#The Pinnacle Gecko Protocol: Idea → Ship → Feedback in Minutes

A synthesized, opinionated protocol for compressing the full idea-to-validated-learning loop to the shortest possible window. Built from 100+ sources spanning YC, Sequoia, a16z, the founders of Stripe, Airbnb, Facebook, LinkedIn, Basecamp, and the modern indie-hacker generation.

#Part 0 — The Core Insight (what every source agrees on)

The single thing every authority converges on: the unit of competitive advantage is not features, code, or capital — it is the cycle time between a hypothesis and validated user behavior.

• Sam Altman: "If you can get 2% better every iteration cycle, your iteration cycle is every four hours rather than every four weeks, and you compound that over the course of a few years, you'll be in a very, very different place. Make it one of your top goals to build one of the fastest iterating companies the world has ever seen."
• Paul Graham: "I've seen a lot of startups die because they were too slow to release stuff, and none because they were too quick."
• Patrick Collison: "I place more value on decision speed. If you can make twice as many decisions at half the precision, that's often better."
• Reid Hoffman: "If you're not embarrassed by your first product release, you've launched too late."
• Guillermo Rauch (Vercel): "The most important metric for growth is Time to URL. We need our pit crew to not take 6 hours to change the tires; we need to get back on the road for the next lap immediately."
• Naval: "Ship something into live production every week — worst case, two weeks. If you just joined, ship something."
• Brian Chesky: "You can ship every hour of every day, but then package it and tell a story if you want." (Founder mode = setting the pace, not the strategy.)
• Kent Beck: small batch sizes, "tidy after," continuous integration.
• Eric Ries: the unit of progress is validated learning, not output.

The protocol below is engineered to make that cycle as close to minutes as physics and humans allow.

#Part 1 — The 4 Permanent Pre-Investments

These exist before you have an idea. They are why a 12-hour idea-to-revenue cycle is possible at all. Louis Pereira built AudioPen to $15K MRR in 12 hours, but only because of years of preloaded scaffolding. Marc Lou ships a startup in 31 hours because he uses the same Next.js boilerplate every time.

#1.1 The Pre-Loaded Stack (set up once, reuse forever)

Pick a stack you have shipped on at least three times. Never learn a new framework during a sprint (Obsidian Clad Labs, 72-hour playbook). Recommended 2026 default:

#1.2 The Push-Deploy-Verify Loop (the universal CI/CD)

Configure, day one, on every project:

• Every push to a branch → Vercel preview URL (~45–60s build).
• Every merge to main → production deploy with instant rollback.
• One smoke test asserting the "magic moment" path still works (Lean Pivot's "Lean Way" smoke test).
• A kill-switch flag wrapping the new feature (LaunchDarkly: ~200ms global rollback vs. minutes for a redeploy).

Without this, your minimum cycle time is hours. With it, it's the build + test time — typically 30–90 seconds. (DHH/Vercel/STEEPWORKS push-deploy workflow.)

#1.3 The Pre-Wired Distribution Surface (own this before you have a product)

Per Pieter Levels and Marc Lou: an audience built before the product is the difference between "ship and hear crickets" and "ship and have 20 paying users by midnight."

• A Twitter/X account where you already build in public.
• 3 Discord/Slack/subreddit memberships where your ICP lives — built up before you need them.
• An email list of 20–50 people who said "tell me when you ship."
• A Product Hunt account with a few launches under it (so the algo knows you).
• A relationship with 1–2 micro-newsletter writers in your space.

This is the part founders skip and then wonder why their 12-hour MVP got 3 visits.

#1.4 The Pre-Loaded AI Toolchain

• Claude Code or Cursor with a project-specific CLAUDE.md/.cursorrules already written.
• A scratch repo with your standard schema patterns, RLS policies, error handling.
• Your own prompt library: PRD-from-idea, Postgres-schema-from-PRD, landing-page-copy, demo-video-script.

Naval (2026): "I give Claude a two-line description. It builds me an app. It ships it to my app store. 30 seconds later, I have a working app on my phone."

#Part 2 — The Two Speeds

You actually want two loops, used at different stages, not one. Confusing them is the #1 reason founders waste weeks.

#Loop A — The Hypothesis Loop (target: minutes to hours)

Used before you commit to building anything. Goal: validated learning.

#Loop B — The Ship Loop (target: minutes per change)

Used after the hypothesis is alive. Goal: 100 micro-deploys per day, each invisible to the user but informed by real telemetry.

Most startups try to run Loop B before Loop A is settled. That's why they ship features no one wants. Eric Ries's entire framework rests on this distinction.

#Part 3 — The Hypothesis Loop (Loop A)

#Step 1 — The 60-Minute Hypothesis Spec

Force yourself to fit the entire idea on one page, in 60 minutes maximum (Sequoia Arc + Lean Pivot + Makers Page 5-hour sprint converge here):

Riskiest Assumption (RAT):  ___ (the one thing that, if false, kills the idea)
Target user (specific person, not "small businesses"):  ___
Hair-on-fire problem:  ___
Magic moment (the "aha" — must be reachable in <30 seconds):  ___
Success metric (one number, measurable in <24 hours):  ___
Cutoff list (everything explicitly NOT shipping):  ___
Distribution test (where the first 10 will come from):  ___
Price (always non-zero — Marc Lou, ZenVoice):  ___

If you can't write this in 60 minutes, the idea isn't sharp enough. Stop and re-form the idea, don't start building.

#Step 2 — The 30-Minute Demand Test (BEFORE writing code)

The fastest possible feedback on whether anyone cares. Pick one — measured in minutes:

Rule from Marc Lou & Andrey: Sell first. If you can't get one person to commit money or a calendar slot in 24 hours, the problem isn't real enough.

#Step 3 — Concierge / Wizard-of-Oz the Solution (hours to days, not weeks)

Per Kromatic's distinction:

• Concierge MVP (user knows you're a human): use when you don't know the right solution shape. Generative.
• Wizard of Oz (user thinks it's automated, you're behind the curtain): use when you know the shape and need to test the workflow.

Both ship in hours and produce real revenue. Wealthfront, DoorDash, and Airbnb all started as concierge/manual operations. Rule of thumb: if the AI/algorithm hasn't been built, do it by hand for the first 10 users — you'll learn what to actually build.

#Step 4 — The 12-Hour Build (only if Steps 2–3 show signal)

Apply the AudioPen / 72-hour / Vibe-Coding constraint:

• Hours 0–2: Landing page first (Obsidian Clad Labs rule). If you can't write the headline, the idea isn't clear.
• Hours 2–4: Schema + auth + the one core action. Skip everything else — no settings, no admin, no password reset, no email confirmation.
• Hours 4–8: The "magic moment" path — input → AI/processing step → output. Add a payment link from minute one (Louis Pereira: "I had to have a payment link in the app from day 1"). Ship ugly: "the first version of every product we ship looks rough."
• Hours 8–10: Deploy to production. Wire PostHog autocapture + session replay + one funnel event for the magic moment.
• Hours 10–12: Open it to your pre-loaded distribution surface — 10–20 users from the people who answered Step 2. Live tweet the build. This is what made AudioPen work.

By hour 12 you have: a live URL, real users, real session recordings, possibly real revenue.

#Step 5 — The 48-Hour Decision (kill / pivot / persevere)

Use your pre-defined success metric from Step 1. Don't move it after the fact (Eric Ries's #1 rule, also the segmentos/Darko Kolev 48-hour framework):

• Persevere: ≥3 paying users from ≥2 acquisition channels OR Sean Ellis "very disappointed" >40%.
• Pivot: users get the concept but stall in the flow → fix the friction, retest in 48 hours.
• Kill: zero conversion, zero engagement, no one finishes the magic moment → start a new RAT next week (Marc Lou's 21-products-in-2-years strategy, Pieter Levels's 12-startups-in-12-months).

The decision must be pre-committed before you see the data, or you will rationalize.

#Part 4 — The Ship Loop (Loop B): Minutes Per Change Once Live

Once you've crossed the "real users" threshold, switch to elite-DORA cadence. The 2025 DORA report data: top 16% of teams deploy multiple times per day; only 9% achieve <1 hour lead time. Match them.

#4.1 Trunk-Based, Always

• One branch (main). Short-lived branches for ≤2 days max.
• Every developer merges to trunk ≥1× per day (Google, Facebook, Etsy practice).
• All work-in-progress hidden behind a feature flag (LaunchDarkly's "decouple deployment from release" — 200ms global rollback).
• Etsy: 50+ deploys/day via Deployinator. Facebook: tens-to-hundreds of diffs every few hours, 100% from master since 2017.

#4.2 The Anatomy of a Single Change (target cycle time: 5–15 min)

1. Cursor/Claude Code generates the change against your pre-written CLAUDE.md rules. (Aipedia: ~45 min on tasks that used to take 4 hours.)
2. One smoke test asserts the magic-moment path.
3. Push → preview URL in ~45s → eyeball the Playwright screenshot diff (STEEPWORKS workflow).
4. Merge → production in ~60s, behind a flag set to internal-only.
5. Toggle flag to 1% → watch PostHog error rate / funnel for 10 min.
6. Toggle to 25% → 50% → 100%, with auto-rollback if error rate >X% (Argo Rollouts / LaunchDarkly Guarded Rollouts).
7. Kill flag once at 100%, schedule cleanup within 30 days (Knight Capital's $460M lesson).

#4.3 Real-Time Feedback Sources Wired From Day One

• PostHog session replay with rage-click + dead-click + JS error detection. Watch the first 5 sessions of every new feature personally — Brian Chesky-style "in the details."
• PostHog AI (or Claude Code via PostHog MCP): "summarize sessions where users dropped off at checkout." This collapses the measure phase from days to seconds.
• PostHog surveys triggered when a user hits the magic moment OR fails to: one question, one answer, in-app.
• A /feedback button shipped in version 0 (Marc Lou: "There's no password reset form, but there's a feedback button so I know what users need.").
• Direct DMs to first 10 users within 24 hours (Collison Installation: "Right then, give me your laptop.").

#4.4 The Mini-Demo Cadence (David Mack / SketchDeck → Brian Chesky's CEO reviews)

Every working day:

• Anyone with new work shares it on a screen / Loom / channel within ≤24 hours of starting.
• Standard format: "here's what I built in the last day, here's what's wrong with it, here's the data."
• This is how Airbnb went from 0 to 70 launches × 2/year cadence with no PMs (Brian Chesky on Lenny's Podcast).

#4.5 The Continuous Discovery Cadence (Teresa Torres)

Match feedback cycle to delivery cycle:

• ≥1 customer interview per week, every week (automate recruiting inside the product — 80% of teams do it this way).
• Every 3–4 interviews → update the opportunity solution tree.
• Every iteration → one assumption test, not a feature build.

#4.6 The Founder Mode Layer

Brian Chesky / Sam Altman / Paul Graham overlap exactly here:

• CEO is the chief product officer.
• CEO reviews the work, not the metrics dashboard.
• Decisions made in hours, not weeks (Patrick Collison: "think about making it in an hour, and getting it done in the next hour.").
• "Fast decisions are an asset and a capability in their own right."

#Part 5 — The Compressed Cycle Times: What's Actually Achievable

Honest benchmarks from the source data, not marketing copy:

The ceiling for "minutes from idea to user feedback" is a tweet (15 min) or a Stripe checkout link on a Carrd page (1–2 hours). The ceiling for "minutes from idea to live working product" is whatever Naval is doing right now with Claude — sub-30-minute, single-user app.

#Part 6 — The 12 Anti-Patterns to Refuse (collected from every source)

1. Learning a new framework during a sprint. Use what you've shipped on 3× before.
2. Building auth/billing/email from scratch. Boilerplate or die.
3. Big-bang launches with embargoed press. PG: "It's easy to see how little launches matter. How many launches do you remember?"
4. Branches longer than 2 days. Trunk-based or you don't ship daily.
5. Settings pages, admin panels, password reset flows in v0. Use SQL, not UI.
6. Free plans before validation. Marc Lou: "You need 10,000 users to make $3,000." Charge from minute one.
7. Subscription billing in v0. One-time payment is faster to validate.
8. Zero telemetry. Lean Pivot: "A build without telemetry is a wasted build."
9. Skipping Concierge/Wizard of Oz to "just build it." You'll build the wrong thing.
10. Letting flags linger after 100% rollout. Knight Capital: $460M loss from a stale flag in 45 minutes.
11. Treating launches as a single event. Kat Mañalac (YC): "Launch again and again."
12. Hiring before product-market fit. Sam Altman: "My first piece of advice about hiring is don't do it."

#Part 7 — The 30-Day Cadence Once Loop B Is Running

This is what differentiates the indie hackers and elite teams from everyone else:

• Every day: ≥1 deploy to production (DORA elite). One mini-demo. Watch ≥1 session replay.
• Every week: ≥1 customer interview. One micro-launch (Marc Lou: launch 7 free tools/year). Review your 3–5 leading indicators (activation, time-to-first-value, retention, conversion).
• Every 2 weeks: One pivot/persevere/kill review of the active hypothesis (Eric Ries innovation accounting cadence).
• Every 6 weeks: Shape Up cycle boundary (37signals) — what got shipped, what got scope-hammered, what's next. Two weeks cool-down before the next cycle.
• Every quarter: Review DORA-style metrics: deployment frequency, lead time, change failure rate, recovery time, rework rate.

#Part 8 — The Honest TL;DR

Idea to user feedback in minutes is real, but only if you've front-loaded the work.

The 12-hour, 31-hour, and 72-hour success stories all share the same prerequisites: a stack the founder has shipped on dozens of times, an audience built over months or years, a CI/CD pipeline that deploys in seconds, and the discipline to ship something embarrassing rather than something polished.

If you have those four things in place, the protocol collapses to:

1. Hour 0: Tweet the headline. Watch reactions for 60 minutes.
2. Hour 1: Post a Stripe link on a one-page Carrd. Drive your existing audience to it.
3. Hour 2–10: Build the magic-moment path with Claude Code on your boilerplate.
4. Hour 10: Push to production. PostHog wired. Flag-gated.
5. Hour 11: DM 20 ideal users with the live URL. Watch their session replays.
6. Hour 12: First payment, first feedback, first decision: persevere, pivot, or kill.

If you don't have the four prerequisites, your real first project is building the prerequisites — not the product. That's a 30–60 day investment that pays back forever after.

Reid Hoffman, again: "If you're not embarrassed by your first product release, you've launched too late." Make peace with embarrassment, and the cycle compresses to the speed of your typing.

#Sources surfaced (representative — 100+ total scanned)

Paul Graham (Do Things That Don't Scale, The Hardest Lessons for Startups to Learn); Sam Altman (Startup Playbook, Startup Advice, Super Successful Companies, Startup Advice Briefly); YC Library (YC's Essential Startup Advice, The Best Way to Launch Your Startup, How to Build an MVP, Order of Operations, The Art of Shipping Early and Often); a16z (Shipping is a Feature, Scaling Your Technical Org, What Working Means in the Era of AI Apps, Programming Your Culture); Sequoia (The Arc Product-Market Fit Framework, Terrifying Questions Framework, Evolution of a Product); Brian Chesky (Decoder interview, Lenny Rachitsky podcast, Rotman interview, multiple founder-mode pieces); DHH (Seven Shipping Principles, The One Person Framework, You can't get faster than No Build, Myth #1: Rails is hard to deploy); Kent Beck (Tidy First?, Batch Sizes, XP foundational paper); Patrick Collison (Fast, Farnam St interview, Stripe leadership profile); Naval Ravikant (Build a Team that Ships, A Return to Code, Product and Media are New Leverage); Reid Hoffman (If you're not embarrassed, Masters of Scale w/ Zuckerberg); Eric Ries (The Lean Startup, Innovation Accounting, Build-Measure-Learn); Marty Cagan + Teresa Torres (Continuous Discovery Habits, opportunity solution tree); 37signals (Shape Up, Six Week Cycle, Set Boundaries, Decide When to Stop); Marc Lou (How I Built 21 Products in 2 Years, Micro SaaS in 31 Hours, ShipFast docs); Pieter Levels (12 Startups in 12 Months, Indie Hackers profile); Louis Pereira (How I Built AudioPen, Half Day Build, Bootstrapped Founder transcript); Guillermo Rauch / Vercel (Iteration Velocity Playbook, Push-Deploy Workflow); Etsy (Quantum of Deployment, Deployinator, Divide and Concur); Facebook (Rapid Release at Massive Scale); Google trunk-based development; LaunchDarkly (Dark Launching, Releasing Features, Trunk-Based Development 101, How LaunchDarkly Works); Databricks SAFE; PostHog + Fullstory session replay docs; Bolt.new / Lovable / v0 comparisons; ShipFast / Marc Lou boilerplate; Cursor / Claude Code workflow guides; 2025 DORA Report + DORA capabilities pages; Concierge MVP / Wizard of Oz canonical pieces (Kromatic, Netguru, Shortform, learningloop); 48-hour validation sprints (segmentos, darkokolev, Stormy AI, Digital Applied); 7-day MVP playbooks (StarterPick, Makershot, Aizecs, Novara Labs, Amir Brooks, businessideasdb, makebolt); the 5-hour validation sprint (Makers Page); plus dozens of 2025–2026 indie-hacker, AI-builder, and engineering-velocity write-ups.