Source note: This is a summary of Pieter Levels’ interview on YouTube, Arvid Kahl — The Bootstrapped Founder.
TL;DR:
“Indie hacking” isn’t dead; it matured into the default way to build. That raises the bar: better products, stronger distribution, and faster iteration. In AI, technical moats are thin, churn is high, and GPU/API costs squeeze margins—so the pragmatic play is sell first → do it manually → automate later, use a stack you ship fast with, let SEO + TikTok/creators pull demand, and wrap your infra so you can swap providers quickly. Surf the short-term waves while cultivating a Lindy asset for stability.
Pieter Levels: Indie hacking isn’t dead — it’s the new normal
“Indie hacking is dead.” Pieter Levels tweeted that once, and most people stopped at the first line. What he meant: indie hacking grew up. It moved from a scrappy subculture to the default path many founders take. When something becomes normal, competition thickens, expectations rise, and Big Tech starts copying features fast. The game shifts from clever hacks to disciplined product and distribution.
When “hacking” becomes the norm
Back in 2014–2016, indie hacking was still novel. By 2023+, it’s mainstream enough that larger teams watch indie makers for ideas. User expectations climbed with it: smooth UX, solid social proof, and real distribution. Winning as an indie isn’t about obscure tech tricks; it’s about learning, shipping, and iterating faster than everyone else.
The AI era: thin moats, hard retention, rising costs
Most AI startups sit on the same foundations: LLMs like GPT, image models like Stable Diffusion, and similar training/inference APIs. That makes moats very thin—a smart competitor can rebuild what you did in a few months. Retention is tough because many users just “try and leave.” Meanwhile, GPU and platform costs are volatile, which quickly compresses margins. Pieter recalls a provider bumping fine-tuning from $3 to $20 per run right when revenue spiked. The lesson is simple: don’t bet your model on prices going down; design so you survive even when they go up.
Pieter’s process: sell first, manual first, then automate
Pieter’s approach is deliberately unromantic. Launch a minimal page, take real payments, and do the work by hand before writing systems. For AvatarAI, early orders were fulfilled manually through the night. Only after the workflow and demand were clear did he automate the backbone within a week. This “concierge first” stance answers two critical questions at once: will anyone pay, and what steps actually create value.
Technology: pragmatic, Lindy, and anti–over-engineering
Pieter refuses to “worship” stacks. PHP/jQuery, modern JS—use whatever makes you ship and fix fast. Repeat work a number of times, then abstract; avoid early elegance that slows learning. That’s the Lindy spirit: tech that’s survived reality is often the most useful. Pragmatic doesn’t mean sloppy—keep code structured so others can take over and modules can be swapped without drama.
Distribution is the moat: SEO and creators beat press
Media logos are nice; customers come from SEO and creators. One creator video took MRR from $12k to $40–50k and kept it elevated. Package your problem–solution in the language of search (pages for money keywords) and the formats algorithms prefer (short, curiosity-driven clips). The best social proof is real metrics, real customer outcomes, and real workflows—not a glossy press kit.
De-risking platform dependencies
AI startups stack dependencies: GPUs, model vendors, APIs, infrastructure. Defense looks boring but saves companies: abstract your model/provider layer, write adapters so you can switch quickly, and multi-source GPU/compute. Track price and performance like business KPIs, not mere technical details. Guarding gross margin belongs right next to building features.
X/Twitter has new rules: long-form, sharp hooks, watch time
The algorithm now favors watch time and early engagement signals. That opens the door for long-form posts on X—but only if you lead with a tight 1–2 sentence hook that earns the first few seconds. Pieter still “says it straight,” but he formats for the platform. If you build in public, think like an editor: one message, many formats.
Cycles and balance: ride waves, plant Lindy assets
Economies and cultures move in cycles; AI is the current wave. Pieter leaned in—trading stability for speed and pressure. On the other side, Nomad List shows the value of Lindy assets: low cost, steady revenue, slow-and-sure improvements. The practical equilibrium is to surf waves for upside while growing something durable to anchor your time and mind.
A practical path for indie devs today
Start with a painful, wallet-attached problem you can observe in subreddits, TikTok, and niche communities. Don’t build a system—fake the door with a landing page, payment, and a form. Fulfill the first dozen orders by hand to learn the real workflow. When a common pattern emerges, automate the spine and keep a few steps manual to keep learning. In parallel, build search-intent pages and short demos (before/after, real outcomes). Always measure unit costs (GPU/API per task), gross margin, and churn/retention; adjust pricing and packaging to actual usage frequency. And set a work–rest cadence so you don’t burn out chasing model releases.
Conclusion
Indie hacking didn’t die—it graduated. Today, the edge comes from how fast you learn and ship, how well you distribute, how little you depend on any single platform, and how honestly you tell the product’s story. Sell first, do it manually, automate later, then optimize. Surf to go fast; plant to go far.