Hacker News: thatjoeoverthr

New comment by thatjoeoverthr in "Not everyone is using AI for everything"

thatjoeoverthr — Sun, 14 Jun 2026 18:20:52 +0000

Unwise design. “It talks, CSRs talk, it’s the same thing”. The fact CSRs talk is incidental. Nobody contacts support to talk. Customer service is a kind of “exception handler” for that which you failed to automate. If your system exists, works and is legible, conversation is avoided.

New comment by thatjoeoverthr in "Not everyone is using AI for everything"

thatjoeoverthr — Sun, 14 Jun 2026 16:42:43 +0000

Last time I was in Germany I saw what appeared to be homeless children

New comment by thatjoeoverthr in "If you are asking for human attention, demonstrate human effort"

thatjoeoverthr — Fri, 12 Jun 2026 06:04:27 +0000

So he’s redundant. You call Uber Eats and you don’t pay a salary for that.

New comment by thatjoeoverthr in "Dopamine Fracking"

thatjoeoverthr — Mon, 08 Jun 2026 13:16:46 +0000

We have social media-like systems going arguably Compuserve and the like, as well as games. There's a matter of "refinement", like how some older people describe the change of drugs over the decades. TikTok, Twitter and many of the games are just "too strong", and it matters. Nobody gets "addicted" to Mario 3 or IRC to the point it resembles alcoholism.

New comment by thatjoeoverthr in "GPT Guesses Between 1 and 100"

thatjoeoverthr — Mon, 25 May 2026 13:48:10 +0000

Is there a reason this was done with such a large sampling when you can read the logits one-shot?

I did this for an article, like so:

https://joecooper.me/blog/gptprimer/food.webp https://joecooper.me/blog/gptprimer/math.webp https://joecooper.me/blog/gptprimer/butts.webp

OpenAI removed this interface from their newer models, but IIRC you can still do this against 4.1 and 4o.

New comment by thatjoeoverthr in "New arXiv policy: 1-year ban for hallucinated references"

thatjoeoverthr — Fri, 15 May 2026 08:52:03 +0000

No mercy to brain slugs.

New comment by thatjoeoverthr in "AI could be the end of the digital wave, not the next big thing"

thatjoeoverthr — Mon, 13 Apr 2026 15:09:47 +0000

It's threatening to "unwind" the entire digital sector back to 1990. Online shopping damaged, job interviews done in person, essays by hand, exams proctored. Cover letters obsolete. There could be a "cognitive waterline" effect where older people who can't tell will continue living in an AI-generated bubble. Cover letters already are generated on demand specifically because people still claim to require them, even though we know they're not real anymore.

Could be an advantage to knowing this because you can step around it.

_You_ know it's AI, so you go in person to a store. Likewise, next time you hire, you can simply refuse to accept "cover letters".

New comment by thatjoeoverthr in "EFF is leaving X"

thatjoeoverthr — Thu, 09 Apr 2026 21:45:05 +0000

Link suppression is contested. Nikita says they're not deboosted. This guy tested it, found evidence they're no longer deboosted: https://x.com/phl43/status/2041893735827460446?s=20

Nikita says they were "never" deboosted, but Musk said they were going to do that and it was a huge topic...?

https://x.com/nikitabier/status/2041911302541730237?s=20

He says here about an interface change. I've noticed this change. The sites are opening in a kind of sub window with the feedback UI still visible. I found this annoying but now I see the point.

New comment by thatjoeoverthr in "AI may be making us think and write more alike"

thatjoeoverthr — Tue, 07 Apr 2026 13:19:27 +0000

I've been calling them "meat condoms". In the workplace, it's one or two warnings before completely ejecting them. On social media, instant block.

New comment by thatjoeoverthr in "What came after the 486?"

thatjoeoverthr — Thu, 26 Mar 2026 11:46:53 +0000

IIRC III was also 6th gen microarchitecture. Pentium IV was the 7th.

New comment by thatjoeoverthr in "Allow me to get to know you, mistakes and all"

thatjoeoverthr — Sun, 15 Mar 2026 19:40:11 +0000

You can easily deny this upside. Your playing field isn't level because instead of grammatical mistakes, you have the online equivalent of talking like a used car salesman.

New comment by thatjoeoverthr in "Allow me to get to know you, mistakes and all"

thatjoeoverthr — Sun, 15 Mar 2026 19:37:11 +0000

The em dash is famous but I've noticed since (I think) December every hustler suddenly at once started using drama dots.

Like... That. Rhetorical ellipsis. Like you see in a 12 year old's fanfic.

I know one of the AIs had a style change. I think Grok. But it started using drama dots so now they are everywhere.

And unlike the em dash, _nobody_ notices. _Nobody_ sees it.

New comment by thatjoeoverthr in "“Car Wash” test with 53 models"

thatjoeoverthr — Tue, 24 Feb 2026 11:32:14 +0000

While that's true, the tokenizer is half the problem. The important fault demonstrated is it doesn't _know_ it can't see the letters, and won't express this unless it has been trained or instructed to. "I can't see letters through the tokenizer" never appears in a corpus of human writing.

New comment by thatjoeoverthr in "I need AI that scans every PR and issue and de-dupes"

thatjoeoverthr — Sun, 15 Feb 2026 22:42:47 +0000

It's surprisingly difficult, and the "obvious" techniques (just do embeddings) don't really work. I wrote about it and did benchmarks here: https://joecooper.me/blog/redundancy/

New comment by thatjoeoverthr in "I started programming when I was 7. I'm 50 now and the thing I loved has changed"

thatjoeoverthr — Tue, 10 Feb 2026 15:48:09 +0000

Speaking directly, if I catch the scent of ChatGPT, it's over.

People put out AI text, primarily, to run hustles.

So its writing style is a kind of internet version of "talking like a used car salesman".

With some people that's fine, but anyone with a healthy epistemic immune system is not going to listen to you.

If you want to save a few minutes, you'll just have to accept that.

New comment by thatjoeoverthr in "Stargaze: SpaceX's Space Situational Awareness System"

thatjoeoverthr — Fri, 30 Jan 2026 13:00:29 +0000

Kessler problems require Kessler solutions

New comment by thatjoeoverthr in "The Hallucination Defense"

thatjoeoverthr — Fri, 30 Jan 2026 01:58:53 +0000

You're stretching it. It's more like if you train your dog to start the car and accelerate, open the door and turn your back.

Everything an AI does is downstream of deliberate, albeit imperfect, training.

You know this, you rig it all up and you let things happen.

New comment by thatjoeoverthr in "Microsoft forced me to switch to Linux"

thatjoeoverthr — Wed, 28 Jan 2026 18:45:44 +0000

Incredible moment when you have to ward off Windows, macOS and iPhone updates like a bouncer.

I’ve gone over the years from Visual Studio fanboy to writing everything in vi, entirely due to software decay.

Our culture and economy can no longer maintain complex GUIs.

New comment by thatjoeoverthr in "LLM-as-a-Courtroom"

thatjoeoverthr — Tue, 27 Jan 2026 22:37:53 +0000

If you do want a numeric scale, ask for a binary (e.g. true / false) and read the log probs.

New comment by thatjoeoverthr in "I know you didn't write this"

thatjoeoverthr — Mon, 22 Dec 2025 18:55:39 +0000

“Any time saved by (their) AI prompting gets consumed by verification overhead, …”

This

When I receive a PR, of course it’s natural an AI is involved.

The mortal sin is the rubber stamp.

If they haven’t read their own PR, I only have so many warnings in me. And yes, it is highly visible.