Comper - Understand your code landscape at AI speed

How it started: "GeoGuessr for code"

The first idea for Comper was "GeoGuessr for code". We were playing GeoGuessr in the office quite a lot and then my friend Nike said "why don't you build that, but for code". So you see a snippet of code and have to guess where in your company's codebase this is from. I built an early prototype and it was a lot of fun!

Then this "map of all software in your company" idea turned serious because I realized that very very few people in the company understand the software landscape. Who knows about these microservices? Which apps depend on our service? How much technical debt do we actually have? These were questions that as CTO, I had a good handle on, but very few others did.

By now, Comper is a real-time map where you can see the entire shape and history of a codebase, and make sure that everyone has the same mental model. We believe that humans are visual creatures and we believe that you should show, don't tell, if you want to convince or teach others.

What we're launching: See and Explore

Today, we're showing the world what we've been building. A new way to deal with the mountains of software in your organization.

Comper is building in 4 phases. See, Explore, Understand, Action. We're still building Understand and Action with our early customers, but today you can already See and Explore on two public boards. We're launching with the Rust and OpenClaw landscapes, because of two simple reasons: Comper is built on Rust (and we love it), and OpenClaw is interesting because of how explosive the growth of this repo is.

5 interesting insights: Try the boards yourself

So here are 5 interesting insights from these boards, but most importantly: please give it a spin yourself at comper.io/rust and comper.io/openclaw.

1

The OpenClaw repo is bizarre

This is the kind of board that makes you stop scrolling for a second. Even at a glance, you can see that the repo is growing at a pace that is hard to hold in your head without a visual map.

The growth and scale is insane: 55M LoC for skills and 2.4M for OpenClaw, within a couple of months.

The skills repo is a piece of work. Would be good as a piece of art!

2

AI maintained documentation is the future.

Most documentation goes stale the moment it is written. If the map, diagrams, and context can be generated straight from the codebase, teams get something that is actually useful on a random Tuesday instead of a document nobody trusts. We always keep docs in sync after a push.

3

Servo has been growing steadily and the repo is only 4% Rust

That makes it a great example of why language labels alone do not tell the full story of a codebase. The interesting part is not just what percentage is Rust, but how the project has evolved over time and where that momentum is concentrated.

There seems to have been a near-total activity stop in 2021/2022, which makes the overall arc of the project even more interesting. And most of the repo is test files, mostly for CSS. Yes, CSS is hard. Apparently, parsing HTML is still the biggest chunk of code, and canvas needs more code than CSS. Who knew stream and webcrypto were this complicated!

4

Smol is really small

There is something refreshing about looking at software that still feels legible in one view. Smol looks like the opposite of sprawl: compact, understandable, and not yet buried under layers of historical complexity.

2.1k lines, mostly Rust, not duplicated. A good example of "software that is just finished".

5

The game is still really fun

We're calling the game GitShuffle, and just a hint: it's more fun on your own repos. Here's a quick video to show you how it works.

It turns out that guessing code location is not just a gimmick. It is a surprisingly good way to test whether a team really shares the same mental model of a codebase.

What we added to the boards

For OpenClaw, we've now added the entire OpenClaw organization. For Rust, we've added the following projects, organized by group: