How to launch a simple first version fast

1 The bottom line

Ship a bare-bones first version built around one core action, and do it in weeks, not months. Real users teach you more than polish, and waiting lets someone else take the market.

2 Key takeaways

• Build the first version around one core action. A long feature list hides what the product is for.
  • Zero began as a one-screen timer with no registration and no monetization. A basic subscription later reached about $5 million ARR within half a year Mike Maser, Zero22:05.
  • Rootd launched with only the red panic attack button and waited about a year before adding features Ania Wysocka, Rootd9:57.
• Cut your feature list hard, then cut it again. Teams plan far more work than the test needs.
  • Teams commonly attempt 100 times more work than is necessary to test their assumption Eric Ries, Long-Term Stock Exchange43:06.
  • Bolt reached $20M ARR without a mobile responsive view Eric Simons, Bolt30:30.
• Ship in weeks and let real users teach you. Theory cannot predict what people will actually do.
  • DoorDash built DashPass in three or four weeks with one pod of four or five engineers Jimmy Liu, DoorDash13:17.
  • Cal AI's bare-bones app reached $1.5M MRR about 8 months after launch Daniel Heintzman, Cal AI6:08.
• Charge from the start with a simple paywall. Early revenue buys momentum and frees you to work on the product.
  • Coconote crossed $100K ARR in 45 days and $2M in 5 months Brett Bauman and Zack Hargett, Coconote16:40.
  • BoldVoice launched with a standard $10 a month plan and a one-week trial Anada Lakra, BoldVoice14:50.
• Raise the bar when trust or safety is at stake. A half-baked launch can cost more than it saves.
  • Superhuman had only 10 to 15 users when it modified a font for readability, because email is mission-critical Rahul Vohra, Superhuman42:40.

just build the smallest thing you freaking can. It can be ugly as hell and just launch it

Koby Conrad, Sunflower

2.1 Choose the one action your product exists for

A first version should do one thing that users come back to repeat.

• Name the one action and build the first screen around it as Mike Maser, Zero22:05 did.
  • Get people to take one simple action a few times a week, and make it rewarding to watch.
  • The first thing Zero users see today is still a timer, running or waiting to be started.
  • Breathwrk's older home page put breathing exercises on top and personalized content beneath, based on goals picked in onboarding Steve P. Young, Breathwrk7:38.
• Find what your audience does every day and build around it Koby Conrad, Sunflower22:11.
  • Sunflower's most important feature is the sobriety timer, not the AI.
  • Version one had three features: a timer, Fiverr graphics and a simple journal. It grew to 10,000 paying subscribers and $1M annualized revenue.
• Ask what makes the product indispensable, not flashy or fun Ami Vora, WhatsApp5:19.
  • Get that part 100% right before adding more. A long roadmap leaves users unsure what the product is for.
  • WhatsApp spent its whole roadmap making sure calls and messages went through for free every time.
  • Sony left recording out of the Walkman so the buy decision stayed binary Rory Sutherland, Ogilvy1:07:05.
• Validate the core before you add anything Ania Wysocka, Rootd9:57.
  • Rootd collected reviews and feedback for about a year, which confirmed people needed the core tool before she spent on more.
  • Opal spent about two years in the dark after trying to fork Android into a full operating system, with no MVP and no testing Kenneth Schlenker, Opal4:58.

2.2 Cut the feature list

Most of what you plan to build is not needed to test whether the product works.

• Write every feature, cut the list in half, then in half again Eric Ries, Long-Term Stock Exchange43:06.
  • Build only what is left. The MVP curve is U-shaped, so landing near the optimum is fine.
  • Behance launched with groups, a tip exchange and more as hedges. Killing the tip exchange and then groups raised project publishing Scott Belsky, Adobe20:50.
  • After portfolio color controls were removed, complaints lasted 24 hours, then stopped, and people published more.
• Resist the big V1 and strip each feature to its core hypothesis Jackson Shuttleworth, Duolingo1:10:43.
  • Simpler V1s ship faster, design faster, get approved faster and show which parts matter.
  • The streak goal started as a single message and grew over many iterations.
• Solve a single problem, because every extra feature costs time and money Jaime, B7 Dev0:52.
  • Add features after you have revenue and user feedback.
  • Instagram and Facebook launched stripped down and added features over years.
  • Ship a one-year feature set, not the five-year vision, and let the market decide Vitaly Davydov, App Masters1:58.
• Unship anything built for debugging before launch Roman Ugarte, Grok Bot19:14.
  • Two weeks before launch, the product still showed exposed model thinking and stored memories.
  • The team trimmed what users must see and hid anything they should not have to think about.

2.3 Build it rough and build it in weeks

A deadline, a tiny team and borrowed shortcuts get the first version out the door.

• Scope work by time, not effort Daniel Lereya, monday.com52:10.
  • Pick a deadline such as three weeks, and cut scope to fit. Planning without one turns into a sport of what can go wrong.
  • Have something working after about 10% of the time budget Nan Yu, Linear9:34.
  • For a month-long feature, that means something to test internally after week one.
• Use a small team and a simple offer Jimmy Liu, DoorDash13:17.
  • The first DashPass was unlimited free delivery over $20, with the whole service fee removed.
  • It was hacky and expensive, but it tested one value proposition and gave enough signal.
  • Google's AI Mode began with 5 to 10 people building a blank page, and one rough moment of brilliance gave them conviction Robby Stein, Google45:00.
• Borrow infrastructure and fake the hard parts so you can test the idea sooner.
  • Suno started as a Discord bot, so there was no UI or auth to build, and users tested, gave feedback and told others in one place Mikey Shulman, Suno8:37.
  • Anchor used college interns to do the Apple Podcasts submissions behind a single distribute button Maya Prohovnik, Anchor31:14.
  • Mimic the AI with a Figma prototype instead of training a model to prove a market Marily Nika, Meta14:54.
  • OpenArt's first product crawled AI images and showed their prompts. It took under a week to build and hit number one on Hacker News Coco Mao, OpenArt8:16.
• Aim for an embarrassing first version Laura Schaffer, Twilio1:17:15.
  • If the first iteration is not embarrassing, you have gone too far. This helps teams celebrate unpolished work.
  • Build only the core use case plus a small settings screen with notifications and a review prompt, and go live fast Bechir Arfaoui, App Masters4:24.
  • Cal AI's first version had only a plus button to take a photo and show a meal's calories and macros, with no daily total Daniel Heintzman, Cal AI6:08.

2.4 Launch, charge and find your first users

Put a simple price on it and get it in front of people right away.

• Charge from the start and keep the paywall simple Anada Lakra, BoldVoice14:50.
  • A standard plan freed the team to build onboarding and the core product instead of running paywall experiments.
  • Coconote offered one free note and then required a trial, essentially a hard paywall Brett Bauman and Zack Hargett, Coconote16:40.
  • BoldVoice now leads with an annual plan of about $150 and a higher $200 tier with AI chat.
  • Users who want to pay deserve to, and revenue funds a better experience Shawn Gong, Tinder9:02.
• Seed your first users before or at launch.
  • Submit a bare-bones V1 to Apple so the app can take pre-orders, which download automatically at release Caleb Dean, Runafy8:32.
  • Start marketing in month two, even with a rough app, and let distribution drive learning Daniel Heintzman, Cal AI6:08.
  • Runafy got about 3,000 downloads on release from roughly 2 weeks of pre-order, and its 1,000-plus leaderboard filled within an hour.
  • Margins launched around December 20th with whatever it had so it would not miss the New Year's reading wave Paul Warren, Margins27:30.
• Treat a launch as a tool for your next 30 users, not a make-or-break event Maor Shlomo, Base4443:26.
  • His first Product Hunt launch was weak but brought about 15 users and his first paying customer.
  • The second launch, backed by an engaged community, topped the day, week and month.
• When you can rank for a keyword, ship minimal and fix monetization after downloads arrive Steve P. Young, App Masters16:06.
  • Then work on the paywall, onboarding and pricing, and add features once the app hits about 5K or 10K.
  • A basic remote app could be built over a weekend with AI. One earns about $3.2 million over the last year with very few features Steve P. Young, Universal TV remote app13:22.

2.5 Learn from real users and iterate

The first version is a best guess, and real use tells you what to build next.

• Let users, not theory, decide Tara Seshan and Nan Yu, OpenAI1:27.
  • Nothing compares to watching users try something and then iterating.
  • Define the single most important question, get something in front of users fast, and feed the results into a sharper hypothesis Tara Seshan, OpenAI7:17.
  • ChatGPT shipped before the holidays as a learning exercise, expecting to wind it down, and people retained instead Nick Turley, OpenAI16:25.
• Release in stages and label early work as early.
  • Brand features as research previews so users know they may not last, which lowers the commitment of shipping Cat Wu, Anthropic8:08.
  • Put the rough feature into production for your own team first, then one to ten opt-in customers, and polish only before general release Karri Saarinen, Linear9:13.
  • Feature timelines on Cat Wu's team dropped from 6 months to 1 month, and sometimes to 1 week or even 1 day.
  • Google tested the tabbed inbox on the team's own inboxes, then other Googlers, then external testers Itamar Gilad, Google53:19.
• Validate one stage at a time and perfect only the part you are testing Nikita Bier, TBH and Gas1:12:07.
  • Gas first made polling and notifications perfect, then sharing and virality, then hopping between schools.
  • Keep it to about four things that must be true, since each extra layer adds risk.
• Let feedback reshape your foundation and expect three versions.
  • Cursor launched in about three months with a hand-rolled editor, then switched to a VS Code base after the rush of feedback Michael Truell, Cursor26:06.
  • Make the product, fix the product, then fix the business Tony Fadell, Build Collective31:07.
  • The iPod only took off in its third generation, with Windows support and the iTunes music store.
  • dbt's first scheduler was a naive for loop, and it has been rebuilt several times as scale arrived. It now manages 10 million runs a month Julia Schottenstein, dbt Labs53:20.

2.6 Where they disagree

Founders split on how much to build first

	Build one thing	Build many, trim
Advocate	Max Schoening, Notion	Koby Conrad, Sunflower
Approach	Smallest core, push rest back	Ship several, watch data
Why	One exceptional core wins	AI makes building cheap
Evidence	iPhone, GitHub, Dropbox cores	AI sponsor usage surprised him
Risk	May miss a surprise hit	Diluted focus

How rough the first version can be depends on your market, your users and what a mistake costs.

• Minimum viable or minimum lovable?
  • In a crowded market, do five things really well instead of 15 clunkily Jiaona Zhang, Webflow20:45.
  • Lovable says differentiation now comes from emotional feel and delight, so it will not ship anything that is not lovable Elena Verna, Lovable43:50.
  • Bolt shipped without basics like a mobile responsive view and still reached $20M ARR Eric Simons, Bolt30:30.
  • Strong distribution can make good enough good enough, because you can fix everything else later Steve P. Young, Universal TV remote app13:22.
• Mission-critical products need a higher floor.
  • Email is not something you can launch half-baked, since a Mailbox bug double-sent a founder's follow-ups Rahul Vohra, Superhuman42:40.
  • Set the non-negotiable quality floor first, then iterate above it Shweta Shrivastava, Waymo10:42.
  • In a network-effect marketplace race, speed matters above all Rahul Vohra, Superhuman42:40.
• Sometimes the first version must be big.
  • When you try something completely new, a small MVP can give a false negative. A new UI without the algorithms cannot tell you whether the idea or the machine learning was poor Gustav Söderström, Spotify57:42.
• You cannot have quality, features and a deadline all at once.
  • Pick two. Software can be iterated, so you can usually ship features and a deadline and improve quality later Dylan Field, Figma29:37.
  • Figma took three and a half years to launch, which Dylan calls way too long, but he says to make it a little awesome Dylan Field, Figma43:50.