Hacker News: dehsge

New comment by dehsge in "America will come to regret its war on taxes"

dehsge — Sat, 18 Apr 2026 19:08:13 +0000

There’s a bit of a double edged sword here. Removal of taxes leads to more coupled private/government relationships. Where external interests fund politicians to protect their own monopolistic interests. That money needs to come from somewhere. Wealthy people don’t want to pay taxes. But they sure do like to pay money to members of the government to further their own interests. Remember it was about taxation without representation. Wealthy people pay a lobbying ‘tax’ and get representation. In this way they are just paying a tax in a different way, where they get benefits and you get very little to nothing.

New comment by dehsge in "The dead Internet is not a theory anymore"

dehsge — Thu, 12 Mar 2026 21:02:48 +0000

Yeah that’s the trade off of this implementation. Lobste.rs already uses this implementation https://lobste.rs/about#invitations The comments are considerably better. I’m not even a member but get more out of reading those comments than hn, and I’ve worked at multiple YC’s. This place is not what it used to be.

New comment by dehsge in "The dead Internet is not a theory anymore"

dehsge — Thu, 12 Mar 2026 18:40:38 +0000

Because you have an initial user who invited the bots. The whole invite tree of this user can cull all invites given by the user who added bots.

New comment by dehsge in "The dead Internet is not a theory anymore"

dehsge — Wed, 11 Mar 2026 21:18:03 +0000

Members only comment blogs. Where you need an invite to comment also solves the problem. You need to know a real human to get access.

New comment by dehsge in "SkillsBench: Benchmarking how well agent skills work across diverse tasks"

dehsge — Tue, 17 Feb 2026 00:16:37 +0000

Compilers can never be error free for non trivial statements. This is outlined in Rices theorem. It’s one of the reasons we have observability/telemetry as well as tests.

New comment by dehsge in "Case study: Creative math – How AI fakes proofs"

dehsge — Mon, 26 Jan 2026 01:07:51 +0000

There are some numbers that are uncomputable in lean. You can do things to approximate them in lean however, those approximates may still be wrong. Leans uncomputable namespace is very interesting.

New comment by dehsge in "Don't fall into the anti-AI hype"

dehsge — Sun, 11 Jan 2026 15:49:45 +0000

Most math books do not provide solutions. Outside of calculus, advanced mathematics solutions are left as an exercise for the reader.

New comment by dehsge in "“Erdos problem #728 was solved more or less autonomously by AI”"

dehsge — Sat, 10 Jan 2026 18:42:26 +0000

There are other bounds here at play that are often not talked about.

Ai runs on computers. Consider the undecidability of Rices theorem. Where compiled code of non trivial statements may or may not be error free. Even an ai can’t guarantee its compiled code is error free. Not because it wouldn’t write sufficient code that solves a problem, but the code it writes is bounded by other externalities. Undecidability in general makes the dream of generative ai considerably more challenging than how it’s being ‘sold.

New comment by dehsge in "AGI fantasy is a blocker to actual engineering"

dehsge — Fri, 14 Nov 2025 22:25:27 +0000

LLMs are bounded by the same bounds computers are. They run on computers so a prime example of a limitation is Rices theorem. Any ‘ai’ that writes code is unable (just like humans) to determine if the output is or is not error free.

This means a multi agent workflow without human that writes code may or may not be error free.

LLMs are also bounded by runtime complexity. Could an llm find the shortest Hamiltionian path between two cities in non polynomial time?

LLMs are bounded by in model context: Could an llm create and use a new language with no context in its model?

New comment by dehsge in "Show HN: Why write code if the LLM can just do the thing? (web app experiment)"

dehsge — Sat, 01 Nov 2025 20:50:04 +0000

There still maybe some variance at temperature 0. The outputted code could still have errors. LLMs are still bounded by the undecidable problems in computational theory like Rices theorem.

New comment by dehsge in "ProofOfThought: LLM-based reasoning using Z3 theorem proving"

dehsge — Sat, 04 Oct 2025 20:06:59 +0000

LLMs and its output are bounded by Rices theorem. This is not going to ensure correctness it’s just going to validate that the model can produce an undecidable result.