What Pieter Levels Has Said Publicly About Building Solo With AI

The Short Answer

Pieter Levels (@levelsio) builds and runs his internet businesses on his own. His approach has been consistent for more than a decade: launch small products quickly, let paying customers decide what survives, and keep costs low. What AI has changed, by his own account in September 2026, is how far one person can take that — he writes that he now codes almost entirely with Claude Code on a server, has replaced paid software services with his own AI-coded tools, and runs his businesses at a profit margin of about 94.5%. He is also blunt about the downside for marketers: once AI can make a type of content as well as people can, that content stops getting attention.

Who Pieter Levels Is

On his own site, levels.io, Levels describes himself as an independent entrepreneur who builds internet startups by himself and has blogged since 2013. The products he lists include Nomads.com (a digital-nomad community), Remote OK (remote jobs), Hoodmaps, Photo AI, Interior AI and Hotelist. Photo AI's own site names him as founder and describes it as a fully owned, independent business without investors: users upload selfies, train a personal AI model and generate photos of themselves. (Both read 30 September 2026.)

On Shipping Fast

The habit predates AI. In "12 startups in 12 months" (1 March 2014) he set himself the challenge of launching a new product every month. His reasons: perfectionism stops people finishing, fear of how the market will react stops them launching, and unfinished projects are wasted potential.

He described the same method ten years later on the Lex Fridman Podcast, episode #440 (20 August 2024): he builds an idea as a small product and launches it within a couple of weeks, handling the design, code and logo himself. He also defended a deliberately plain technology stack — PHP, jQuery and SQLite — on the grounds that it keeps working and does not need constant change.

On AI Coding and Staying Lean

In "I passed $10M/y in revenue and investment gains with a 94.5% profit margin" (23 September 2026) he reports that his business revenue has been stable at about $200,000–$250,000 a month, that his investment returns now exceed it, and that the two together passed $10 million a year. The AI part is operational: he writes that he has coded almost solely on a remote server with Claude Code for close to a year, and that replacing paid software-as-a-service subscriptions with services he built himself saves him about $25,000 a month.

Note what the figure is and is not: it combines business revenue with investment gains, and he says most of those gains are unrealised. It is his own self-reported number, not an audited one.

On AI and Content: When Everyone Can Make It, Nobody Watches It

In "Every time AI can do something as well as humans it stops being special" (28 September 2026) he argues that AI pushes the cost of each media format — graphic design, then photography, then video — towards zero, and that once anyone can make a polished marketing video, feeds fill up with them and people stop paying attention. His conclusion for marketers is to stand out with work that is original, personal and, increasingly, rooted in the real world.

That is a notable position from someone who runs an AI photo business, and it is worth taking seriously if your plan is to win attention with AI-generated content alone.

On Experimenting in Public

He still treats new ideas as public experiments. In late August 2026 he wrote about building Infinite Slop, an interactive, AI-generated live stream, and followed it with a post saying 37,000 people had watched it before he registered a domain for it. The pattern is the same as in 2014: build it, show it, and let the response decide what happens next.

What to Take From This if You Are Building Alone

This part is ToppAgent's reading, not Levels's. Three practical tests fall out of his writing:

• Can you launch a first version within two weeks? If not, cut scope until you can. Tools such as Replit Agent exist for exactly that rough first version; the judgement about what to build stays yours.
• Which paid tools could you replace, and which should you keep? His SaaS savings came from building replacements; for most people the better first step is simply cancelling what they do not use.
• Would your content still get attention if everyone could make it? If the answer is no, add what AI cannot copy: your own data, experience and face.

For distribution, a scheduler with AI caption drafting such as Vista Social helps one person keep several social channels running, and PixVerse is an option for short AI-generated video — these are ToppAgent's picks for the task, not tools we attribute to Levels, and given his warning, use them to publish your own ideas faster rather than to replace them. For the wider toolkit, see the solo freelancer's AI stack and our Claude Code beginner's guide.

Sources: levels.io (site description and product list) and the posts "12 startups in 12 months" (1 March 2014), "I passed $10M/y in revenue and investment gains with a 94.5% profit margin" (23 September 2026), "Every time AI can do something as well as humans it stops being special" (28 September 2026) and "I built Infinite Slop" (August 2026); the Lex Fridman Podcast #440 episode page and transcript (20 August 2024); and photoai.com. All read on 30 September 2026. Paraphrased throughout; no direct quotations.

Related profiles: What Sam Altman Has Said About OpenAI's Future | What Dario Amodei Has Said About Safety-First AI | What Reid Hoffman Has Said About AI and Jobs | What Arvid Kahl Has Said About Bootstrapped AI | All Founder Perspectives