Skip to content

Timeline

The plan of record for the Impulse launch, mirrored from vault/docs/marketing/launch-plan.md §6–§7 and vault/docs/marketing/content-push.md §1–§4, §11.

Target ship: September 2027 (communicate October to backers) Kickstarter: 1 March 2027, 30 days Advertising budget: $2,000–5,000 total, Aug 2026 → Mar 2027 (BUD-ADS)

This file holds no status. It is the reference that Progress.md is measured against. /parselogs may edit it only when a log states a plan change as a decision — never to make slippage disappear. If the plan itself changes, say so in that day's log, so the change is visible rather than silent.

Month granularity comes from the source docs; the week-level splits come from the content playbook. Anything more precise than that is a window, not a commitment.

Plan change, 2026-08-21 — the campaign moves four months and the ship date six

Decided by the author and recorded in the 2026-08-21 log. School has started and the day job has ramped up; the hours the August–November plan assumed no longer exist. The dates move on purpose, which is a different thing from the three Week-2 milestones that were about to be missed by accident.

Was Now
Kickstarter launch early November 2026 1 March 2027
Campaign closes early December 2026 31 March 2027
Go/no-go gate check 2026-10-31 2027-02-14
Ship March 2027 (communicate April) September 2027 (communicate October)
Pre-launch run-up 12 weeks 27 weeks
Advertising spend unbudgeted $2–5k, phased (BUD-ADS)

The strategic reason, in the author's framing: market and finish the nRF redesign for six months, prove demand with the campaign, and only then start FCC certification — the expensive, irreversible, third-party-dependent part. The 2022 FCC rule change already permits advertising and conditional sales before authorization (launch-plan.md §5), so nothing about this ordering costs legality. It costs calendar, and buys the ability to not spend certification money on a product nobody has committed to.

Three consequences that are easy to miss:

  1. The old plan's fatal compression is gone. December close → March ship gave 3.5 months for a 6–10 week certification with 3–6 week failure penalties, plus pilot run, mass production and two app-store reviews, with zero slack. March close → September ship gives about five months. That is the first version of this schedule that is buildable.
  2. The failure mode changes from "not enough time" to "not enough sustained attention." launch-plan.md §12.32 already names founder burnout on the content cadence as a killer at twelve weeks; this plan runs twenty-seven, at reduced hours, through a school term. Every cadence rate below has been cut. A slower rate that gets hit beats a faster one that stops in month three — and the metric that decides this (emails per week, trending) does not care about post count.
  3. One regulatory item does not get to wait for the campaign. See SEP-13 and the note under Phase 1. The WM02C's grant condition problem is a negotiation with Fanstel, not a purchase, and it has no schedule you control.

ID changes

Every AUG-* identifier is unchanged and still means what it meant. The rest were renumbered to their new month. No number has been reused — where an old ID moved, the old number is retired, not reissued.

Old ID New ID Where it went
SEP-01 5 posts/week — folded into CAD-POST; it was always a rate, not a milestone
SEP-02 open $15 reservations DEC-01 1 Dec 2026
SEP-03 60 creator emails — folded into CAD-CREATOR
SEP-04 weekly Hackaday.io logs — folded into CAD-LOG
SEP-05 second YouTube long-form NOV-03 Nov 2026
SEP-06 first newsletter/podcast pitches NOV-04 Nov 2026
SEP-07 HN power post NOV-02 Nov 2026 — and it is better there, see Phase 2
SEP-08 Hackaday tip line #1 DEC-02 Dec 2026
OCT-01 do only what won FEB-01 Feb 2027
OCT-02 paid creator videos DEC-03 Dec 2026 – Jan 2027
OCT-03 third HN post JAN-01 Jan 2027
OCT-04 Hackaday tip line #2 JAN-02 Jan 2027
OCT-05 book the FCC lab slot MAR-05 mid-Mar 2027, during the campaign
OCT-06 pre-scan FEB-05 Feb 2027 — read the note, it is the one FCC item worth paying for early
OCT-07 final-week push FEB-03 Feb 2027
OCT-08 gate check (go/no-go) FEB-06 2027-02-14
NOV-01…NOV-04 campaign MAR-01…MAR-04 Mar 2027
Old Phase 5 (DEC-*, JAN-*, FEB-*, MAR-01) SHIP-01…SHIP-11 Apr–Sep 2027

New items added by this revision start at SEP-10 and OCT-10 so that no reader of an older document can land on the wrong milestone.

The gates

Two changes. The go/no-go moves to 2027-02-14 — two weeks before launch, so a no-go is still actionable rather than an autopsy. And a new early gate fires at the end of November, because a single go/no-go at the end of a twenty-seven-week runway is the wrong shape: it tells you in February what you needed to know in November.

Gate Target Date Why it is a gate
GATE-TRAJECTORY 600+ emails 2026-11-30 New. Not a launch decision — a hook decision. 600 by the end of November is the trajectory that reaches 2,500 by mid-February at a compounding rate. Miss it and the answer is to change the hook, not to post more.
GATE-EMAILS 2,500+ emails 2027-02-14 Campaigns taking 20% of funding in 48h have a 78% chance of hitting goal, and that velocity comes from your own list.
GATE-DEPOSITS 250+ paid reservations ($15, refundable) 2027-02-14 Deposit-holders convert to backers at 40–60% vs 3–5% cold. Still has no mechanism — see DEC-01.
GATE-CREATOR ≥1 creator video that visibly outperformed 2027-02-14 Evidence the framing converts before money goes behind it. DEC-03's paid placements are how this is bought.

If GATE-EMAILS, GATE-DEPOSITS and GATE-CREATOR are not met, do not launch. A failed campaign burns the one launch news cycle you get; six months of organic that produced nothing costs only time.

2,500 emails is now a lower bar than it was, and should be read that way. It was aggressive over twelve weeks. Over twenty-seven it is ~93/week — if that is not clearing comfortably by January, the problem is the hook or the channel, and the gate is doing its job by saying so.

Funding requirement ≈ $60–75k, needing 400–500 backers at ~$165 average pledge.

Phase 1 — 21 Aug → 30 Sep 2026 · Restart and instrument

Six weeks. The job of this phase is to finish the assets Phase 1 already had open, put a number on everything, and get the two Fanstel conversations moving.

Nothing here is new work. Every content item below was already due; they now have realistic windows instead of a week each.

ID Milestone Due Note
AUG-01 Film the Sunrise Lock master footage — #content ✅ 2026-08-11 Residual, and still open: 11 takes exist in one copy, on a camera. Getting them onto a disk is minutes and unblocks AUG-04 and AUG-06.
SEP-10 Install site analytics — #campaign 2026-08-31 New. Ten minutes. Every number in this file is unmeasurable until it exists, and AUG-09 has no instrument without it.
AUG-02 Site + email capture live — #campaign 2026-08-31 Live since 2026-08-09; §4's UTM acceptance test is what closes it.
AUG-05 Hackaday.io project page + first two build logs — #content 2026-09-07 Done when: the page exists and log #1 (what Impulse is, why the phone can't be the enforcer) states the commercial intent plainly.
SEP-11 Stand up the blog — #content 2026-09-07 New, and it was always an unstated dependency. AUG-07's done when requires submitting your own blog's URL; the site has eight routes and none is a blog. Every HN submission in BUD-HN needs this to exist first.
AUG-06 Begin posting — #content from 2026-08-24 Start of CAD-POST, at its revised rate of 2/week.
AUG-04 Founder video published — #content 2026-09-20 Once, done well.
AUG-03 Kickstarter pre-launch page live — #campaign 2026-09-30 Copy has been done since 2026-08-11; the page has never been started. Likeliest home for GATE-DEPOSITS.

Running in parallel — regulatory, and not deferred with the rest of certification

The campaign-funds-the-certification decision defers paying for certification. It does not defer the two items that are somebody else's schedule.

  • SEP-13 Open the WM02C change-of-ID conversation with Fanstel — #regulatory · due 2026-09-30 · New, and the longest-lead item in the entire plan. The X8WWM02C grant is a mobile grant: 20 cm from all persons. The watch breaks that condition, and no SAR arithmetic changes it because it is a clause on the grant rather than a consequence of a power figure (launch-plan.md §5, 2026-08-11). The fix is a Class II permissive change only Fanstel may file, or a change of ID in which Fanstel authorises you to hold your own FCC ID. That is a negotiation with a company that took days to answer a one-line question, and it has no schedule you control. Starting it in April 2027 alongside the lab booking is how a five-month buffer becomes zero.
  • SEP-12 Ask Fanstel to scope the open BM20C filing for portable body-worn use — #regulatory · due 2026-08-31, and the date is theirs, not yours Fanstel said on 2026-08-04 that BM20C FCC testing would complete "this month." Ten days of that month remain. Scoping an open filing costs them a paragraph; asking after it grants costs a permissive change. If they file it mobile-only the way they filed the WM02C, the BLE-only fallback stops being SAR-excluded by calculation and the contingency in BLK-15 loses its cheapest property. drafts/fanstel-exposure-category.md has been written and unsent since 2026-08-12. ⛔ 2026-09-20 — this window is shut, and the sentence above describes what happened. The grant issued 2026-09-11, mobile, with the 20 cm clause. The draft was never sent and the free question is now a Class II permissive change only Fanstel may file. The milestone is expired rather than overdue, and what replaces it belongs to SEP-13: one conversation now covering both modules. The date below is left where it is deliberately — this is slippage, not a plan change, and Progress.md should keep saying so.
  • AUG-12 Get three FCC lab quotes — #regulatory · due 2026-09-30 · scoped as one PCB / two host configurations / two exposure categories. 1 of 3 answered. Quotes cost nothing and age well; the quote is not the certification.
  • AUG-11 ✅ Part 15B integration guidance — discharged 2026-08-09; it was in the datasheets already filed.
  • AUG-13 ✅ SAR exclusion analysis — done 2026-08-09. BLE excluded (2.5 vs 3.0), 2.4 GHz WiFi not (26.2 at datasheet power; 53 against the grant's declared 0.17 W), 5 GHz 7.3.

Retired from Phase 1

  • AUG-07 First HN attempt — moved to October (Phase 2). It was four days away with no blog behind it, and BUD-HN allows 3–4 submissions across the whole plan. Spending the highest-variance card in the deck on an artefact that isn't written was the worst available use of it; six months of runway means there is now no reason to.
  • AUG-08 Reach out to 20 creators — folded into CAD-CREATOR, which starts at 10/week rather than 20.
  • AUG-10 Batch-film 20 clips — folded into CAD-FILM.
  • AUG-09 Review analytics; double down on the best hook — moved to 2026-10-31, because SEP-10 has to exist first for it to mean anything.

Phase 2 — 1 Oct → 30 Nov 2026 · Build in public, and actually build

Nine weeks. This is the phase the old plan never had: engineering time between the boards arriving and the campaign. It is also where the technical content comes from, because the two are the same work.

Hardware — the reason the extra time is worth having

  • AUG-14 Validate Board V1 — #hardware · due 2026-10-31 Done when: an assembled unit drives vibration and the LED ring simultaneously. ✅ This half is done — 2026-09-10, ring and motor driven together during a live enforcement window with a human watching. And the second thing, which is not in that sentence: three of five recent ESP32 prototypes died with the 3.3 V regulator holding the rail low, and V1 carries an identical power tree (analysis in docs/agent-notes.md, 2026-08-10). Whatever that failure is, it takes days to appear — a demonstration that runs once proves nothing. The old plan gave this three weeks. It now has ten, which is the first schedule in which "leave it powered for a fortnight and see" is actually possible. As of 2026-09-12 the soak has still not started, and it is now the whole of what AUG-14 has left — with one thing that changed in its favour: a second Board V1 is assembled and working (it runs the anchor role, associates on WiFi, and has been ranged against), so the soak board and the development board no longer have to be the same board. The reason the soak has not started is no longer a hardware constraint.
  • OCT-10 Draw and order the board respin — #hardware · due 2026-11-30 · New. V1 is explicitly a bring-up prototype, not the shipping revision (launch-plan.md §5, the author's own correction). The respin carries the WM02C antenna keepout violation found 2026-08-09 and whatever V1's power tree turns out to be. This is the board that gets certified, so it has to exist and be trusted before SHIP-03's pilot run in April 2027 — and launch-plan.md §12.6 is certifying the wrong board revision.

Content

  • AUG-07 First HN — Show HN (framing) — #outreach · October, Tue–Thu 8–10am ET · spends 1 of 4 Title: "Show HN: Impulse – a commitment device where the adversary is my future self" Done when: posted on your own blog's URL (SEP-11), and you were in the comments for four straight hours.
  • NOV-01 Build the November review into GATE-TRAJECTORY — #content · 2026-11-15 A fortnight before the gate, not on it. Pull the email numbers by source, name the hook that is producing them, and decide what November's last two weeks do differently — so that NOV-05 is a measurement of a decision rather than the first time anyone looked.
  • NOV-02 Second HN — the power post — #outreach · November · spends 1 of 4 Title: "Migrating a wearable from ESP32 to nRF54: what it cost and what it bought" This post is materially better under the new plan than the old one. It was scheduled for September, when no nRF firmware existed and no number had been measured; content-push.md §4 says in its own words that it "needs real measured numbers." Phase 2 is the phase that produces them. This is the strongest single piece of material in the plan and it now has its evidence.
  • NOV-03 Second YouTube long-form — #content
  • NOV-04 First newsletter and podcast pitches — #outreach
  • AUG-09 Review analytics; double down on the best hook — #content · 2026-10-31 · the first one with an instrument behind it
  • NOV-05 GATE-TRAJECTORY check — #campaign · 2026-11-30 · 600 emails. Not a launch decision; a hook decision.

Phase 3 — 1 Dec 2026 → 31 Jan 2027 · Convert

Nine weeks, of which about three are dead. 19 Dec – 3 Jan is deliberately light — creator outreach does not land, newsletters do not run, and attention is elsewhere. Plan around it rather than failing against it.

  • DEC-01 Open $15 refundable reservations — #campaign · 1 Dec 2026 ⚠️ This has no mechanism and hasn't since pre-orders were removed on 2026-08-09. It also acquired a new problem with the date change: a deposit taken 1 Dec 2026 against a September 2027 ship is nine to ten months of holding someone else's money, well outside the ~120-day chargeback window launch-plan.md §12.12 warns about, and long enough to attract rolling reserves. Whatever mechanism is chosen has to survive that; an authorization-and-later-capture arrangement is worth asking a processor about before a hold-the-cash one.
  • DEC-03 Paid creator placements — 5–8 — #outreach · Dec–Jan · BUD-ADS, $300–800 each Treat as data, not reach: you are buying the answer to which framing converts. This is also how GATE-CREATOR gets bought rather than hoped for.
  • DEC-02 Hackaday tip line pitch #1 — #outreach · tips@hackaday.com, four sentences, one photo, one link. Pitch the technique, not the product.
  • JAN-01 Third HN post — #outreach · January · spends 1 of 4. Pick one: RSSI is a liar (best candidate) / Fail-open / Safety limits in firmware because the app will be forked
  • JAN-02 Hackaday tip line pitch #2 — #outreach · different angle from December's
  • JAN-03 Newsletter and podcast placements run — #outreach · pitched in November, run in January, which is the highest-intent month of the year for this product's subject matter

Phase 4 — 1 Feb → 28 Feb 2027 · Concentrate and pre-launch

Four weeks. February is the old October: kill everything that didn't win, and point what's left at one date.

  • FEB-01 Do only what won — #content · kill the rest
  • FEB-02 Retargeting spend — #campaign · BUD-ADS, $500–1,500 · retargeting and pre-launch-page audiences only, never cold
  • FEB-03 Pre-launch page push + "we launch in 10 days" sequence to the list — #campaign
  • FEB-04 Campaign assets final — #campaign · video, page copy, tiers locked at six (BUD-TIERS)
  • FEB-05 Pre-scan, paid from deposit money — #regulatory · optional, and the one FCC item worth paying for before the campaign ⚠️ Read this before deferring it. The whole point of the new plan is that certification money follows demand, and a pre-scan is the exception that proves the rule: it is ~$1–2k against 250 deposits ($3,750), it is the cheapest possible instance of launch-plan.md §12.4 (don't spend customer money on certification that might fail), and a failure found in February changes the campaign, whereas the same failure found in June changes the ship date. The argument for deferring it to March is real — it is money spent before demand is proven — but it is the smallest such bet available and the one with the highest information value. Decide it explicitly; don't let it slide.
  • FEB-06 Gate check — #campaign · 2027-02-14 · evaluate GATE-EMAILS, GATE-DEPOSITS, GATE-CREATOR. This is the go/no-go, and it sits two weeks before launch so that "no" is still a decision rather than a post-mortem.

Phase 5 — 1 Mar → 31 Mar 2027 · Campaign

  • MAR-01 Kickstarter launches — #campaign · 1 March 2027 · goal $15,000; six tiers maximum
  • MAR-02 All traffic to Kickstarter for 30 days, no exceptions — #campaign · site pre-orders dilute the velocity that drives placement
  • MAR-03 Daily campaign-native posting — #content · funding milestones, stretch goals, backer questions
  • MAR-04 Press follow-ups to everyone who didn't reply in Nov–Jan — #outreach · a live campaign is a news hook a pre-launch isn't
  • MAR-05 Book the FCC lab slot — #regulatory · mid-March, once the campaign is visibly funding · binding constraint: 2–6 weeks lead time. Booking before the funds land is the one certification commitment worth making on partial evidence, because the lead time is otherwise added to the critical path in April.

Chinese New Year 2027 falls on 6 February, with the usual two-to-four week factory shutdown around it. Nothing in this plan needs a factory in February, which is one of the incidental reasons a March campaign is cheaper than a February one — but it does mean no manufacturing conversation started in Phase 4 will get an answer until March. Do the MOQ and tier-pricing work (BLK-10) in December or January, not February.

Phase 6 — Apr → Sep 2027 · Certify, manufacture, ship

Five months from funds landing to shipping. The old plan had 3.5 and no slack; this has genuine room for one certification failure, which launch-plan.md §5 prices at 3–6 weeks each.

  • SHIP-01 Campaign closes; funds land ~14 days later — #campaign · 31 Mar → mid-Apr
  • SHIP-02 Pay the lab — #regulatory · mid-Apr
  • SHIP-03 Pilot run of production-representative units on the shipping board revision — #hardware · Apr–May · labs can reject hand-soldered prototypes with different stackups. This is OCT-10's respin, built properly.
  • SHIP-04 Long-lead component orders — #hardware · Apr
  • SHIP-05 WM02C change-of-ID / permissive change concluded — #regulatory · by May · started in SEP-13 eight months earlier. If this is still open in May, the ship date is not September.
  • SHIP-06 Formal Subpart B + SAR testing — #regulatory · May–Jul · 6–10 weeks booking to signed report if everything passes; each failure adds 3–6 weeks
  • SHIP-07 Mass production in parallel — #hardware · Jun–Aug
  • SHIP-08 App store submission and review — #app · Jun–Jul · two platforms; the least visible dependency and the one most likely to slip (launch-plan.md §12.20)
  • SHIP-09 Authorization in hand; import, label, fulfil — #regulatory · Aug
  • SHIP-10 Physical creator seeding begins — #outreach · Aug · now legal; shipping units before authorization is prohibited
  • SHIP-11 Ship — #campaign · September 2027 · communicated to backers as October

Cadence commitments

Rates, not one-off events. /parselogs scores these over trailing weeks rather than marking them done.

Every rate below was cut on 2026-08-21, deliberately. The old rates were written for a twelve-week sprint by someone with the hours. This plan runs twenty-seven weeks through a school term and a busier job. A rate you hit for six months beats a rate you abandon in month three, and launch-plan.md §12.32 says exactly this about the old numbers. If these turn out to be too easy, raise them in January — that is a much better problem than the alternative.

ID Activity Target rate Period
CAD-POST Short-form posts 2/week Sep–Oct → 3/week Nov–Jan → 5/week Feb → daily in March ongoing
CAD-FILM Sunday batch-film 8–10 clips, weekly ongoing
CAD-LOG Hackaday.io build log 1 per fortnight Sep–Dec → 1/week from January from Sep
CAD-CREATOR Creator emails 10/week Sep–Nov → 20/week Dec–Feb (Tue–Thu) from Sep
CAD-REVIEW Friday numbers review — kill the worst hook, clone the best 1/week ongoing
CAD-OFF Saturday off. Actually off. 1/week ongoing

Sunday batch-filming is the single mechanic that keeps the plan alive past week five, and it is the one rate that was not cut in kind — only in volume, from 12–15 clips to 8–10. If it lapses for two consecutive weeks, that is the earliest reliable signal the plan is failing — flag it loudly.

Budgets (things you can spend too much of)

ID Budget Cap Why
BUD-HN Hacker News submissions 4 across the whole Aug 2026 – Mar 2027 period Show HN is once per project. Domain throttling is real; repetition reads as marketing and gets flagged, not downvoted. The cap did not rise with the runway, but the spacing improved a lot — four posts across seven months instead of four across four.
BUD-TIERS Kickstarter tiers 6 maximum Decision paralysis is a defining trait of the beachhead audience.
BUD-ADS Advertising spend $2,000–5,000 total New, 2026-08-21. See the phasing below.

BUD-ADS — how the $2–5k is phased

Window Spend On what
Aug – Nov 2026 $0–200 Tooling only — analytics, scheduling, a link tracker. No paid acquisition. There is no proven hook to amplify and no list to retarget; money spent here buys noise you will then mistake for data.
Dec 2026 – Jan 2027 $1,500–3,000 DEC-03 — 5–8 paid creator placements at $300–800. Bought as data on which framing converts, and doubles as the only reliable route to GATE-CREATOR.
Feb 2027 $500–1,500 FEB-02 — retargeting site visitors, the email list, and pre-launch page follows. Highest-converting spend available and the only one this budget can afford to make work.
Mar 2027 $0–500 held in reserve Spent only if the campaign is already funding. Advertising never rescues a stalled campaign; it amplifies a moving one.

One rule that decides most of this: no cold paid social, at any point, in this budget. At $2–5k you cannot buy a signal above noise from a cold audience — median electronics CPA is ~$49 (launch-plan.md §9) and you would be buying at most a hundred conversions across six months, with no attribution good enough to learn from. The two defensible uses of this money are creators (which produce assets you keep and evidence you need) and retargeting (an audience you already own). Everything else in the plan is organic on purpose.

And the Meta rule stands regardless of budget: ADHD language lives in organic content, site copy, email and creator videos. It never appears in paid creative, and never boost a creator post where they say "as someone with ADHD, this is for us." One rejected ad is recoverable; a restricted business account in March is not.

Blockers and open decisions

From launch-plan.md §13. These gate other work; several block ship rather than marketing.

Ship blockers

  • BLK-01 Spec the hard override into firmware (§11.2) — an unconditional stop that works on the watch alone and cannot be disabled. A device that locks and escalates with no unconditional stop is a physical safety defect. 2026-08-08: the requirement is unchanged; the mechanism is open. "Both buttons 10s" is not implementable on Board V1 (one user button, P0.05, plus reset) and is judged the wrong mechanic regardless — it puts a screwdriver between a distressed user and relief. Design work in Concepts/Safety.md. This is the blocker most likely to be quietly deferred and least safe to defer. 2026-08-21 — the new timeline gives this the time it needed and removes the last excuse. It was competing with a November campaign; it now has Phase 2 and Phase 3 and no competing deadline.
  • BLK-02 Decide and document the primitive bounds (§11.3) — max window duration, min gap between windows, daily cap, escalation cap
  • BLK-03 ✅ Board revision unsharing motor and LED ring — demonstration discharged 2026-09-10. Design satisfied 2026-08-07 (Board V1 ordered, ring P0.02 / motor P3.03); an assembled V1 has now driven both at once with a human watching, which was the stated condition, and launch-plan.md §10's claims fence moved with it. The blocker is not fully closed, because its own second clause is the power tree — see AUG-14. See AUG-14 — V1 inherits the power tree behind a 3-in-5 prototype failure rate, which is a second thing this blocker's demonstration has to survive.
  • BLK-15 The BM20C is now FCC-certified — and it granted mobile. Asked on 2026-08-04, Fanstel replied that FCC testing would complete "this month" and that they have no confirmed ship date for the nRF54LM20A. Gates AUG-12, MAR-05 and everything in Phase 6. Get the grant number in writing before booking a lab slot. 2026-09-20 — the grant issued on 2026-09-11, and it carries the same 20 cm condition as the WM02C's. X8WBM20C, original equipment, DTS, 2402–2480 MHz, ≈8.4 dBm conducted / 15.4 dBm EIRP, granted to Fanstel with the note "The antenna(s) used for this transmitter must be installed to provide a separation distance of at least 20 cm from all persons." The certification half of this blocker is discharged and the exposure-category half doubled: the watch now breaks a grant condition on both modules, and SEP-13/SHIP-05 are a two-module negotiation rather than a one-module one. The silicon ship date is untouched and still unconfirmed. Read from the public record, not from Fanstel — see docs/agent-notes.md, 2026-09-20, including why it went nine days unnoticed and the caveat on the source. *2026-09-22 — the grant note read verbatim; the co-location half is not a second violation, and the ship date has a vendor answer for the first time. The clause ends "except in accordance with FCC multi-transmitter product procedures"*, so the BM20C matches the WM02C: the 20 cm exposure category is the unqualified half on both modules, and co-location of the two radios on one board is permitted under both grants. The doubling recorded above is real but it is a doubling of one condition across two modules, not two conditions — say it that way to a lab. Separately, Fanstel's own page now lists ISED 4100A-BM20C and TELEC R201-260622 and states "Availability: Production 09/2026". That is a vendor availability line, not a confirmed ship date, and it does not discharge this blocker's remaining half — but it is the first plausible date in writing, and confirming a published date is a far easier ask than requesting one, so it belongs in SEP-13's conversation. Detail in docs/agent-notes.md, 2026-09-22. 2026-08-21 — the schedule pressure came off this and the negotiating position improved. An extra six months is six months for Fanstel to grant the BM20C, and — more importantly — six months in which a fallback is a decision rather than an emergency. The supply half was discharged 2026-08-10 (both modules shipped). What remains is the grant, plus the WM02C grant-condition problem now tracked as SEP-13 and SHIP-05, which is the half with a third party's calendar in it. 2026-08-12 — the fallback's exposure category is verified, and it is the good answer. Fanstel's nRF54L15 filing states *"According to §2.1093 this is a Portable device" and clears SAR by exclusion at 5 mm (2.82 vs 3.0). Two corrections travel with it: "with SAR"* means an exposure exhibit, not a test, and the exclusion holds only at a near-zero declared tune-up tolerance once antenna gain is counted. Worked through in Concepts/Module Alternatives.md §5. 2026-08-11 — the fallback is a contingency rather than a candidate. v1 ships with WiFi* (firmware_spec_v3_nrf.md §1.5, launch-plan.md §5), so a BC15C build is what happens to* this design if the grant fails. The lab quotes still carry the BLE-only delta because that is what prices the contingency.

Marketing blockers

  • BLK-04 Resolve the battery vs example-day contradiction. 2026-08-11 — half resolved. Figure is 8 hours minimum; the 9 pm window is deleted from impulse_overview.md's example day. What remains is arithmetic: 8 h from 6 am is flat at 2 pm, and the example day still enforces a gym window at 5:30 pm. Still blocks any copy pairing the two.
  • BLK-05 Regulatory/claims review with a lawyer before the first ad — promoted 2026-08-21: BUD-ADS means there will now genuinely be a first ad, in December, and this blocker is written as a precondition to it. It has a date now: before DEC-03.
  • BLK-06 Write the prepared response to the coercion narrative — before launch, not during

Decisions

  • BLK-07 $199 outright vs $149 + subscription
  • BLK-08 Main crowdfunding tier at $125 or $135 (plan recommends $135)
  • BLK-09 Confirm $15/$16 are BOM, not landed
  • BLK-10 MOQ tier pricing in writing from the PCB manufacturer — do this in Dec–Jan, not February; see the Chinese New Year note under Phase 5
  • BLK-11 Anchor BOM reduction pass — compounds across a 1–5 attach rate

Before code goes public

  • BLK-12 Draft SAFETY.md — turns removal of the safety bounds from a cleanup task into an argument someone must make in public
  • BLK-13 Trademark filing on "Impulse" — the lever for open-source safety bounds, since a license can't restrict use. $350; may be deferred to campaign money, but publishing code before filing spends the lever
  • BLK-14 Clinician hour — one psychologist working with eating disorders or OCD, asked: how would a patient of yours misuse this?

Metrics worth logging

If a day's log mentions any of these numbers, /parselogs will pick them up and trend them.

  • Emails captured (total, and by source — UTM everything)
  • Deposits placed, and deposit rate off the list
  • Per-clip: views, completion rate, profile clicks, link clicks
  • Creator emails sent → replies → placements
  • Kickstarter pre-launch page follows
  • Ad spend to date against BUD-ADS, and cost per email where it is attributable (new)

The one number that matters: emails per week, trending. GATE-TRAJECTORY exists to force that reading at the end of November. If it isn't compounding by then, the hook is wrong — change the hook, not the volume.

Expected conversion, for sanity-checking

Source View → email
TikTok / Reels 0.05–0.3%
YouTube (description link) 0.5–2%
HN front page 5–10% of ~10–40k visits
Hackaday feature 3–8%
Newsletter mention 2–5%
Retargeting (FEB-02) 1–4% of reached, and the only paid line that clears organic

500k TikTok views ≈ 500 emails. One HN front page ≈ 1,000–3,000. Emails come from the technical channels, not the social ones.

Honest range over 27 weeks rather than twelve: 2,500–6,000 emails, still with enormous variance. Flat outcome ≈ 800 (don't launch). One breakout ≈ 10,000+. The longer runway raises the floor and the median; it does not change the shape of the distribution, because that is set by whether one technical post lands.