Tweeks.io Lets You Fix the Internet One Browser Tab at a Time

The Macro: Websites Are Broken and You Already Know How to Fix Them

Every person who uses the internet regularly has opinions about how websites should work. YouTube’s autoplay is annoying. LinkedIn’s feed is a dumpster fire of engagement bait. Amazon buries the actual product specs under seventeen sections of sponsored recommendations. These aren’t edge cases. These are the daily frustrations of a billion people who have no ability to change any of it.

Browser extensions exist, of course. Greasemonkey has been around since 2004. Tampermonkey is popular with developers. uBlock Origin handles ads. But the gap between “I wish this website worked differently” and “I can write a userscript to make it work differently” is enormous. It requires JavaScript knowledge, DOM manipulation skills, and enough patience to debug things when the website pushes an update that breaks your script at 2am.

The no-code movement promised to close gaps like this, but it mostly stayed focused on building new things rather than fixing existing ones. Lovable, Bolt, and a0.dev let you create apps from scratch with AI. What nobody was really building was a tool that lets you modify apps that already exist. Tweeks.io is going after that specific problem, and the framing is smart: you shouldn’t need to be a developer to make websites work the way you want.

The browser extension market itself is substantial. Chrome Web Store has over 200,000 extensions. Firefox has tens of thousands more. But the creation side has always been gated by technical knowledge. If Tweeks can genuinely turn natural language into reliable browser modifications, they’re opening that creation layer to everyone who has ever muttered “why does this website do this” under their breath.

The Micro: Describe What You Hate, Get a Fix That Sticks

The core loop is simple. You describe what you want changed on a website in plain language. Tweeks generates the modification. You install it. It persists across page reloads and browser sessions. No coding, no debugging, no reading Stack Overflow threads about content security policies.

They call individual modifications “tweeks,” and there are already over 4,000 public ones in their community library. That’s a meaningful number for a product this early. It means people aren’t just trying the tool and leaving. They’re building things and sharing them. The library covers the usual suspects: YouTube, Google, Amazon, Reddit, LinkedIn, X, ChatGPT. Basically every website people love to complain about.

The technical underpinning is Greasemonkey-compatible userscripts, which is a smart architectural choice. It means the modifications are built on a format that’s been battle-tested for two decades. Advanced users can view and edit the source code of any public tweek. Beginners never have to see it. That layered approach is how you build for two audiences without alienating either one.

Browser support is broad. Chrome, Edge, Brave, Opera, Vivaldi, Arc, Comet, Firefox, and anything else built on Chromium. That covers essentially the entire browser market minus Safari, which has always been hostile to extensions anyway.

The numbers they’re reporting are solid for a YC W25 company: 15,000 active users and over a million pages modified. The free tier covers installation, browsing, and basic creation. Premium features exist but the details are thin.

Jason Madeano, co-founder and CEO, previously led recommendation systems as an ML engineer at Pinterest and holds MIT degrees in Computer Science and Brain & Cognitive Science. His NeurIPS best paper on cultural transmission and reinforcement learning suggests he thinks seriously about how AI models learn from human behavior. Co-founder Matt Stallone led post-training for IBM Research’s Granite foundation language models, also out of MIT with published first-author LLM research. This is a team with deep model training experience, not just API wrapper skills.

The security model matters here more than in most products. Browser modifications that run on every page load have real attack surface. Tweeks runs modifications in a security sandbox, and the fact that all public tweeks have viewable source code helps with transparency. But the AI-generated modification pipeline will need to prove it can’t be manipulated into producing malicious scripts.

The Verdict

I like what Tweeks is doing. The insight that people want to modify the web, not just build new things on it, is underserved and correct. The community library approach creates network effects that most browser extension tools don’t have. And the AI generation layer genuinely lowers the barrier from “needs to code” to “needs to complain.”

The questions I’d push on: how reliable are AI-generated modifications when websites push updates? YouTube changes its DOM structure regularly. A tweek that works today might break Thursday. If users have to keep regenerating modifications, the magic fades fast.

Competitors in adjacent spaces include Stylus for CSS modifications, Arc browser’s built-in boost features, and the entire Tampermonkey/Greasemonkey ecosystem for technical users. None of them have the AI generation layer, which is the moat Tweeks is building. Whether that moat holds depends on how good the generated scripts actually are in practice, not just in demos.

At a million pages modified, something is clearly working. The question is whether it scales from power users and early adopters to the mainstream audience that would make this a big business.