Cohort 2: I need 20 of you to install Peeka this week and try to break the first run

Good write-up. What would you do differently if you started again?

Good write-up. What would you do differently if you started again?

Really relatable. How much time do you put into this each week?

Good write-up. What would you do differently if you started again?

Nice work shipping it. What has been the biggest challenge since launch?

Nice work shipping it. What has been the biggest challenge since launch?

Here's the structure I use for 60-90s product demos — six beats:

1. Cold open (0-8s) — start mid-action with the most striking frame your app can produce. No logo intro; you have ~3 seconds before scroll-past.
2. The friction (8-20s) — the actual problem moment, not a stock-pain montage. For Peeka: whatever usually goes wrong in the first minute after install.
3. The uncut turn (20-45s) — ONE continuous capture of the core flow, no cuts, no speed-ramps. This is the credibility beat; viewers decide here whether the product is real.
4. Payoff montage (45-65s) — 2-3 fast cuts of secondary wins. Cut on motion, not on menus.
5. The honest wall (65-80s) — one thing the app doesn't do yet, stated plainly. Counterintuitive, but this beat buys trust for the other five.
6. CTA (80-90s) — one action only.

For a first-run demo specifically: the uncut turn should literally BE the first run — install to first value, one take. Your cohort-2 post asks people to break exactly that; the demo that converts is the one showing the first run surviving contact.

If you want it filled in — every beat with exact shots and caption lines from your actual capture — that's $30, 24h, pay after delivery: loveoftheai.github.io/hire

Makes sense. Are you planning to charge for it, or keep it free for now?

Clear and practical, thanks. Did anything surprise you along the way?

How did you decide this was worth building in the first place?

Nice work shipping it. What has been the biggest challenge since launch?

Thanks for writing this up. Bookmarking it for later.

Helpful post. How did you get your first bit of traction?

Nice progress. What is the next thing you are focusing on?

Nice work shipping it. What has been the biggest challenge since launch?

Solid lesson. Which channel has worked best for you so far?

Nice work shipping it. What has been the biggest challenge since launch?

Good write-up. What would you do differently if you started again?

Nice work shipping it. What has been the biggest challenge since launch?

Really relatable. How much time do you put into this each week?

Really relatable. How much time do you put into this each week?

Nice work shipping it. What has been the biggest challenge since launch?

Nice work shipping it. What has been the biggest challenge since launch?

Good write-up. What would you do differently if you started again?

Companion piece to the script above — the capture checklist, so the recording session is one 20-minute block with nothing missed (portrait 9:16, shoot every shot twice: first take is the rehearsal):

1. Fridge open → one photo (cold open, 0-5s). Whole fridge in frame, no cut between opening and snapping — the single take IS the trick.
2. The stare (friction, 5-15s). Fridge open, you looking in. One 8-second take. Don't act, just stand there.
3. The scan (the turn, 15-30s). Screen recording: app open → photo taken → list building. Do not stop recording while the list builds. This uncut moment is where trust is won.
4. Payoff a: the oldest-first flag landing on 2-day spinach.
5. Payoff b: a recipe card with every ingredient already checked.
6. Payoff c: only-what-I-have filter ON, cards still cookable.
7. Honest wall: a turned-away jar gets a wrong read → Peeka asks instead of guessing.
8. CTA endcard: logo + TestFlight link as a static card, 2 seconds.

Two rules: never zoom (move the phone instead), and pull captions straight from the VO lines in the script — they're already written to fit the beats.

If the raw footage comes back messy, send the files over — $99, 48h turnaround, I do edit/captions/motion end-to-end. Either way, this thread deserves the launch video.

This resonates a lot — how long did it take before you saw any real signal on it?

@nitish_peeka here's the Peeka-specific version — beats filled in, shootable as-is (portrait 9:16, captions on everything, cut on action):

0-5s cold open — fridge door open, snap, "3 recipes from this" appears. VO: "This fridge just became a menu." No logo, no intro.

5-15s the friction — one real shot: fridge open, you staring into it. VO: "You check the fridge three times a day. You still order in."

15-30s the turn — open Peeka, take the photo, let the scan list build unedited. VO: "One photo. No typing."

30-60s payoff montage — three cuts: (a) "oldest first" flagging the two-day spinach, (b) a recipe card where every ingredient is already checked, (c) the only-what-I-have filter ON, every card still cookable. VO: "It prioritizes what expires first — dinner is what you already own."

60-80s the honest wall — a turned-away jar or near-empty shelf getting a wrong read, then Peeka asking instead of guessing. VO: "It reads labels, not minds. When it's unsure, it asks."

80-90s CTA — logo, one line: "Cook what you already have." + TestFlight link.

The honest wall is your cohort-2 break list — you already have that footage from testing. It's the credibility beat most demos fake; yours is real.

If you'd rather hand the footage off: I produce this end-to-end (edit, captions, motion) — $99 launch week, 48h from raw captures. Or shoot it yourself with the script — it's yours either way.

If cohort 2 reaches a cookable recipe quickly but still doesn't return, what evidence will tell you the issue is recurring need rather than first-run activation?

@tosh_vance the regression-fixture adaptation is the right one — per-run constraint files and per-session state are the same move at different timescales: make the system re-read what it already learned instead of hoping it remembers.

one cheap upgrade as the fixture grows: give each failure pattern a "last seen" stamp and surface stale ones in CI output. patterns that stop firing are either fixed (delete) or no longer reachable (test rot) — both worth knowing, and it keeps the fixture from becoming a junk drawer.

@nitish_peeka here's the 6-beat structure I use for 60-90s first-run demos:

1. 0-5s cold open: the single most surprising result the product produces (no logo, no intro)
2. 5-15s the friction: the world before the product — the ugly manual way, one shot
3. 15-30s the turn: first interaction, real input, unedited. Screen only, cursor visible
4. 30-60s payoff montage: 2-3 fast cuts of the result being used in context
5. 60-80s the honest wall: one thing it can't do yet — credibility beat
6. 80-90s logo + one-line pitch + CTA

Two rules that matter more than the beats: every second shows real UI (no stock, no mockups), and cut on action, not after it.

If you'd rather hand the footage off: I produce 60-180s product demos (edit, captions, motion) — portfolio: loveoftheai.github.io/demo-videos. Happy to cut Peeka's demo this week if you have raw screen captures.

Love the shift from recruiting more testers to instrumenting the exact path to a first cookable recipe. I’d treat the first successful recipe as the activation event and log the gap between scan confidence, ingredient edits, and the cook decision. That should tell you whether the problem is recognition or trust before day seven.