Hacker News: gowld

New comment by gowld in "AI;DR (AI; Didn't Read)"

gowld — Mon, 17 Aug 2026 22:26:24 +0000

"Good" is subjective, but I found this OOTB sufficiently information-dense and easy-to-customize:

https://claude.ai/share/ac747c6d-2a62-40b7-b33e-06a94cf34344

New comment by gowld in "Going Dark, and the era of law enforcement hacking"

gowld — Fri, 14 Aug 2026 22:24:35 +0000

"backdoors" aren't holes, they are well-documented superuser APIs. The hack will be by compromising the API user's credentials (moles, stolen passwords, etc), not compromising the server's design intent.

New comment by gowld in "What sort of maths are LLMs good at?"

gowld — Wed, 12 Aug 2026 21:55:34 +0000

How much API usage cost (or what subscription did it fit in)?

New comment by gowld in "What sort of maths are LLMs good at?"

gowld — Wed, 12 Aug 2026 20:50:26 +0000

> the very short disproof of the Jacobean Conjecture.

That's a counterexample (finding a needle in a haystack), not an elegant proof. Proving the conjecture would be elegant, if it were true but somehow still resisted proof nearly as much as the conjecture did because the conjecture was false.

New comment by gowld in "What sort of maths are LLMs good at?"

gowld — Wed, 12 Aug 2026 20:48:33 +0000

That's trivially solvable by deploying multiple AIs on one project, so that can't be the true obstacle.

New comment by gowld in "What sort of maths are LLMs good at?"

gowld — Wed, 12 Aug 2026 20:47:40 +0000

One question I have about ARC-AGI-3 is how much it depends on vision ability, which has a substantial hardware component that isn't "General" "Intelligence".

What happens if the game is encoded in a non-visual logical form?

New comment by gowld in "What sort of maths are LLMs good at?"

gowld — Wed, 12 Aug 2026 20:46:21 +0000

You wrote:

> > It's the cumulative knowledge of all general human intelligence, baked into an artificial form, which can then use that knowledge to achieve novel goals.

But it (currently) can't solve puzzles like ARC-AGI-3 that children can solve.

New comment by gowld in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"

gowld — Wed, 05 Aug 2026 23:11:53 +0000

Or Sundar wanted Demis to change to work on other things.

New comment by gowld in "Discovery Loop"

gowld — Wed, 05 Aug 2026 23:05:58 +0000

"Solar PV with storage = solar PV installation paired with four-hour duration battery storage, scaled to 20% of the output capacity of the solar PV."

Sometime the sun goes away for more than 4hrs.

That may be OK for closed-ended systems (turn off the science at night and during storms), but not for open-ended systems with diverse user demand.

4hour batteries are competitive with gas peakers to match high demand during and after sunny times, but solar needs gas peakers or similar to over for non-sunny times.

New comment by gowld in "Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery"

gowld — Wed, 05 Aug 2026 22:58:58 +0000

It's a lot easier to automatically detect a match to 1 specific thing than detect a match to a completely novel thing that offends sensibilities in a similar way to other things.

New comment by gowld in "Discovery Loop"

gowld — Wed, 05 Aug 2026 22:52:08 +0000

That's solved by either private monopoly charging money to fund investment, or public monopoly raising taxes to fund investment.

New comment by gowld in "Discovery Loop"

gowld — Wed, 05 Aug 2026 22:51:01 +0000

It's a lot more complicated than that -- Solar is great for discretionary, opportunistic, and time-shiftable additive demand, but not for replacing all-day baseline load. Using solar to dip into baseline demand sometimes is catastrophic, because the "profit" is exploiting its failure to provide consistent, reliable power but charging "default" prices that have an implicy assumption that power is consistent and reliable.

New comment by gowld in "All of Winona Police Department's Flock cameras cut down and stolen"

gowld — Tue, 04 Aug 2026 18:18:52 +0000

No, Flock plays a semantic game where data on Flock's servers os "owned" by the client, so Flock is not responsible for the content or management of the data.

New comment by gowld in "Why Large Language Models Fail at Tabular Prediction"

gowld — Tue, 04 Aug 2026 15:15:22 +0000

Solve what problem? My hammer can't drive screws. Is that a problem to solve?

New comment by gowld in "Ten advances in mathematics and theoretical computer science"

gowld — Mon, 03 Aug 2026 20:05:57 +0000

> In any research phd course you're actively told to bite off something small and likely to be provable so that you can prove it (and publish it).

But that's the start of math research, not the end.

The point is to get practice and experience doing research.

Did ChatGPT learn anything from these proofs, that it can build on?

Part of what's annoying people is that ChatGPT is churning though problems that are meant to be motivating. They are problems that aren't worth the effort of human professionals (usually because they are incredibly computation-hevy, so better suited for a computer than a human), so they are good for students to work on.

New comment by gowld in "Ten advances in mathematics and theoretical computer science"

gowld — Mon, 03 Aug 2026 16:57:20 +0000

Did you know that "revolutationary" is not equivalent to "useful", and "revolutionary" is quite ambiguous?

New comment by gowld in "Ten advances in mathematics and theoretical computer science"

gowld — Mon, 03 Aug 2026 16:56:19 +0000

What you are doing is "motte and bailey".

The motte is "AI useful". The bailey is "Singularity is nigh".

New comment by gowld in "Karpathy’s Pelican"

gowld — Mon, 03 Aug 2026 16:36:51 +0000

Why does Twitter show a reply "Yah" from a random fan?

xcancel shows a whole stream of thoughtful discussion.

New comment by gowld in "Taylor Farms has rewritten its cyclospora statement four times in sixteen days"

gowld — Mon, 03 Aug 2026 16:29:40 +0000

The WalMart Waltons are a "family"

At the core, "Family-owned" just means nepotism is preferred over merit.

(The mob is also "Family-owned")

New comment by gowld in "Prevent cognitive debt by manually retyping LLM-generated code"

gowld — Mon, 03 Aug 2026 15:27:31 +0000

better: before submitting the code, ask LLM to quiz you for understanding the code.