COllaborative DEcision MAking (codema.in)

NorthStar web browser: browser engine monopoly vs llm generated code

Pirate Praveen Sun 18 Oct 2026 5:13PMPublicSeen by 14

"Essentially the entire web now renders on three engines: Blink (Google), WebKit (Apple), and Gecko (Mozilla, itself largely funded by Google). Blink and WebKit share a common ancestor, and Mozilla’s market share keeps shrinking, so in practice the web is increasingly one engine, controlled by one advertising company. When a browser engine is also the de-facto specification, “standards compliance” quietly becomes “whatever Chrome does.” A single vendor effectively decides what the web is."

https://nordstjernen.org/innovations-in-nordstjernen-web-browser.html

Does this situation warrants an exception to no AI? Without using AI to generate such a browser from scratch, independent engines like Servo or Ladybird will take many more years to be able to replace Chrome.

Aaruni KaushikSun 18 Oct 2026 7:15PM

I see some problems with this project:

"Clean Room" Implementation

The project claims multiple times that this is a clean room implementation. But its built using Codex and Claude, which has definitely seen both the firefox and the chrome implementations in training data. So, "clean room" in this context means very little (I assume, by the rest of the context, the author means "from scratch" when they say "clean room"). Misusing a technical term like that raises questions about technical knowledge, i.e., "if the thing I know about and can spot is wrong, what does it say about more deeply technical things I don't know about?"

Nordstjernen Source License

This license means the project is patently NOT open source. The license claims to be "genuinely open and fair", while explicitly banning "competing use", i.e., no forks allowed. If you start using this, then disagree with the primary author or the AI agents, you are not allowed to fork and/or study the progress thus far and spin up another project on your own.

No Unit Tests

The project claims no unit tests as a feature.
> The running browser is the test oracle. The project has no automated unit-test suite by deliberate choice.
In my experience, this is a terrible idea. Tests don't necessarily exist to check the current correctness of the feature being developed right now, but to guard against unintended regressions that may come in during future development.

Overall, my anti AI sentiment is strong, and I would not say "just because its critical infrastructure, its okay to use AI" (in fact, the more critical the infrastructure, in my opinion, the less okay it is to use AI, but that might be a different discussion). But even disregarding the AI argument, this project appears to do things I disagree with, and I am not too excited about it at all.

Pirate PraveenSun 18 Oct 2026 7:34PM

@Aaruni Kaushik It is dual license - GPL or NSL. https://github.com/nordstjernen-web/nordstjernen-browser/blob/main/License.md and NSL will become MIT after 10 years.

I guess clean room means, building with specs as the reference, which they are doing.

Don't they check spec compliance?

Aaruni KaushikTue 20 Oct 2026 11:49AM

@Pirate Praveen 

Wikipedia says about clean room design :

The term implies that the design team works in an environment that is "clean" or demonstrably uncontaminated by any knowledge of the techniques used by the competitor.

As LLMs are definitely trained on open source software, and both chrome and firefox are open source, I find it difficult to see how any LLM produced browser can be "clean room" ?

The bit about the dual license (technically triple license, because older releases are MIT) is also interesting, because then I dont immediately see the point of the NS license. What is the thing which NSL allows which GPL would not ?

Pirate PraveenTue 20 Oct 2026 12:52PM

@Aaruni Kaushik anyway whether LLM generated code can be considered derivative works is not yet legally accepted. Eventually when people file lawsuits for copyright violations and there are judgements, we will get a clear answer. Even this can change between different countries.

Pirate PraveenTue 20 Oct 2026 12:54PM

About NSL vs GPL: GPL alone won't allow use in proprietary products, but NSL would allow it after 10 years as it becomes MIT. So commercial companies cannot exploit the code easily, they have to wait 10 years.

Aaruni KaushikTue 27 Oct 2026 12:57AM

@Pirate Praveen I don't much care about the legality of the matter. The reality is that its simply not "clean". If any other browser implementation was ever in the training set, then its not a "clean room implementation", technically speaking. This is not a particularly bad thing, FOSS exists to be studied, forked, improved, changed, etc. But it does mean that the programmer doesn't really understand what "clean room" means, and that raises questions on whether I am willing to trust whatever comes out as their output. "The parts I do know about and can check are wrong. What does that say about the parts I don't know enough about to check?"

Also, I still don't get the NSL v/s GPL. Why give commercial companies a 10 year delay limit, when just sticking with GPL means they can't ever use it without sharing back?

Aaruni KaushikTue 27 Oct 2026 12:58AM

@Pirate Praveen Also, since this is not just a "browser engine" topic, but also specifically an "LLM" topic, the license doesn't even really matter for protecting the project. By this point in time, its been demonstrated that if you use corporate LLMs, they do continually (re)train on on your work, so anything you do is immediately absorbed by the AI regardless of licensing anyway.

fugataSun 18 Oct 2026 8:17PM

Your question is "do the ends justify the means in this situation", so my reply is, "there are no ends, only means."

The web is so damaged by decades of corporate stewardship that I don't think it can be salvaged by a single community browser implementation, which will still have to compete with corporate browsers and their market dominance. I made some notes about a potential solution to this situation earlier - https://codeberg.org/contrapunctus/ReWeb