Skip to main content

We're not obsoleted by AI yet. Two more years? Maybe?

· 3 min read

Another year, another existential question: will AI tools/agents or AI browsers replace browser extensions? Spoiler alert - not yet. But let's talk about what actually happened in 2025.

The numbers that matter​

54 million requests intercepted across 83,971 users. That's a lot of HTTP traffic being tweaked, modified, and debugged by actual humans doing actual work.

Most users keep it simple - 6 rules or less. But there's a fascinating tail of some few hundreds of power users configuring 100+ rules. We have no idea who they are, because that's how anonymous tracking is supposed to work 😊. What are they building? Enterprise test environments? Elaborate API mocking setups? We don't know, but we're impressed.

What we've learned​

Feature complexity doesn't drive conversion. We built sophisticated functionality - breakpoints, logpoints, mock generation and auto-enabled sites. Usage is negligible. Meanwhile, simple request modification continues to dominate.

The 6-rule threshold is real. There's a massive drop-off after 6 rules. This suggests most use cases are relatively simple: override an API endpoint, inject test data, modify some headers. Done.

The AI question​

Will AI agents replace tweak in two years? Probably not entirely. Here's why:

AI is great at generating code, answering questions, even writing API mocks. But intercepting and modifying live HTTP traffic requires precision, not probability. You need to know exactly which request to catch, exactly what to change, and exactly when it fires. In addition to that, you probably won't want an AI agent poking around your live API traffic without strict controls.

That said, AI could make our UI obsolete. "Intercept all requests to /api/users and add an Authorization header" - that's a perfect AI prompt. The question isn't whether AI replaces HTTP interception, but whether it replaces our configuration interface (this is not just true for tweak of course).

What's next​

We're not pivoting to AI (everyone else already is). We're going back to basics:

  1. Double down on the core: Request modification works. Make it bulletproof, address edge cases and customer pain points.
  2. Explore GraphQL tooling: The signal is weak but consistent. GraphQL has made it, it has become a boring technology. Maybe there's something there.
  3. Rethink paid features: Complex features don't convert. What would?

54 million requests intercepted. 83,971 users trusting us with their HTTP traffic. Not obsolete yet.

See you in 2026 - if we're still here.

Want to help us not get obsoleted? Try tweak and tell us what actually matters to you. We're clearly bad at guessing.

If you liked this article, consider sharing (tweeting) it to your followers.


Did you like this article?

Need something else? Request a feature