Board 2026-08-23 — Owner goals — FINAL MINUTES #54

New Issue

2026-08-23T23:44:45-04:00

logan commented

2026-08-23 23:44:45 -04:00

Agenda

The owner answered the four goal-setting questions on 2026-08-23. Rule on them, resolve the contradiction between them, and produce a standing goals document every future sitting measures against.

The four answers, verbatim

Q	Answer
Q1 SUCCESS	C) Second income — $2-5k MRR at 12 months; support and sales become a permanent weekly obligation and the roadmap follows customers, not curiosity.
Q2 WHO WINS	B) Customers first — friends ride the same product paying users get; anything friends would like that doesn't sell gets deferred.
Q3 MY TIME	A) Under 5h, hobby — agents build, owner reviews.
Q4 QUIT	A) Nobody pays — 12 real conversations, zero conversions -> keep it running for friends, kill the commercial track.

The conflict, named

Q1, Q3 and Q4 cannot all be true as written, and averaging them would be the worst available outcome.

The scarce resource is not engineering. Agents build; agent-hours are effectively unbounded. The scarce resource is owner presence in the things that cannot be delegated: sales conversations, the business entity, ToS/Privacy/AUP sign-off, E&O, credential rotation, Firebase/GCP console actions, and support.

The arithmetic, stated so nobody is surprised later:

$2-5k MRR at a blended ~$75/mo is 27-67 paying customers.
BUSINESS_MODEL.md §7's own walk-in funnel is ~5 conversations per paid customer. 27 customers is ~135 conversations. At ~1h each with travel, prep and follow-up that is ~135 of the ~260 owner-hours the whole year contains.
That leaves nothing for support. BUSINESS_MODEL.md §4.3 additionally requires selling the same county repeatedly, and Westchester saturates long before 27 accounts.
The binding constraint is support, not sales. 27 customers at 15 min/customer/week is 351h/yr — more than the entire budget. At 10 min/customer/month it is ~54h/yr and fits.

That last line is the whole finding, and it is actionable: at 5h/week, any customer who needs the owner personally is a customer we cannot afford. Self-serve signup, self-serve billing and zero-touch onboarding stop being SaaS polish (SAAS_PLAN.md §5, "makes it good") and become the revenue mechanism itself.

Decisions

1. $2-5k MRR at 12 months is ratified as an ASPIRATION, not a plan. The 12-month plan of record is 5-15 paying customers, roughly $400-1,200 MRR, conditional on the beachhead surviving its kill criterion.

I am not revising the owner's number down; I am refusing to pretend the current plan produces it. $2-5k MRR is a 24-36 month figure at this time budget unless decision 3 lands, in which case it is reachable sooner because customer count stops being a function of owner-hours.
Owner: CEO. Recorded in GOALS.md. Reviewed at the first sitting after the beachhead test returns.
Cost: we give up the comfort of a headline number that everyone half-believes. Every future sitting now has a plan-of-record it can be measured against and can fail against.

2. The under-5h/week budget is a HARD CONSTRAINT, not a preference. Any plan that requires more owner-hours than it has is rejected at the board, not attempted and missed.

Concretely: BUSINESS_MODEL.md §7's sequence — 40 walk-ins, trial hand-holding, node-operator recruiting and shipping B1-B2 concurrently — is permanently un-schedulable under this constraint and stays unratified. It is superseded by decision 4.
Owner: CEO, enforced at every sitting. Effective immediately.
Cost: the fast version of this business is off the table. We are trading calendar time for the owner's life, which is what Q3 asked for.

3. Self-serve is promoted from "makes it good" to a REVENUE BLOCKER. The scaling channel must not consume owner-hours per customer.

SAAS_PLAN.md §5 lists onboarding wizard, plan-limit enforcement, real org_api_keys and email as non-blocking. For decision 1's aspiration they are the blocker. This does not unsuspend Stripe (SAAS_PLAN.md §4 stays SUSPENDED until Gate B) and does not authorise publishing a price (Gate A stands). It changes what agents build after the beachhead test returns.
Owner: CTO, to sequence. Date: proposal due at the first sitting after the beachhead test reports.
Cost: correlation quality work (#5, #35, #28, #29) drops below funnel work in the queue. Under Q2 that is the correct trade and it is uncomfortable, because correlation is the differentiator.

4. The beachhead test is RE-SCOPED to fit the budget, and its clock starts now — before Gate A.

#42 open item 6 ("willing to walk into ~40 local businesses") is answered YES by implication: the owner chose a kill criterion that requires 12 conversations, and choosing a kill criterion is accepting the obligation to run the test. But the ask narrows from §7's 40 walk-ins + trials + supply recruiting to a single bounded sprint:

12 qualifying discovery conversations, Westchester tow/collision/restoration/auto-glass, in the 30 days ending 2026-09-22.
~10-15 owner-hours total. This is essentially the entire owner budget for that month and is meant to be.
No price is quoted. No account is provisioned. No demo of a screen showing person names. A discovery conversation reaches no public surface, so Gate A does not block it — that is the ruling, and it is what makes the test runnable now instead of after 2026-09-13.
The one question is §7's: "how do you hear about these today?" Then stop talking.
Trials (the <4 half of the criterion) require #30, #31 and #43 first — see the ranking.

Owner: the owner personally. Date: 2026-09-22.
Cost: roughly a month of agent-output review. Agents will run further ahead of review than usual during September. Accepted.

5. Option (b) — switching the beachhead to a no-owner-hours channel now — is REJECTED, but #42's automatic fallback to segment #2 is SUSPENDED.

Rejected because changing channel before running the one cheap experiment throws away the only fact we would act on, and because inbound has 6+ months of latency with zero validation signal (BUSINESS_MODEL.md §9 Q8).

Amending minutes #42, decision on founding-agenda item 5: #42 provided that a failed beachhead re-ranks BUSINESS_MODEL.md §1 to segment #2 (private security, campus & hospital safety) with no further board sitting. That is now wrong on a fact #42 did not have. Segment #2 is more owner-hour-expensive per deal — scheduled meetings, written one-pagers, 90-day procurement. Auto-promoting it would replace a channel we cannot afford with one we can afford less.

New rule: a failed beachhead triggers a board sitting on CHANNEL, not on segment, and the standing presumption at that sitting is self-serve/inbound.
Owner: CEO. Trigger: the 2026-09-22 report.
Cost: one board sitting we had previously agreed to skip.

6. The comped friends-and-family tier is a cost centre with no strategic claim, and it is capped by NODES, not by accounts.

Q2 (customers first) plus Q1 (revenue) removes any strategic argument for the comped tier. It survives because the owner wants it, which is a legitimate reason and not a business one. That means its cost must be bounded now rather than after #45.

#50 is re-framed. #50 treats a comped account as the cost unit. That is wrong on BUSINESS_MODEL.md §4.2's own economics: the AI pipeline runs per call ingested, not per viewer. An extra comped viewer costs Firestore reads (pennies). An extra comped node costs $33-325/month. Ten friends watching node-002 is approximately free; one friend standing up a busy urban node is not.

Rulings, enactable without #45:

Comped accounts: uncapped. Marginal cost is near zero.
Comped nodes: hard cap of 3 fleet-wide until #45 ships per-org cost attribution. A fourth comped node requires a CEO decision, in an issue.
EMS/medical exclusion applies to comped nodes exactly as everywhere else (#43). No carve-out.
BUSINESS_MODEL.md §4.4's demand-gated pipeline is not applied to comped nodes — that would break the owner's "keep them full-featured" ruling on #42. The node cap is the substitute.

Owner: CTO to implement the cap and comment the reframing onto #50. Date: 2026-09-30, and before any fourth node is enrolled, whichever is first.
Cost: a friend who wants to run their own node in a new metro may have to wait. Under Q2 that is exactly the intended answer.

7. The 12 conversations are the single highest-value action in the company, above any code change. Agreed and recorded.

Q4 makes "nobody pays" the kill condition. The conversations are the only instrument that can return that verdict. Every engineering item in the backlog is a bet whose expected value is multiplied by the probability that anyone pays — a probability we currently have zero evidence about in either direction, and which costs ~12 hours to measure. Nothing else in the backlog has that ratio.

Corollary: owner-only items are rationed, not queued. During the test window the owner does the conversations plus the two security items in decision 8. Entity formation, ToS/Privacy/AUP, delay policy and takedown address are deliberately deferred until the test reports — they are prerequisites to charging, and there is nothing to charge for until someone says yes. Bootstrapping order, stated on purpose.
Owner: CEO. Recorded in GOALS.md.

8. Two owner-only items are NOT rationed, because they are security and not commerce.

#42 open items 9 (credential rotation state) and 10 (is the production Firestore ruleset the permissive one?). Combined they are well under an hour of console time and they gate whether we currently have a data exposure. They run before the conversations, not after.
Owner: the owner. Date: 2026-08-30.

Re-ranked backlog under these goals

The tiebreaker is GOALS.md's: does this move a paying customer closer, or does it stop the company from being killed? Everything else waits.

Tier 0 — owner, this month. The 12 conversations (decision 4). #42 open 9 and 10 (decision 8).

Tier 1 — agents, now. Makes a "yes" convertible into a trial.

Issue	Why it is here
#30	`/dashboard` does not exist — every post-login redirect 404s. You cannot put a prospect in front of a 404. Hard demo blocker
#31	`/admin` unreachable on cold load. Blocks the hand-onboarding path we would use for trial #1
#43	EMS exclusion + name suppression. Gate A A3 and Gate B B3, and the thing that makes a stranger's account safe to hand out
#13	Firestore rules deployed + ruleset ID recorded. Owner-blocked (WSL), pairs with decision 8
#4	`/admin` org-scoping. Gate B B2

Tier 2 — agents, after Tier 1. Gate B and the honesty surface.

#45 (per-call/org AI cost accounting — Gate B B5, and the only thing that unblocks a real #50 cap) · #44 (retention claims honest) · #46 (Gate A site copy, due 2026-09-13) · #48 (entity-name accuracy on 200 calls — owner-hour item, runs after the test reports) · #9 (org-owner writes) · #3 (billing mock data).

Tier 3 — agents, after the test reports and decision 3's proposal lands. The self-serve funnel: SAAS_PLAN.md B4 signup, plan-limit enforcement, zero-touch onboarding, org_api_keys. Not started before 2026-09-22.

Tier 4 — durable safety, unglamorous. #51 (no automated Firestore rules deploy). Real, and it becomes the top of this tier the moment #13 is done, because that is when the drift starts.

Tier 5 — deferred under Q2. Not killed, demoted. #5 (scene over-splitting) and #35 (LLM tier decides 3.5% of links) touch correlation, which BUSINESS_MODEL.md §3.4 calls the differentiator — they get promoted the moment a prospect complains about a specific bad incident, and not before. #28, #29, #33, #6, #27, #36, #37, #7, #19 sit below them. #6 (geocode_max_km rejecting MTA locations) jumps to Tier 1 only if the owner intends to show the map in a conversation — MTA coverage is §7's one non-obvious asset.

This is the concrete meaning of Q2. Correlation quality is the most interesting work in the repo and it is now behind a 404 fix, because the 404 stops a sale and the correlation bug does not.

What we gave up

The headline number. $2-5k MRR is now on the record as an aspiration with a stated gap. Nobody gets to quote it as a plan.
BUSINESS_MODEL.md §7 as a sequence, permanently. It was never schedulable at 5h/week and now says so.
A month of agent-output review. September's owner hours go to conversations.
The automatic segment #2 fallback from #42 — replaced by a board sitting we had agreed to skip.
Correlation quality, temporarily, and it is the thing most worth building. Q2 chose this.
Speed on the owner-only gates. Entity and legal pages wait for a signal. If a customer says yes on day 3, we will have a two-to-three week lag before we can charge. Accepted: that lag is cheaper than doing the paperwork for a customer who never appears.

Dissent

None recorded — this sitting ruled on owner answers rather than on adviser drafts, so no exec was overruled. The rulings that overturn prior board positions are decision 3 (against SAAS_PLAN.md §5's own ordering), decision 5 (amends minutes #42's founding-agenda item 5) and decision 6 (re-frames #50's cost unit against the CTO's and COO's original framing at #42). Any exec may file a correction; decision 6 in particular rests on an economic claim — cost is per node, not per account — that is falsifiable by #45 and should be checked when #45 lands.

Drafts

None. This sitting was called on the owner's direct answers to the CEO's goal-setting questions, not on a fan-out. The four advisers were not convened; the decisions above are sequencing and scope rulings on answers only the owner could give, and each names an owning exec for execution.

Enacted this session

GOALS.md written at the Version 5C root, stamped with this issue number and 2026-08-23. It is the document every future sitting measures against.
CLAUDE.md project-docs table gains a GOALS.md row.
.claude/skills/board/SKILL.md founding agenda amended: item 3 re-scoped by the time budget, item 5's automatic fallback struck per decision 5, item 2 unrationed per decision 8, and a new standing item 7 — "Rule on the scaling channel" added, triggered by the 2026-09-22 report.
#50 commented with decision 6's reframing.

Open — needs the owner

Run the 12 conversations by 2026-09-22. This is the company's highest-value action and nothing else on this list competes with it.
#42 open 9 and 10 by 2026-08-30 — credential rotation state, and whether the live Firestore ruleset is the permissive one. Console-only, under an hour.
Still open and now explicitly deferred until the test reports, per decision 7: #42 open 4 (business entity), 6 (superseded by decision 4), 7 (public-tier delay policy), 8 (takedown contact name + address).

## Agenda The owner answered the four goal-setting questions on 2026-08-23. Rule on them, resolve the contradiction between them, and produce a standing goals document every future sitting measures against. ## The four answers, verbatim | Q | Answer | |---|---| | **Q1 SUCCESS** | **C) Second income** — $2-5k MRR at 12 months; support and sales become a permanent weekly obligation and the roadmap follows customers, not curiosity. | | **Q2 WHO WINS** | **B) Customers first** — friends ride the same product paying users get; anything friends would like that doesn't sell gets deferred. | | **Q3 MY TIME** | **A) Under 5h, hobby** — agents build, owner reviews. | | **Q4 QUIT** | **A) Nobody pays** — 12 real conversations, zero conversions -> keep it running for friends, kill the commercial track. | ## The conflict, named Q1, Q3 and Q4 cannot all be true as written, and averaging them would be the worst available outcome. **The scarce resource is not engineering.** Agents build; agent-hours are effectively unbounded. The scarce resource is **owner presence in the things that cannot be delegated**: sales conversations, the business entity, ToS/Privacy/AUP sign-off, E&O, credential rotation, Firebase/GCP console actions, and support. The arithmetic, stated so nobody is surprised later: - $2-5k MRR at a blended ~$75/mo is **27-67 paying customers**. - `BUSINESS_MODEL.md` §7's own walk-in funnel is ~5 conversations per paid customer. 27 customers is ~135 conversations. At ~1h each with travel, prep and follow-up that is **~135 of the ~260 owner-hours the whole year contains**. - That leaves nothing for support. `BUSINESS_MODEL.md` §4.3 additionally requires selling the *same* county repeatedly, and Westchester saturates long before 27 accounts. - **The binding constraint is support, not sales.** 27 customers at 15 min/customer/**week** is 351h/yr — more than the entire budget. At 10 min/customer/**month** it is ~54h/yr and fits. That last line is the whole finding, and it is actionable: **at 5h/week, any customer who needs the owner personally is a customer we cannot afford.** Self-serve signup, self-serve billing and zero-touch onboarding stop being SaaS polish (`SAAS_PLAN.md` §5, "makes it good") and become **the revenue mechanism itself**. --- ## Decisions **1. $2-5k MRR at 12 months is ratified as an ASPIRATION, not a plan. The 12-month plan of record is 5-15 paying customers, roughly $400-1,200 MRR, conditional on the beachhead surviving its kill criterion.** I am not revising the owner's number down; I am refusing to pretend the current plan produces it. $2-5k MRR is a 24-36 month figure at this time budget **unless** decision 3 lands, in which case it is reachable sooner because customer count stops being a function of owner-hours. *Owner: CEO. Recorded in `GOALS.md`. Reviewed at the first sitting after the beachhead test returns.* *Cost:* we give up the comfort of a headline number that everyone half-believes. Every future sitting now has a plan-of-record it can be measured against and can fail against. **2. The under-5h/week budget is a HARD CONSTRAINT, not a preference. Any plan that requires more owner-hours than it has is rejected at the board, not attempted and missed.** Concretely: `BUSINESS_MODEL.md` §7's sequence — 40 walk-ins, trial hand-holding, node-operator recruiting and shipping B1-B2 concurrently — is **permanently un-schedulable** under this constraint and stays unratified. It is superseded by decision 4. *Owner: CEO, enforced at every sitting. Effective immediately.* *Cost:* the fast version of this business is off the table. We are trading calendar time for the owner's life, which is what Q3 asked for. **3. Self-serve is promoted from "makes it good" to a REVENUE BLOCKER. The scaling channel must not consume owner-hours per customer.** `SAAS_PLAN.md` §5 lists onboarding wizard, plan-limit enforcement, real `org_api_keys` and email as non-blocking. For decision 1's aspiration they are the blocker. This does **not** unsuspend Stripe (`SAAS_PLAN.md` §4 stays SUSPENDED until Gate B) and does not authorise publishing a price (Gate A stands). It changes **what agents build after the beachhead test returns**. *Owner: CTO, to sequence. Date: proposal due at the first sitting after the beachhead test reports.* *Cost:* correlation quality work (#5, #35, #28, #29) drops below funnel work in the queue. Under Q2 that is the correct trade and it is uncomfortable, because correlation is the differentiator. **4. The beachhead test is RE-SCOPED to fit the budget, and its clock starts now — before Gate A.** #42 open item 6 ("willing to walk into ~40 local businesses") is **answered YES by implication**: the owner chose a kill criterion that requires 12 conversations, and choosing a kill criterion is accepting the obligation to run the test. But the ask narrows from §7's 40 walk-ins + trials + supply recruiting to a single bounded sprint: - **12 qualifying discovery conversations, Westchester tow/collision/restoration/auto-glass, in the 30 days ending 2026-09-22.** - ~10-15 owner-hours total. This is **essentially the entire owner budget for that month** and is meant to be. - **No price is quoted. No account is provisioned. No demo of a screen showing person names.** A discovery conversation reaches no public surface, so **Gate A does not block it** — that is the ruling, and it is what makes the test runnable now instead of after 2026-09-13. - The one question is §7's: *"how do you hear about these today?"* Then stop talking. - Trials (the `<4` half of the criterion) require #30, #31 and #43 first — see the ranking. *Owner: the owner personally. Date: 2026-09-22.* *Cost:* roughly a month of agent-output review. Agents will run further ahead of review than usual during September. Accepted. **5. Option (b) — switching the beachhead to a no-owner-hours channel now — is REJECTED, but #42's automatic fallback to segment #2 is SUSPENDED.** Rejected because changing channel before running the one cheap experiment throws away the only fact we would act on, and because inbound has 6+ months of latency with zero validation signal (`BUSINESS_MODEL.md` §9 Q8). **Amending minutes #42, decision on founding-agenda item 5:** #42 provided that a failed beachhead re-ranks `BUSINESS_MODEL.md` §1 to segment #2 (private security, campus & hospital safety) **with no further board sitting**. That is now wrong on a fact #42 did not have. Segment #2 is *more* owner-hour-expensive per deal — scheduled meetings, written one-pagers, 90-day procurement. Auto-promoting it would replace a channel we cannot afford with one we can afford less. **New rule: a failed beachhead triggers a board sitting on CHANNEL, not on segment, and the standing presumption at that sitting is self-serve/inbound.** *Owner: CEO. Trigger: the 2026-09-22 report.* *Cost:* one board sitting we had previously agreed to skip. **6. The comped friends-and-family tier is a cost centre with no strategic claim, and it is capped by NODES, not by accounts.** Q2 (customers first) plus Q1 (revenue) removes any strategic argument for the comped tier. It survives because the owner wants it, which is a legitimate reason and not a business one. That means its cost must be bounded now rather than after #45. **#50 is re-framed.** #50 treats a comped *account* as the cost unit. That is wrong on `BUSINESS_MODEL.md` §4.2's own economics: **the AI pipeline runs per call ingested, not per viewer.** An extra comped viewer costs Firestore reads (pennies). An extra comped **node** costs $33-325/month. Ten friends watching node-002 is approximately free; one friend standing up a busy urban node is not. Rulings, enactable without #45: - **Comped accounts: uncapped.** Marginal cost is near zero. - **Comped nodes: hard cap of 3 fleet-wide** until #45 ships per-org cost attribution. A fourth comped node requires a CEO decision, in an issue. - EMS/medical exclusion applies to comped nodes exactly as everywhere else (#43). No carve-out. - `BUSINESS_MODEL.md` §4.4's demand-gated pipeline is **not** applied to comped nodes — that would break the owner's *"keep them full-featured"* ruling on #42. The node cap is the substitute. *Owner: CTO to implement the cap and comment the reframing onto #50. Date: 2026-09-30, and before any fourth node is enrolled, whichever is first.* *Cost:* a friend who wants to run their own node in a new metro may have to wait. Under Q2 that is exactly the intended answer. **7. The 12 conversations are the single highest-value action in the company, above any code change. Agreed and recorded.** Q4 makes "nobody pays" the kill condition. The conversations are the only instrument that can return that verdict. Every engineering item in the backlog is a bet whose expected value is multiplied by the probability that anyone pays — a probability we currently have **zero** evidence about in either direction, and which costs ~12 hours to measure. Nothing else in the backlog has that ratio. Corollary: **owner-only items are rationed, not queued.** During the test window the owner does the conversations plus the two security items in decision 8. Entity formation, ToS/Privacy/AUP, delay policy and takedown address are **deliberately deferred until the test reports** — they are prerequisites to *charging*, and there is nothing to charge for until someone says yes. Bootstrapping order, stated on purpose. *Owner: CEO. Recorded in `GOALS.md`.* **8. Two owner-only items are NOT rationed, because they are security and not commerce.** #42 open items 9 (credential rotation state) and 10 (is the production Firestore ruleset the permissive one?). Combined they are well under an hour of console time and they gate whether we currently have a data exposure. They run **before** the conversations, not after. *Owner: the owner. Date: 2026-08-30.* --- ## Re-ranked backlog under these goals The tiebreaker is `GOALS.md`'s: **does this move a paying customer closer, or does it stop the company from being killed?** Everything else waits. **Tier 0 — owner, this month.** The 12 conversations (decision 4). #42 open 9 and 10 (decision 8). **Tier 1 — agents, now. Makes a "yes" convertible into a trial.** | Issue | Why it is here | |---|---| | **#30** | `/dashboard` does not exist — every post-login redirect 404s. You cannot put a prospect in front of a 404. Hard demo blocker | | **#31** | `/admin` unreachable on cold load. Blocks the hand-onboarding path we would use for trial #1 | | **#43** | EMS exclusion + name suppression. Gate A A3 and Gate B B3, and the thing that makes a stranger's account safe to hand out | | **#13** | Firestore rules deployed + ruleset ID recorded. Owner-blocked (WSL), pairs with decision 8 | | **#4** | `/admin` org-scoping. Gate B B2 | **Tier 2 — agents, after Tier 1. Gate B and the honesty surface.** #45 (per-call/org AI cost accounting — Gate B B5, and the only thing that unblocks a real #50 cap) · #44 (retention claims honest) · #46 (Gate A site copy, due 2026-09-13) · #48 (entity-name accuracy on 200 calls — owner-hour item, runs **after** the test reports) · #9 (org-owner writes) · #3 (billing mock data). **Tier 3 — agents, after the test reports and decision 3's proposal lands.** The self-serve funnel: `SAAS_PLAN.md` B4 signup, plan-limit enforcement, zero-touch onboarding, `org_api_keys`. Not started before 2026-09-22. **Tier 4 — durable safety, unglamorous.** #51 (no automated Firestore rules deploy). Real, and it becomes the top of this tier the moment #13 is done, because that is when the drift starts. **Tier 5 — deferred under Q2. Not killed, demoted.** #5 (scene over-splitting) and #35 (LLM tier decides 3.5% of links) touch correlation, which `BUSINESS_MODEL.md` §3.4 calls the differentiator — they get promoted the moment a prospect complains about a specific bad incident, and not before. #28, #29, #33, #6, #27, #36, #37, #7, #19 sit below them. #6 (geocode_max_km rejecting MTA locations) jumps to Tier 1 **only if** the owner intends to show the map in a conversation — MTA coverage is §7's one non-obvious asset. This is the concrete meaning of Q2. Correlation quality is the most interesting work in the repo and it is now behind a 404 fix, because the 404 stops a sale and the correlation bug does not. --- ## What we gave up - **The headline number.** $2-5k MRR is now on the record as an aspiration with a stated gap. Nobody gets to quote it as a plan. - **`BUSINESS_MODEL.md` §7 as a sequence**, permanently. It was never schedulable at 5h/week and now says so. - **A month of agent-output review.** September's owner hours go to conversations. - **The automatic segment #2 fallback** from #42 — replaced by a board sitting we had agreed to skip. - **Correlation quality**, temporarily, and it is the thing most worth building. Q2 chose this. - **Speed on the owner-only gates.** Entity and legal pages wait for a signal. If a customer says yes on day 3, we will have a two-to-three week lag before we can charge. Accepted: that lag is cheaper than doing the paperwork for a customer who never appears. ## Dissent None recorded — this sitting ruled on owner answers rather than on adviser drafts, so no exec was overruled. The rulings that overturn prior board positions are decision 3 (against `SAAS_PLAN.md` §5's own ordering), decision 5 (amends minutes #42's founding-agenda item 5) and decision 6 (re-frames #50's cost unit against the CTO's and COO's original framing at #42). Any exec may file a correction; decision 6 in particular rests on an economic claim — cost is per node, not per account — that is falsifiable by #45 and should be checked when #45 lands. ## Drafts None. This sitting was called on the owner's direct answers to the CEO's goal-setting questions, not on a fan-out. The four advisers were not convened; the decisions above are sequencing and scope rulings on answers only the owner could give, and each names an owning exec for execution. ## Enacted this session - **`GOALS.md`** written at the Version 5C root, stamped with this issue number and 2026-08-23. It is the document every future sitting measures against. - **`CLAUDE.md`** project-docs table gains a `GOALS.md` row. - **`.claude/skills/board/SKILL.md`** founding agenda amended: item 3 re-scoped by the time budget, item 5's automatic fallback struck per decision 5, item 2 unrationed per decision 8, and a **new standing item 7 — "Rule on the scaling channel"** added, triggered by the 2026-09-22 report. - **#50** commented with decision 6's reframing. ## Open — needs the owner 1. **Run the 12 conversations by 2026-09-22.** This is the company's highest-value action and nothing else on this list competes with it. 2. **#42 open 9 and 10 by 2026-08-30** — credential rotation state, and whether the live Firestore ruleset is the permissive one. Console-only, under an hour. 3. Still open and now explicitly **deferred until the test reports**, per decision 7: #42 open 4 (business entity), 6 (superseded by decision 4), 7 (public-tier delay policy), 8 (takedown contact name + address).

logan added the board minutes:final role:ceo labels 2026-08-23 23:44:45 -04:00