pliny

>Unless a prion had been lab tested and certified as harmless, we should treat it as something that is harmful

In this case the virus escaped during the testing process to certify or turn the virus harmless, so it's unclear what you mean by "treat it as something that is harmful" other than testing it and trying to make it less harmful.

FWIW I don't understand the point of the virus analogy since LLMs are not very similar to viruses and most people (on HN and in general) do not have much better intuitions about security in biolabs as opposed to security in ML research environments.

Your description of the HF attack as being merely "the elite task of discovering 14 Hugging Face API tokens that careless developers had committed to public GitHub repositories" does not match the description in the technical report[1].

[1] https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c78... - page 9

This is an AI written post and the details are wrong (the description of the HF incident as involving Irregular is wrong and the description of the incident as only involving stealing public credentials is wrong, per the technical report the agents got access to internal HF infrastructure).

They had very large put positions on stock that they had very small long stock positions in, I think they were net short in almost everything they held puts on (except the smh etf put which I guess is an attempt to cancel out sector beta)

I don't have anything to say about their concentration beyond that if you read their early 13Fs (ex https://13f.info/13f/000204572425000006-situational-awarenes...) they were in fact "that concentrated" which is why it's plausible they got 100s of % returns without leverage. Re options being leverage - everything you said is true but unfortunately the public filings dont have strikes or durations so it's not possible to say whether they bought short dated otms with 0.05 delta or leaps with close to 1 delta or something in between.

I was answering his point that 400% returns are impossible without leverage, and my argument was that they were in the right stocks at the right time to achieve those kinds of returns without leverage. Since a lot of their portfolio is/was options without strikes or durations listed and the exact dates of their buys/sells is unknown it's not possible (AFAIK) to reconstruct their P&L exactly but it seems plausible to me that, given the stocks they were concentrated in and the quarters they start reporting those positions, that they could have gotten 400% returns in a year without leverage. If I were an LP reading their filings saying that (for example) they held 25% of the fund in Intel during a half year period where it went up 200%, plus some other similar holdings, resulting in overall 400% returns I wouldn't automatically conclude they were levered.

>Returns like that are not asymmetrical and can only be produced with leverage, at least when you're trading paper.

Definitely not true, looking at their last 13f (filed in may 2026) their top3 holdings were BE, SNDK and CRWV which had 1y returns of 1,500%, 2,000% and 500% at the time of the filing. They accounted together for 15% of the fund (plus an unknown amount of exposure through options). These tickers show up in the 2 and 3 previous filings so they had exposure to some of that run up, and looking at the filings further back they had some very concentrated exposure to INTC in a half year period where the stock went up about 200%

The other side could be someone with natural exposure to the question that wants to hedge, for example people traveling to/from the middle east were exposed to the Iran war question(s) and could get insurance against airspace closure through prediction markets.

This argument doesn't work for 90+% of the volume on PM/Kalshi but I think most of the questions there are just gambling.

>What was slow?

Everything, huge input delay in every interaction, clicking on anything, opening menus, typing, tabbing between windows, everything had 1-2s of delay.

>disable variable refresh rate

I think I tried this but dont recall, there were a few things related to monitor refresh I tried that probably included this

I got a new computer a couple weeks ago, with a 5070, and installed ubuntu on it and it was incredibly slow. I looked online and found some claim that 24.04 has some incompatability with nvidia, tried installing a bunch of different driver versions and nothing helped, tried turning everything off in gnome tweaks and still slow, tried installing 26.04 and 22.04 but the installer hangs forever in both, tried linux mint 24.04 still slow, gave up and installed windows with WSL :/

I asked this (in a less accusatory tone) of an NSO employee once and he said something about how the big tech companies also spy on people and do unethical things.

There are already cheap, domestic robots for cleaning dishes, cleaning the floor, cleaning clothes, making coffee, heating and cooling food, turning screws, drilling holes and so on. All those robots represent a greater than 90 percent (and sometimes a greater than 99 percent) savings in time relative to doing the same tasks manually. You still have to move the objects they operate on around within your house but that's mostly the only part of the task you have to do.

>All photographs or videos you have seen have been placed there for you to see them

The source of most of the videos from both sides is random social media users.

Even the videos and info from the IDF I would regard as credible, since they released similar videos and info from the Lebanon operation last year that was consistently corroborated by evidence from social media (there was no internet blackout in Lebanon so every IDF strike on an urban area had multiple videos from different perspectives).

The network Assaf got from founding and selling a big cybersecurity company and then being a VP-equivalent at Microsoft for 5 years (immediately before found Wiz) is more relevant than what he had from being an IC in the army 15 before that.

The purpose is to injure enemy combatants

For public companies you don't need a heuristic, as the balance sheet is included in quarterly earnings reports.

The price of the share stays the same but there are fewer shares, so the market cap would go down.

I saw a comment (not in the linked thread) that notes that the system prompt probably includes something along the lines of "You are a helpful AI assistant...", which does feel like a bit of a giveaway in the sense that there are many documents on the internet that discuss testing AI assistants and therefore a document who's topic is an AI assistant is likely to contain discussion of testing the assistant (and if the document is written from a first person perspective, of the assistant, then it's likely to contain text from the perspective of the assistant discussing it's own testing).

The secret challenge exists and it is the phone number / email address / VC account of CFO. If CFO wants to order EMPLOYEE to send money, then EMPLOYEE should only do the action after making an outgoing call to CFO.

This is a weird objection because if the expensive insulin did not exist (effective price = $inf) it would not apply.

This law was passed specifically to allow Aryeh Dery to ignore the terms of his plea agreement and serve as a minister. It's not related to the Jewish-Arab demographic balance (which, within the 67 borders, is about 85-15).

1. There seems to be some suggestion that X

2. ???

3. In no world can I imagine ¬X

launch scrubbed

Double quotes hasn't worked in a long time. It used to be the case that if you search "foo" "bar" it would only return documents that had the words "foo" and "bar".

I will frequently search for z foo bar z, get documents that have synonyms of foo and bar (and the synonyms will be highlighted in the snippet), then I will search "foo" "bar" and get the same documents except now it doesn't highlight the synonyms in the snippet. Maybe I am in some awful cohort where this feature is intentionally broken to see if it reduces engagement.

If you ask ChatGPT to compose a layoff email with a quote from MLK that is the quote it uses (ref https://twitter.com/mattstratton/status/1618018859763798016)

I don't understand the article:

Posted October 07, 2022

Subtitle:

  An IT outage hit the airline's check-in and boarding systems \*last month\*

First sentence:

  Aer Lingus is seeking compensation after an outage to its IT systems led the airline to cancel more than 60 flights \*last night\*.

Which is it? Are there two separate incidents?

Then later:

  The flight cancellations occurred on September 10

Given where the markets are today, Twitter is a ~$30B write down for him once the deal closes, so if he could pay $1B to make it go away he would have done so already.

The difference is between something obvious and uninteresting (equity comp goes down when the value of the equity goes down) and something interesting and untrue (tech salaries are crashing). I get that post writer needs a hook in case they have something interesting to say about the equity comp situation (having read the article I don't think this is the case), but lying in the title to trick people into reading your article is still bad.

Why is using a token better than using money, in this case?