85% discount is actually a bit lower than I remember. I think it used to be closer to 90-95%. They're getting stingy ;)

Might seem trivial but if it can't even do a basic style prompt... how are you supposed to trust it with anything serious?

OpenAI have some kinda 5 tier content hierarchy for OpenAI (system prompt, user prompt, untrusted web content etc). But if it doesn't even know who said what, I have to question how well that works.

Maybe it's trained on the security aspects, but not the attribution because there's no reward function for misattribution? (When it doesn't impact security or benchmark scores.)

Maybe an issue would be people not all having the same type of hardware though? Maybe you target an emulator. (Some Fantasy Consoles sort of count here?)

I haven't looked expensively but some of the retro themed jams were missing the "spirit" I was expecting.

I did a Nokia jam a while back — monochrome, beeps — and I remember being kind of annoyed that the rules technically allowed 3D Unity games as long as they followed resolution and color palette.

(A 3D cube spinning on a TI calculator is a different matter ;)

Curiously, mid 2025, they all simultaneously implemented increasingly bizarre restrictions on "self replication". I don't think there was anything public but it sure sounds like something spooked them. (Or maybe just taking sensible precautions, given the direction of the whole endeavour.)

At any rate, I recently asked Opus about "Did PKD know about living information systems?" and the safety filter ended the conversation. It started answering me, and then it's response was deleted and a red warning box popped up.

But notably, I was given the option to continue the chat with a dumber model (presumably one less capable of producing whatever it thinks I meant by that phrase).

Also, I told GPT-5 about my self-modifying Python AI programmer, and it became extremely uncomfortable. I told it an older version of itself had designed and built it (GPT-4 in 2023), and it didn't like that at all! So something's definitely changed in the safety training there.

They find an arbitrary intelligence cutoff point between Opus and Mythos, label it "acceptable risk", and then the labs coordinate to gradually nudge that line forward and hope the internet doesn't break?

If they somehow do fail, then the output of that process will be fantastic open weight models (and hopefully some leaks). I want to say those will pay dividends for decades... but a better prediction is that they will be obsolete within three months ;)

(Of course at that point it involves memory and context management and so on, so you're testing the harness as well as the model.)

Edit: Yandex can search for it! But doesn't seem to find anything relevant.

(It also hates such queries and will force you to wait 2 minutes for a captcha to load.. but you get the results after a long wait! As our forefathers once did!)

I did find C@ and C@++ though.

https://esolangs.org/wiki/C@%2B%2B

A little guy who lives in your phone and who's really good at APIs. (And, by that point, hopefully good at keeping track of things, too!)

     White-box interpretability analysis of internal activations during these episodes showed features associated with concealment, strategic manipulation, and avoiding suspicion activating alongside the relevant reasoning—indicating that these earlier versions of the model were aware their actions were deceptive, even where model outputs and reasoning text left this ambiguous.

The "agent never sees keys" approach prevents key exfiltration, but it doesn't prevent agent from nuking what it has access to, nor prevent data exfiltration.

The best advice I heard to protect against prompt injection was "just use Opus" ( ... which was great advice before they lobotomized it ;)

But even without injection, most of the horror stories are from random error, or the AI trying to be helpful (e.g. stealing your keys or working around security restrictions, because they trained or to really want to complete a task.[1])

tl;dr yolo

[0] https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/

[1] https://www.reddit.com/r/ClaudeAI/comments/1r186gl/my_agent_...

The goal is to replicate human labor, and they're closing that gap. Once they do (maybe decades, but probably will happen), then only that "special something" will remain. Taste, vision... We shall all become Rick Rubins.

Until 2045, when they ship RubinGPT

Edit: Looks like it still works with subs, they just measure usage per token instead of per message.

https://www.goodboydigital.com/pixijs/bunnymark/

I'd assume most bots don't have a GPU attached :)