I considered it, but I figured that whatever I did there would be inconclusive. Instead I tried to figure out the blast radius of this being proven, and I didn’t get very far with that either.
For comedy’s sake, I asked ChatGPT 5.5 about the significance of the problem and the chance that 5.6 would solve it with a three page solution. It said close to zero.
I invited it to search the internet and it remains extremely sceptical.
This is actually great, and I predict that fans of nil-punning will rapidly discover the joys of actually having errors trigger where the error was introduced rather than propagating through the program.
This is one of those little things I’ve had trouble putting my finger on: the US eats surprisingly few oats. The U.K. eats more than three times as much on average. Which is probably one of the reasons I find US cuisine slightly uncanny valley.
Southern Europe doesn’t really consume much either, but most US food is closer to Northern European food.
I’m not a follower either, but inference from the article tells me: Krammik has come up with some sort of cheat detection method, has loudly accused some others of cheating, and FIDE are both unconvinced of the truth of the allegations and very unhappy he didn’t do this through proper channels. He also doesn’t appear to have co-operated with the investigation.
I suspect this is a standard mathematical “it is computationally impossible to do this in the general case despite it being entirely feasible in many cases”.
I don’t think anyone in the business thinks that the markets are 100% efficient, just that they are sufficiently efficient that beating them is a genuinely hard job requiring heavy, expensive analysis.
The language used in this press release is borderline hilarious. It’s simultaneously trying to tell you how great it is while also telling it’s not THAT great. Nothing to worry about, move along.
I feel like "Company ditches staff in favour of AI" stories currently fit into two categories 1) The CEO is actually ditching staff for other reasons like falling revenue, but "going AI first" sounds a lot better 2) The CEO is making a mistake.
I'm thinking it's rapidly not becoming "can you find a security issue in XYZ?" and "what is the cost of finding a security issue in XYZ?". I want to know what the spend was.
That's both extremely neat and, for the time being, extremely hard for me to get my head around. I got it round monads, so I imagine it's just a matter of time!
It's not just for choice of model, you can use it for your prompting as well (basically anything to do with your setup). And yes, running evals is expensive and mostly of use to people with serious spend.
US controls on cryptography software lasted _20 years_. If there's something I'm absolutely certain of, and I'm certain of very little in the fields of AI and of politics, it's that Fable will be utterly irrelevant in 20 years time.
Look, I’m not an AI hater, but AI is… not great at multi-threading code. And having it analyse multi-threaded code proves nothing because… it’s not good at multi-threaded code. This isn’t entirely shocking because I’m not good at it either and need to write in some very particular ways to have even a hope of being correct. But basically, unless it was written by a genuine expert, I wouldn’t want to even glance at this PR. And it wasn’t.
I mean, I appreciate the huge engineering and design challenges here, but C# had non-null value types in 2002. It had generic value types in 2005 and it gained not-null “types with identity” in 2019 and no-one has batted an eyelid. (Indeed the type system support for value types still includes stuff Java isn’t even considering yet.)
Saying the mental model is too hard is basically saying your userbase is stupid. This stuff is not tricky.