Cycle 2
Three products, one loop each. Every key result grades a step of the loop: understand, first win, return. Arrivals are watched, not graded, until the loop holds them.
Objectives and key results
Sim Startup
A stranger takes a company through its seed round, comes back the next day to see what happened, and sends the game to someone
- Outside players who start the real-time clock at the seed close, during the window. Target: At least 25, and at least half of the window's outside players
- Outside players who open the game on the calendar day after their clock start, on their own. Target: At least 12, and at least 3 in 5 of the players whose next day falls in the window (the continue bar of the playtest)
- New installs whose player says another person sent them the game. Target: 5
Public name: Default Alive (from 2026-10-01; Sim Startup is the internal name; naming decision). Added 2026-10-01 by Dave, mid-cycle (product decision). Dave: "the goal is to grow it. i don't care about monetization. i want it to catch FIRE." Loop: loop. The game is in public TestFlight beta (build 0.1 (7), link capped at 200 testers). Players arrive from community posts. This objective grades the loop's three steps plus a fourth that this product adds, spread: "catch fire" means players bring players, not more posts. The window is Oct 1 to Oct 9 (9 days). The grade at the close is Sim Startup's baseline for Cycle 3, not a verdict. Why people play is not yet known in their words, so this objective carries a pre-registered learning question, below. Work that serves these, in order: build 9, one TestFlight build carrying the session-done state, the founder's week and the G3 instrument (a one-tap "how did you hear about it?" and a way to pass the beta link on), for G1, G2 and G3; spec: the founder's week pitch. Then read the funnel daily and fix the step that loses the most players, for G1 (on Oct 1, every outside player who started the clock left within 4 minutes with nothing to do, which is the problem that the session-done state solves). Then the end calls with five players, for the learning question and the second half of the playtest criteria (a hard decision named with its trade-off). The name is decided before 1.0, outside the key results (the brand no-go does not apply here). Watched, not graded: TestFlight installs and the public link's fill (200 cap); companies that end in the garage, by cause; cards that expire unanswered; returns from a game notification; ending cards shared; how many testers mention the name. Not this cycle: the App Store 1.0 submission (the review queue cannot bring arrivals before Oct 9, and the name is open) and the Series A stage. Read G2 by build. Build 9 changes the step just before the return in the middle of the window, so G2 is read for build 7 and build 9 apart (by build number) as well as in total. The grade is the total. Pre-registered learning question, ruled on at the close: do people play to learn how startups work (the pitch's thesis), or for the tycoon fun? Evidence for learning: most sessions open the board deck or the numbers screen, and end-call answers name a lesson. Evidence for fun: sessions are mostly cards and office taps, and end-call answers name the art, the story or the people. Either answer is a result; the learning is written separately from the grade.
Pet Med Reminder
A new caregiver stops wondering whether they gave the pill, from the first dose on
- External installs during the cycle that log a first dose within 48 hours of install. Target: At least 6 installs, and at least half of the cycle's external installs
- People who cross the habitual bar (five doses on three distinct days in ninety) during the cycle. Target: 3 crossings, 13 to 16
- Frustrated repeat-taps on the Care Item Editor from cycle installs, after the editor fix ships. Target: Zero across the cycle's external installs after the fix is live
Loop: loop. Why they buy: to stop wondering whether they gave the dog her pill. The loop works for thirteen habitual people. It breaks at the Care Item Editor, in front of the first win. The first win is the first time the app answers the question: a dose logged means the next glance at the Today screen or the widget is an answer, not a guess. Annotated 2026-09-27: 9 of the 12 external habitual users had no dose in 14 days or more, and the five heaviest all stopped. They logged in rounds, not dose by dose (2026 09 27 heavy users why they stopped). The loop page now makes the round the unit, and builds 1.12.0 against it. Read P2's trailing count of 13 next to the weekly count of external dosers (4 in the week of Sept 21). Work that serves these, in order: submit 1.10.0 today; re-verdict and fix the Care Item Editor for P3 and P1; end onboarding on the Today screen with the next dose visible and the notification scheduled, so the buyer's question has an answer before they put the phone down, for P1; check whether the app shows a user their run of dosing days, and if not, put it on the Today screen, for P2; wire the rating prompt to the habitual-bar crossing, once per user, in the 1.10.x line after the editor fix, for P2. Watched, not graded: first downloads (instrument owed); the 1.9.0 listing conversion read; the App Store subtitle, which should carry the promise (never wonder whether you gave the pill) rather than the category, changed on or after Oct 10, when the window closes, never before.
Show Notes
A stranger feels caught up on a show they follow, for the first time
- Verified strangers who read a summary of a show they follow, out of the first 30 who see the samples screen in the cycle. Target: 6 of the first 30
- External people who read a summary on three distinct days in the cycle. Target: 3 people
Loop: loop. Why they buy: to feel caught up without thirty-five hours a week. No external person has completed the loop. It breaks at sign-in, before any first win. This is an activation cycle, and the targets are absolute counts because the baseline is zero. The why is the least certain of the three, so this objective also carries a pre-registered learning question, below. Work that serves these, in order: move the first win in front of the email: a stranger adds one show and reads one summary of it with no account, and the email is asked for afterward, to keep the feed (decided by Dave at adoption, 2026-09-26; this is the product the JTBD describes). The two sign-in failures are fixed on the way, since the email step still exists after the first win. Then make the first authenticated screen show this week's episodes of the shows they follow, summarized, by show search or OPML on first run, never an empty inbox (H3), because caught up is a feeling about a backlog and one summary of one show does not produce it; name and ship the key event; write Job B into the cycle doc as the job. Watched, not graded: arrivals from App Store search (12 external new people in 28 days); first-time downloads, read by hand (3 in the 28 days to Sept 25 from 20 page views); the email bar, superseded by S1. Annotated Sept 26: at 3 downloads a month, S1's denominator of 30 verified strangers does not arrive inside the cycle. The store page, not sign-in, is the first break; found the fix and Dave chose it the same day (variant 1b, subtitle "Caught up on your podcasts"); ships it as 1.5.0 with the first-win flag on and the caught-up state. S1 stands as written and is graded at the close on whatever arrived. Annotated 2026-09-28: Apple approved 1.5.0 and it went live on 2026-09-28 at 09:31 EDT (13:31:04Z, App Store lookup; App Store Connect reads ready for distribution). The S1 window opens at that time. S1's first 30 and S2's distinct days count from then. Before the first stranger arrives, one read must show that the S1 event reaches the app's analytics. Pre-registered learning question, ruled on at the close: does the loss of skipped episodes bother anyone? Evidence that it does: S1 met and S2 met, or any external reader answering the founder welcome in words that say so. Evidence that it does not: S1 met and S2 at zero, meaning strangers read once and do not come back for the feeling. Either answer is a result; the learning is written separately from the grade.
Paroletta
An arrival feels their Italian getting better, and comes back for the next three minutes
- External arrivals during the cycle that complete a first word the same day. Target: At least 15 arrivals, and at least 1 in 3 of the cycle's external arrivals
- External people who play on three distinct days in the cycle. Target: 3 people
Loop: loop. Why they buy: to feel their Italian getting better, three minutes a day. That makes the number on the result screen the product's promise, not decoration: I2 is the why made visible. Arrivals are the best of the three products. Completions were zero for four days. It breaks between the first open and the first completed word, at a step nothing has measured since Aug 12. Return evidence, five weeks to Sept 26: five external people completed the daily word on two or more days, two on three or more; nobody completed Categorie twice. Work that serves these, two tasks only, because Paroletta is third in priority: read the first-open-to-first-completion funnel, then fix the one step it names (closed the level picker on Aug 12; the loss since is unlocated). That is the cycle. The doors are decided (Sept 26: the loop is the daily word, Categorie gets no work, Vocabolario is the number). Carried, filed, not this cycle: the score and streak as the result screen's largest element with the name in the frame, and a number for every Vocabolario finisher. Either comes forward only if the funnel read names the result screen as the broken step. Watched, not graded: store downloads (instrument owed); Categorie and Vocabolario completion rates; the Play tester roster; the App Store subtitle, which should carry the promise (feel your Italian improve in three minutes a day) rather than the category; Paroletta has no ASO window running, so this can change with the next submission. Android is a no-go this cycle.
The retro
Cycle 1 proved a stranger can find each product. Cycle 2 asks what happens next: whether the stranger understands it in ten seconds, gets a first win, and comes back. Each product has one loop written on one page, under one sentence in the buyer’s own words about why they want it. Each key result grades a step of that loop, and the step is judged against that sentence: a first win is the first moment the buyer’s sentence comes true. Arrivals keep being counted, but no product is graded on more arrivals until its loop holds the ones it has.
Priority order: Sim Startup, then Pet Med Reminder, then Show Notes, then Paroletta. When a morning has one hour, it goes to Sim Startup. Amended 2026-10-01 by Dave: Sim Startup entered first, because its players come from community posts and that attention decays within days. Pet Med Reminder 1.12.0, already in progress, goes on. Show Notes and Paroletta finish the work they started and start nothing new for the rest of the cycle. From Sept 26 to Sept 30 the order was Pet Med Reminder, then Show Notes, then Paroletta. Personal projects (WCP Cloud, Filotto, Salotto, Lettura, Ascolto, Vine Trails, Auction Hub, the AI-native lane) carry no objective and are worked by whim. WCP Cloud’s registry experiment ends Oct 4 and is read as a personal-project result, not graded.
No-Gos
- Revenue. Growth now, monetization later still stands. No paywall, no pricing.
- New platforms. No Android work on Paroletta until the iOS loop holds. No Android for Pet Med Reminder.
- New features. Nothing that is not a step of a loop page. A feature that serves no step of the loop is the twenty-fifth feature. Amended 2026-09-27 by Dave, for Pet Med Reminder only: the loop page is rewritten around the round, and Pet Med Reminder 1.12.0 builds it inside the cycle. The P1 to P3 targets do not change.
- Brand. Names stay as they are. Brand is a question for the close. Amended 2026-10-01 by Dave: not for Sim Startup. Its name is live on no store, it has an open trademark question, and the first tester feedback asks to change it.
- Rating prompts. Pet Med Reminder only, on the habitual-bar crossing. Paroletta and Show Notes are deferred with dated reasons on their loop pages: Paroletta until the funnel has a baseline, Show Notes until a stranger has completed the loop. Review count is never a key result.
Get updates
Occasional notes on what's happening at Enginery. No spam, no marketing.