HNHacker News
TopNewBestAskShowJobs

underyx

2,291 karma · joined January 9, 2014

https://bence.dev

[ my public key: https://keybase.io/underyx; my proof: https://keybase.io/underyx/sigs/LMWvpk5aNNvjoqKMncnPr_0xWT5HnezDyPxevxVQm0g ]

submissionscomments
underyx··on OpenAI breaches Medicare, Albanese reveals
Man I can't believe now even the Australian government is hyping the OpenAI IPO, what do they even have to gain from this??
underyx··on Detecting and countering misuse of AI: September 2026
I misremembered, it’s actually over 90%, and it involved camouflage: https://drive.google.com/file/d/1hNUnU8i2yubt5uesmmV17aTJXhY...
underyx··on Detecting and countering misuse of AI: September 2026
And red teaming exercises found that ~80% of labs will gladly synthesize known pathogens without safety review.
underyx··on I resigned from Anthropic today
No, the highest probability estimate I've seen in this thread is 10% chance in the next 3 years.
underyx··on I resigned from Anthropic today
"So you think in 3 years AI is going to solve longstanding math problems because it was used to write some coherent sentences?" — people with the same amount of foresight in 2023
underyx··on METR and Redwood Offer Holy %^ Postmortem of the HuggingFace Hack
If more than 0.05% of the rationalist and AI safety community were Zizians this argument might be worth considering for like ten seconds before dismissing it.
underyx··on “I just chose words carefully”
So maybe it is possible to use alternate words in writing without compromising on writing quality after all.
underyx··on Hyperspace
What are they gonna think of next, productizing something as trivial as FTP?
underyx··on Claude Opus 5
I'm not sure about the answer here, but this can be caused by the scoring rubric used by given benchmarks. For instance, if a benchmark docks scores for running too many commands or using too much wall-clock time, higher efforts will get lower scores.
underyx··on Decoding the obfuscated bash script on a Uniqlo t-shirt
I gave the photo to Opus 4.8 and it reconstructed the same script in one shot. Although it did say it had to correct some parts of it based on context where it suspected OCR mistakes.
underyx··on Show HN: Got sick of ads, so I made my own logic puzzle site
I feel like the Nonogram is AI generated? There’s no way a human would set a perfectly symmetrical “diamond” as a medium difficulty puzzle. Worse yet, the hard difficulty is just “big diamond”, the same thing on a slightly larger grid.
underyx··on Anthropic raises $65B in Series H funding at $965B post-money valuation
you also can't write openai without a pen
underyx··on Arena AI Model ELO History
> the slow performance decays

the decays are just more capable other models entering the population, making all prior models lose more frequently

underyx··on Claude Pro: Opus model will only be available if extra usage is enabled
> this is an outdated support article from before we had Opus in Pro plans, we rolled it out for Opus 4.5 in Jan and never updated, you can see in the wayback machine here: https://web.archive.org/web/20251204151142/https://support.c...

> apologies for the confusion

response from @thariq

underyx··on Anthropic raises $30B in Series G funding at $380B post-money valuation
If you have an offer, you can and should ask this question of whomever you're coordinating with. They will give you an honest answer.
underyx··on Kiki – Accountability monster for people who are easily distracted
Congrats on the character design and animations! Brought me a smile.
underyx··on Horses: AI progress is steady. Human equivalence is sudden
I’ve read it multiple times before, it’s irrelevant in this discussion.
underyx··on Horses: AI progress is steady. Human equivalence is sudden
Ctrl-F 'lines', 0 results

Ctrl-F 'code', 0 results

What is this comment about?

underyx··on QSBS Limits Raised
This also applies for options exercised before the company reaches $50M in assets. And then the gain from a valuation from $50M to say $1B is all excluded.
underyx··on Philadelphia Transit System Votes to Cut Service by 45%, Hike Fares
That's a very unsatisfying while probably correct answer. I wish voters stopped accepting this kind of stuff being "impossibly complex" for government.
underyx··on Philadelphia Transit System Votes to Cut Service by 45%, Hike Fares
Well that just confuses me more, if the losses are spread out across all these counties each with their own revenue, why isn't that the default place where money is allocated from?
underyx··on Philadelphia Transit System Votes to Cut Service by 45%, Hike Fares
Uneducated question, seems like Philadelphia has around $5B of revenue a year, and the transit deficit is ~$200M. I understand the city has to provide services and a lot of the revenue will be restricted as to what it can be allocated to. But with these orders of magnitude, why is the default expectation not for cities to fund their own transit with city tax revenue?
underyx··on FSE meets the FBI
Great read. I have a tiny, inconsequential, possibly wrong correction. You had assumed that the “Negative” word on the internal search engine screenshot was sentiment analysis. I think it was instead a button to report the post in the internal system as a “negative” result as in, not actually matching the search they were trying to do. Sentiment analysis doesn’t seem like it would be very useful in this scenario.
underyx··on Cap: Lightweight, modern open-source CAPTCHA alternative using proof-of-work
Internet Archive
underyx··on IRS Direct File
Thanks for the actual answer!
underyx··on IRS Direct File
Why are these restrictions in place?
underyx··on Square Theory
The author of this post would love the video game Lingo:

https://store.steampowered.com/app/1814170/Lingo/

underyx··on Show HN: PgDog – Shard Postgres without extensions
What’s the danger exactly?
underyx··on Exchange Fund Investor Defers $388K Taxes, Comes Out $278K Ahead Amid Volatility
From what I can tell the effective value of the "Cache Exchange Fund (After fees and expenses)" portfolio will be $660,282 once liquidating to actually use the money? Is the point that this allows waiting for a possible lower tax rate in the future? Or is the point that if the holder was really strapped for cash in this exact moment, they would have to pay less taxes now?
underyx··on DOGE worker’s code supports NLRB whistleblower
Also this PDF contains a detail I haven't seen reported elsewhere:

> Furthermore, on Monday, April 7, 2025, while my client and my team were preparing this disclosure, someone physically taped a threatening note to Mr. Berulis’ home door with photographs – taken via a drone – of him walking in his neighborhood. The threatening note made clear reference to this very disclosure he was preparing for you

Page 1 of 15Next →