the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor)

Was OpenAI sitting on the zero day, or did the model figure it out during its escapades? Can the model sit on a zero day on its own now?

I’ve seen a few news stories through the years where I thought the stock movement was out of whack with the underlying security, and options are a way to “bet” on that being right. It’s cheap and easy to just watch. Almost every time I was wrong and only found out a few times why. That probably just means I’m bad at it.

A few more things to remember. If you look at the first link at the bottom of my post, there are ways to structure buying/selling the calls/puts so that you can have a strategy that pays off if the stock goes up, down, or stays the same. How to set those up and the math involved is beyond me. The other thing is that for every trade with options someone makes there’s an entity on the other side of it. You might be buying a part of the leg of someone else’s more complex trade where a big investment house is trying to manage their risk and keep it below a certain level.

https://www.cnn.com/2020/06/19/business/robinhood-suicide-alex-kearns/

Stay smart and safe, invest money you can afford to lose in products you understand.

Shorting a stock is a bad plan. You typically need collateral (cash, or a line of credit, or something) in case the short goes bad. Shorting a stock can have a theoretically infinite loss since if you buy a share of a company and it goes bankrupt, your maximum loss is the value of the share. Short selling means you’ve sold a share you don’t own with a pinky swear that you’ll buy one back to replace it. A stock likely won’t, but could, go infinitely high. In that case you’d have infinite losses.

The other way is options trading. You can do this with your normal broker, possibly. You’ll need to make statements about your income level, savings, risk tolerance, and check a box saying you’ve read some mandated documentation. The process can seem like click-through stuff for a normal website, but it’s absolving the brokerage of fault if your decisions go bad.

Options work a little like an insurance contract. If you are worried that you’re too heavily tilted towards a certain company (but still believe in the investment), you can buy “puts” that help mitigate the risk if the underlying security. A “call” is a contract that lets you hedge the other way. They let you get exposure to a security without having to purchase the shares. An options contract is a contract that gives you the right, but not the obligation to buy (call) or sell (put) some number (typically 100) of an underlying security by a certain date. Selling options naked is a topic better left for someone else to explain all the risks involved (and I covered a strategy that has infinite risk).

A call pays off if the stock shoots up before the contract expires, and a put pays off if the stock tanks. To do what you’re saying you’d have to buy puts (I’m not an investment advisor, I’m not your investment advisor, this is not investment advice). If it sounds like gambling that’s because it is. Everybody already had the idea to bet against SPCX. There are platforms out there that let you look and see without having to risk money, and that’s probably the best course of action.

• https://finance.yahoo.com/quote/SPCX/options/?date=1789689600&straddle=true

I picked the date September 18, 2026 at random. If you had a September 18 130P, then it went up 21.19% Friday, kudos. The “last price” field is PER SHARE, and the contract was probably for 100 shares, so that’s $2030, meaning you had four figures on the line there. If none of that makes sense, if the activity in the volume field doesn’t make sense, then it’s better to sit it out. In that realm you’re swapping your money for that contract, and someone else is swapping that contract for your (now their) money. If the contract expires out of the money, you lose.

Some resources:

• https://optionalpha.com/options-strategies
• https://archive.org/details/optionsasstrateg0000mcmi
• https://moneybliss.org/recover-from-financial-disaster/

i think Redditors coordinated to buy shares or something so that it won’t go to bankruptcy or something like that. I could be wrong though.

Close enough. DeepFuckingValue was either the one that caught on to or was the public face of a concern that noticed that GME was extremely short sold which roughly means that they sold shares they didn’t own with the notion that the share price would drop. Then they could buy the shares needed to cover the ones they shorted and make money. Selling short can have theoretically infinite losses, btw. The internet noticed…

So people flooded in and bought shares, call options, whatever, with the thought that the stock price would go up because the short sellers likely took out loans to cover shorting the shares, and the people that loaned them money don’t find cases of infinite losses/risk nearly as funny as the old wallstreetbets.

Some brokerages halted buying GME, and DeepFuckingValue got to testify in front of congress.

When reached for comment the family said, “how could this happen to us? We’re not even trans?”

Have the day you voted for, know that you won’t be the last people in rural areas that this happens to, and that when the AI bubble pops you still won’t get the land back. It’s gone forever.

They typically condemn the land before using eminent domain.

The adult entertainers are the VC investors. They’re pretty world-wise, but can’t be well versed on everything. So when someone sells them on something that sounds pretty good, they bite. The CEOs are the bros laughing about how great everything is, except in real life they don’t have consequences. All the CEOs get paid like it might be their last job so if they never work again they’re still fine.

It’s still impressive to see what the LLMs cook up when asked about programming problems. I’m coming back to programming from some time away from it, and it’ll give you the answer to the question you asked. If you ask it for an old way of doing something, it’ll tell you that. Then it slips and shows you a new way of doing something (I’m specifically talking about std::cout versus std::format and std::print), and the doors are wide open all of a sudden.

Then it gives you a technique for something and you spend hours debugging code only for the LLM to say that the solution it provided won’t work.

Prompt engineering is going to be a real thing whether we like it or not.

I know :-)

Here’s hoping for more red today.

Replying to myself to put up a link to the jenga tower scene of The Big Short in case nobody has seen it: https://www.youtube.com/watch?v=xbiDrzTd8fE

It’s there in the S&P 500 between MSFT and PLTR on the left kind of in the middle (size of the box is the market cap of the company). It’s in practically every 401(k) in the US. BBB is somewhere in the middle of the jenga tower.

You shouldn’t forget about selective enforcement.

Ft. Knox

Did anyone ever go in there and show the gold? I thought someone campaigned with that as a promise.

it would be able to detect

“It” doesn’t have to be any good at it:

https://blog.princelaw.com/2009/07/08/nfa-and-constructive-possession-myth-or-reality/

Most still will. Like I’m sure a lot of people are doing, I was trying to reuse old hardware for a new purpose. Perfectly good computer with 16GB of RAM with an AMD A8-3850. I’m not complaining about progress’s march towards the future, but I missed the warning signs about the changes. I’m sure some other folks probably did as well.

I’m going to drag out my same soapbox: a lot of systems old enough to use DDR3 RAM will have x86_64 v1 or v2 processors. Some projects have already removed support for those, the big one being the RHEL kernel as of RHEL9.

Replying to myself here and including a link that just dropped:

• https://lemmy.world/post/48881740
• https://nsa.2026.action.cr.yp.to/
• https://blog.cr.yp.to/20260405-votes.html

Apparently the debate was more spirited than I thought. The argument appears to revolve around whether it’s OK to jump to the new stuff directly, or use a combination of the old and new.

I think this is how I can message people…

• @DevDave@piefed.social
• @monkeyFromTheLake@programming.dev

reddit cant afford [the V3 captcha system] but google lets them use it in exchange for AI/datamining

Had no idea they used that. I edited all my comments to crap then deleted them around the time the admin monkied with the backend database, and stopped using old.reddit to browse once I found lemmy. I once went through the effort of making a temp account to comment on someone else’s comment there because they had suggested trying something specifically dangerous and didn’t seem to know about it. I doublechecked later and the comment I wrote was caught in some filter, likely the result of the account being too new. I can’t imagine what garbage that site will be in the years to come.

abusive scraping

As opposed to the plain old scraping they do to train AI, and generate revenue by selling user comments for others to train AI.

I read a half-cocked internet theory that a certain someone might’ve purchased twitter just to gain access to an ex-gf’s personal tweets. I judged it as possible but unlikely, as that’s a lot of money to spend on such a thing.

Now, we’ve all heard stories about reddit blocking accounts for no published reason, and tracking folks down across accounts/IP addresses/etc. That code must be pretty expansive to do the things they’ve done. So one has the thought: if you’ve ever reached out to the reddit hive mind for some kind of support with a personal issue of any kind then that data about you is still floating around in their database and tied to whatever alternate accounts you have, even if it was the “good old days” when you did it.

Abusive scraping, my ass.

There are four version of x86_64: v1, v2, v3, and v4.

RHEL 9 dropped support for anything prior to v3. That means RockyLinux doesn’t cover it, either. AlmaLinux has support for v2 in version 10, but there’s no way of knowing how long that will last.

Some binary packages are starting to drop support for earlier version. The latest numpy out of pip will not work on a v1 machine. You can sometimes use the system package manager’s numpy to work around it, or constrain pip to use an older numpy. I don’t know what else is lurking out there.

If you’ve got visions of taking a really old computer that you happened to max out on RAM back in the day and bringing it back to life there are surprises waiting for you.

They’ve been beaten on enough that industry is moving forward. The advantage of knowing and being able to prove that the algorithms are insecure would be incalcuable, so groups who want to be able to break into systems aren’t going to volunteer the information. It’s to the benefit of everyone else that the algorithms be secure. The third section of that paper I linked does a pretty good job explaining the why and why now.

tl;dr: Smart people have dug into it, and we know what we’re going to know for now.

• https://csrc.nist.gov/projects/post-quantum-cryptography
• https://media.defense.gov/2025/May/30/2003728741/-1/-1/0/CSA_CNSA_2.0_ALGORITHMS.PDF