How to start indie hacking
Start by owning a small product, charging for it, and finding one buyer you can actually reach. A repo, a launch post, and a follower count are optional costumes. The record says most attempts stay small. That is the base rate, not a reason to skip the first charge.
Starting indie hacking is not a curriculum. It is the moment a person tries to get paid by a product they own, instead of only by an employer. That is the definition. The person can be solo or not, technical or not, nights-and-weekends or all-in. Courtland Allen’s about page still says the test is customer revenue, and that almost every extra requirement people add is optional.
What follows is what the public record shows people doing when it works, and what it shows more often when it does not. It is not a cohort. It will not check your homework.
The three things that are the start
You have a product. It can be ugly. It has to be something a stranger can touch: a page, a login, a directory, a download, a checkout. A deck is not a product. A bio that says “building” is not a product. The 2016 Indie Hackers launch was interviews with people who already had a business or a side project with a story, including money. The identity came after the attempt, not before.
You charge. A free tool can be a start if the plan to charge is real and soon. The word, as Allen used it, pointed at profitable online businesses. A price teaches you whether anyone wants it enough to pay. It also keeps you out of the hobby pile. How the charge is shaped — subscription, one-time, listing fee, ad — can wait until you know who is paying. The first charge should be collectible. Stripe is the usual pipe in this culture. It is not magic. A manual invoice has started real companies.
You can reach a buyer. This is the part the tools do not include. An audience is one path. A community you already belong to, a person you can email because you have a reason, a marketplace, a partner with a list, a query people still type: those are paths too. If you cannot name the buyer without saying “everyone on X,” you do not have a start. You have a costume. The costume is crowded.
Decide the plan before you decorate it. If a small company you own is the finish, say that. If you need a fund’s money to make the idea true, you are starting a venture startup and the indie label will confuse your obligations. Mixing them is how people raise accidentally, or how they refuse help they actually need.
What a cheap stack changed, and did not
In 2014 the admired start was Levels trying twelve products in a year, most of them dying, one spreadsheet becoming Nomad List. The cost of a failed attempt was months. In 2026 the cost of a failed attempt can be a week and a model bill. He put a figure on the new build cost in July 2026, about $20 a month, in a post about moats. That number is his, and it is about software cost, not about your groceries. Shipping fast is available. It does not pick the product.
ChatGPT from November 2022 and vibe coding from February 2025 mean a person who does not write much code can still put a screen online. Allen already allowed that person. The new risk is thinking the screen is the company. Users, in the complaint Rushab Jain posted on Indie Hackers in February 2026, will not pay for a sloppy MVP the way they might have when a sloppy MVP was scarce. Polish is cheaper. So is the clone next door. A start that is only a generated UI is a start that anyone can repeat tomorrow.
Boilerplates were the 2023 answer to setup time. ShipFast put login and payments in a repo you could buy. By 2026 Lou said that shortcut was less valuable. You do not need one to start. You need a way to take a card and a way to log a person in if the product requires it. How you get there is a weekend decision, not an identity.
Where first customers have actually come from
The record is anecdotes with dates, not a funnel chart.
Nomad List’s first demand was people who wanted a city list Levels was already maintaining for himself and for a scene he was in. The product was the scene’s object. Distribution was usefulness inside a group that existed. The million-dollar year he posted in 2019 was four years later. A start can be that slow and still be the example everyone cites.
Lou’s starts are often a launch in front of people who already follow the work, then a product for those people. When the product is not for them, he has said the audience does not save it. If you are starting without that audience, copying the launch cadence copies the empty half.
Postiz did not start in 2026, but the jump is a lesson for a start that already exists: be sitting where a new habit needs a tool. An API that an agent can call was a small change on top of a scheduler. The article that traveled was someone else’s. A new founder cannot plan on a stranger’s viral post. They can notice when a habit is looking for a default tool and try to be callable. That is a narrower piece of luck than “go viral.”
Tony Dinh’s letters are a start that became a living over many months of a product people opened often, TypingMind, after an earlier product died for platform reasons. The lesson for day one is unpleasant. You may be starting the attempt that does not work. The 36 startups are that fact as a career. Quitting a product is allowed. Quitting the attempt is a different decision. The record cannot tell you which one this month is.
Stripe Atlas’s benchmark belongs here as a ceiling on fantasy. Twenty percent of Atlas companies reach a first paying customer within 30 days of incorporation, up from 8% in 2020. Most do not, even in that population. A first customer is not $1,000 MRR. The cultural milestone comes after a handful of people stay. The benchmarks say the top of the pyramid is thin. Start anyway if you want the life. Do not start because a thread said the median person hits five figures.
What to ignore at the start
Ignore the bidding boards as a business plan. outbid.lol is the live spectacle of 2026, not a first product for someone who wants a customer next year. Paying to sit on one is advertising, maybe, and a hobby, often. Know which you are buying.
Ignore courses that promise the outcome. The shelf is real and mostly sells the method to other starters. Interviews and founder blogs are enough to see the shape. A course is optional apprenticeship. It does not put your checkout in front of a buyer.
Ignore a personal content calendar that replaces the product. A short public log can keep you honest. A performance of a company you have not charged for is how the saturated feed grows. Post after there is something to show. Show the failure if you can do it without handing over the only hard part.
Ignore other founders as your only peers if they are not your buyers. The communities are good for “does this charge flow make sense?” They are bad for “the market exists because fifty people said good luck.”
A start that matches 2026
Something small enough to ship while the cost is low. A price. A named buyer and a place you can stand where they already are. A week where you talk to that buyer more than you talk to the editor. A number you do not inflate. If it is zero, say zero and change the offer or the room. If someone pays, figure out why they stayed before you add a second feature from a competitor’s landing page.
That is the whole method the archives agree on. The disagreement is taste, niche, and how public to be. The year makes the method feel harder because copies are cheap and attention is for sale. It does not replace the method with a new one. People who are making a living from a product they own are still doing a version of those three things. The screenshots are what happened after, for a few of them, over a long time.
FAQ
How do you start indie hacking?
Own something a stranger can use, put a price on it, and get it in front of a person who has the problem. Courtland Allen’s test is customer revenue, or the intent to earn it, not a funding round and not a follower count. A job can pay the rent while that is small.
Do you need to know how to code?
Allen’s about page says it is fine if you cannot code. In 2026 coding agents make a thin product cheaper to draft. You still have to decide what it is, charge for it, and handle the mess when it breaks. Shipping speed is not the same as a business.
How much money do you need to start?
Often very little for the tools: a domain, a host, Stripe, and a model subscription. Levels has talked about a build stack around $20 a month, his figure, as a claim about cost, not as a budget for your life. Living expenses are separate. A cheap stack does not pay rent.
How long does it take to make money?
Stripe Atlas says a larger share of its companies reach a first paying customer within 30 days than in 2020. That is a first customer, in one population. Many products never become a living. Public success stories are usually years, not a launch weekend.
Should you start with an audience?
Only if the audience is the buyer. Otherwise start with the buyer’s room: a niche you are in, a search that still gets used, a marketplace, a partner. Building a personal brand for a year with nothing to sell is a different project.