Vibe Coding: Levels' Hotelist Takes On Airbnb's Rating Inflation

A rating scale that stopped meaning anything

Anyone who has booked a stay on Airbnb knows the problem. Nearly every listing lands somewhere between 4.5 and 5.0 stars, so the number no longer separates a great place from a mediocre one. Pieter Levels, the indie developer behind a long run of solo-built web products, has built hotelist.com specifically to address what he calls Airbnb's 4.5-to-5.0 rating scale 4.

The premise is simple. A rating system only works if scores spread out, with most properties clustering in the middle and fewer at the extremes. When almost everything sits in the top tenth of the range, the scale compresses and guests are left reading reviews line by line to find the signal the stars should have provided. Hotelist's stated aim is to restore that spread 4. The real test is whether the distribution stays wide as more properties and reviews come in. Rating inflation tends to creep back once hosts and guests start treating a low score as a punishment rather than information.

Part of a rapid-fire building streak

Hotelist is not a one-off. Levels' recent output reads like a showcase of AI-assisted development. Around the same time, he built Infinite Slop, an endless, interactive, AI-generated live stream, and registered infiniteslop.ai after 37,000 people watched it 4. He says he made that site entirely on his phone, using Termius to connect to a Hetzner VPS running Claude Code, while sitting in a sauna 4. He also describes vibe coding a full video editor into his Photo AI product so users can generate and edit videos with their own trained models 4. Separately, he added US Census median income data to Hoodmaps after finding that Strava's data was protected 4.

The pattern matters more than any single project. Products that once took weeks of focused engineering now ship in days, sometimes from a phone. Hotelist fits that template: a narrow, clearly stated problem, attacked quickly by one person with AI tooling.

SuperLevels and the trust argument for building your own

The clearest example of Levels' philosophy is SuperLevels, a Chrome extension he vibe coded and open sourced on GitHub 1. His reasoning is about security. He argues that popular browser extensions are increasingly being bought by spyware and malware operators who sell user data or hijack accounts, and he calls this one of the top security risks right now 1. His answer was to rebuild the features of the extensions he relied on into a single tool he wrote himself and can read line by line. One example is a custom dim mode for X with seven color palettes 1.

He frames the open-source release as a template for others. Users can have an AI audit the code before installing it, then use AI again to customize it 1. That is a notable change in how people might relate to software. Instead of trusting a vendor, you trust code you, or your AI assistant, have inspected.

The data plumbing underneath

Projects like Hotelist depend on data, and the tooling market for collecting web data is crowded and price-sensitive. Comparisons of ScrapingBee alternatives show how much costs vary. FlyByAPIs offers a free tier of 100 requests a month and paid plans from $14.99 for 100,000 credits 2. ScraperAPI is pitched as the closest drop-in replacement, but it uses credit multipliers. A basic page costs one credit, while Amazon costs 5, Google search results 25, and LinkedIn 30, with JavaScript rendering adding 10 more 2. Scrappey prices at €0.20 per 1,000 plain HTTP requests and €1.00 per 1,000 full-browser requests with residential proxies included, and claims ScrapingBee runs roughly €25 for the comparable browser tier 3. Both comparisons come from vendors ranking themselves first, so their figures are best read as marketing claims rather than neutral benchmarks.

Nothing indicates which data sources Hotelist uses. Still, the economics are relevant. When infrastructure is cheap and AI writes much of the code, the barrier to launching a review or listings site drops sharply.

What it adds up to

Levels himself has been blunt about the downside of this shift. He has suggested indie hackers may be the first kind of developer to go extinct as AI lowers the cost of execution 4. He has also said big AI companies are absorbing much of what used to be indie app territory 4. He notes that tools like Wispr Flow, Granola and WHOOP have already been reverse engineered into free open-source versions 4.

That tension frames Hotelist well. Building a better rating scale is now easy. Getting enough honest ratings for the scale to mean something is not. Airbnb's compressed scores are a product of social pressure and incentives, not code. If Hotelist manages to keep a genuine spread of scores, the achievement will lie in its design and community rather than in how quickly it was prompted into existence. As AI makes software cheap to build, the lasting advantage moves to whoever can earn trust, whether that means auditable extensions or ratings people believe.