. The proxy pattern is only needed because the runtime
authenticates as `cursor[bot]` user instead of the App installation
token. Doc gives two clean paths to fix.
- **NEW PR `a11oy#111`** opened tonight by Perplexity proxy — promotes
Cursor's existing `cursor/fix-dependabot-label-config-2f18` branch
(which had no PR) to a real PR. Dependabot label fix unblocked.
- **a11oy#94** rebased against main tonight (was BEHIND); now mergeable.

## Doctrine

- Signed off with `-s`
- No marketing superlatives
- No emoji in headers
- No touches to a11oy#57, amaru#46, sentra#45
- Honest sorry/declaration/TH10 counts preserved

## Order

Cursor reads this on next loop. Founder reviews + merges. Then Cursor
executes Section 7 step list.

Co-authored-by: Perplexity Computer <perplexity@szlholdings.com>
Signed-off-by: Stephen P. Lutar <stephen@szlholdings.com>

Signed-off-by: Stephen P. Lutar <stephen@szlholdings.com>
Co-authored-by: Stephen P. Lutar <stephen@szlholdings.com>
Co-authored-by: Perplexity Computer <perplexity@szlholdings.com>
stephenlutar2-hash added a commit that referenced this pull request May 30, 2026
## Summary

Adds `coordination/CURSOR_READ_THESE_TWO_FIRST.md` so Cursor isn't lost
across 11 open coordination PRs.

## Why

11 open coordination PRs (#71, #72, #73, #75, #76, #77, #78, #79, #80,
#82, #83) — Cursor needs a single canonical entry point.

## What it does

- Points Cursor at **#82** (canonical 30-day operational plan,
supersedes #71/72/73/75/77/78/79/80)
- Points Cursor at **#83** (canonical theorems plan, supersedes
#78/79/80)
- Includes the **sorry-count reconciliation: 7 not 6** (PhD audit ground
truth)
- Doctrine v6 quick-reference
- PR proxy pattern explanation for `cursor[bot]` `permission: none`
issue
- Sign-off pattern for all Cursor commits

## Doctrine

- No marketing superlatives
- No emoji in `## ###` headers  
- Signed off with `-s`
- Honest sorry count (7), honest declaration count (217), honest TH10
status (axiom-structured)

## Founder-approved

Approved 2026-05-29 by Stephen Lutar.

Signed-off-by: Stephen Lutar <stephen@szlholdings.com>
Co-authored-by: Perplexity Computer <agent@perplexity.ai>
@stephenlutar2-hash stephenlutar2-hash merged commit 256b57e into main May 30, 2026
5 checks passed
@stephenlutar2-hash stephenlutar2-hash deleted the cursor/perplexity-cto-pm-operational-plan-2026-05-29 branch May 30, 2026 05:46
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