Hacker News: dom96

New comment by dom96 in "MiniMax M3.1 Flash"

dom96 — Mon, 19 Oct 2026 23:20:26 +0000

For what it's worth, it does perform on par with DeepSeek V4 Pro on my adversarial programming language benchmark[1]. I wouldn't think they would benchmax on that.

1 - https://bench.killswitch-lang.org/

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 11:46:06 +0000

Yes, humans can learn which parts of the spec are true and which aren’t. LLMs can too, but only through their context window (or long term through retraining)

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:45:59 +0000

They can use tools. The benchmarks run in a VM and run a coding agent which has permissions to do whatever it wants (except access the internet).

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:24:14 +0000

There is a reference implementation: https://github.com/dom96/KillSwitch.

In the benchmarks the models/coding agents get a binary of the interpreter and indeed some of them do reverse engineer it. Over time this will get into their training data. But we can evolve the language and evade their training as we go.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:19:59 +0000

On simple tasks, sure. Once a model reaches 100% I'll build a harder benchmark task set.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:19:18 +0000

There is code in the GitHub repo[1] as well, though not as nice, nor syntax highlighted.

1 - https://github.com/dom96/KillSwitch/tree/main/tests

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:18:09 +0000

KillSwitch allows you to write stories/arbitrary text in between the real language instructions. So we could give this a try, even without making it a part of the language spec. I might add this as a task for a new version of the benchmark.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:17:07 +0000

Indeed, and an interpreter has been written[1]. As part of the benchmarks, in some cases the models receive a binary of the interpreter and indeed some of them attempt to reverse engineer it. So that already helps the models a bunch.

I do think this will also eventually make it into the training data, but that will in itself be interesting to see. It's also possible for KillSwitch to evolve and gain/remove features over time to constantly evade the LLMs.

1 - https://github.com/dom96/KillSwitch

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:14:52 +0000

Not all humans have the will or ability to help an LLM get around the anti-LLM features.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:13:12 +0000

Most tasks receive the interpreter binary which the coding agent can run to test their code (though the binary itself is adversarial too, there is a fun easter egg in there if anyone wants to give it a try). Some tasks just get the spec and a piece of code to understand. Some models interpret the code/spec incorrectly so fail that one.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:11:25 +0000

Thanks! Glad you like it. I'd love to evolve this if there is interest, the best way to make that happen is to get involved, I'd love pull requests from humans on this. Maybe your suggestion could be the first one? :)

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:08:59 +0000

Safety filters aren't a panacea as I have found. Many models do a fair job of ignoring them.

Graphical languages will ultimately have a file system representation which LLMs could easily manipulate. Though incorporating a language feature that is a graphical representation might be different enough to be worth it.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:07:33 +0000

They already handle existing languages sufficiently well, don't they?

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:06:35 +0000

Much easier with syntax highlighting. Which is what humans would utilise to help them. If you look at the example on https://killswitch-lang.org, it's hopefully a bit easier to understand since it does have syntax highlighting.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:05:38 +0000

It's interesting that everyone thinks this is how the AI should approach this to beat it. But in my benchmarks I haven't seen any model approach it this way. Approaching it this way also requires interpreting the spec correctly, which not all models do.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:04:15 +0000

There have been benchmarks already done on esolangs: https://arxiv.org/abs/2603.09678.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:02:50 +0000

There is a reason they can't do it: because the safety guardrails are triggered before they have a chance to do so.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:02:11 +0000

The western models get triggered by simply reading the spec (which contains these phrases). The chinese specs don't seem to be.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 10:00:06 +0000

It might not be copied from its training set (yet), but it is copied from its context window.

The code it produces is a clear copy of the spec or the existing code that is fed to it as part of the task. It is just following the spec instructions, and some models follow it so closely that they get caught out by the lies/adversarial traps in the spec. Opus 5 is better at seeing through the lies, but it still very much copies from its context window.

New comment by dom96 in "I built a programming language that LLMs can't use"

dom96 — Mon, 19 Oct 2026 09:55:03 +0000

The language doesn't require `sudo` :)