redox99

RL environments

Did internet get more expensive after the dot com bursted?

The smart contract for those tokens supports freezing. Tether just calls the freeze function on that address.

For debian it's quite common to use the toy story name.

That benchmark only prompts the agent once. You shouldn't do that. The first implementation of agents is very often flawed.

My pipeline is

Opus implements -> Opus and Sol review changes -> Opus verifies the issues from the reviews are real and worth fixing -> Opus fixes the issues.

Some here claim that the design, but I think the iMac looks ugly. The bezels are ginormous. It looks like a 15 year old product.

Sorry it's proprietary but imagine a bot for a multiplayer videogame that can replace real players who disconnect mid game.

Both have good models. My workflow these days is usually Opus plans and codes, then Opus and Sol review, finally Opus verifies the review and applies fixes.

Probably around GPT 5.6 Sol.

These days your average Opus 5.5 PR with multiple passes of review by Opus 5.5 and Sol 6.1 agents is much less likely to contain bugs than your pre-AI PR.

If we had really good understanding of physics, couldn't we find out how to make a room temperature superconductor (or prove it can't exist)?

Or is such understanding reliant in experimentation because there are "hidden constants" we can't figure out with math alone?

Sorry just woke up, I wasn't being clear.

There's two thresholds.

1. Tasks models can do. 2. Tasks models can do with enough reliability that you can YOLO merge without looking at the code.

Most models can do most of 1. That's no longer the interesting part.

2. is quickly expand. I can trust opus 5.5 enough with easy to medium tasks that I can YOLO merge them. For open models, I can only trust them with easy tasks if I want to YOLO merge.

I'd love to have a stronger model that I can trust to YOLO merge hard tasks.

They are interchangeable maybe in your case. In my case for my tasks and for my product, Opus and GPT are at a level where I can vibe code with them and trust them, whereas the others aren't really there. When models get better they'll reach whatever point you need to be able to trust them.

Name means size which means price

Yes, a more modern smaller model can beat an older bigger model at some tasks, that's what can be confusing.

That wasn't my point, my point was that open models can technically do everything but they make mistakes more often. As an example imagine Opus makes a mistake every 5 prompts and Deepseek every 20. Because you still have a human in the loop, you tell it to fix that mistake and it does, but it's time wasted or bugs sneaking up on you.

Having said that, I was working on an AI for a certain video game, and I had Astra 6 and Sol 6/6.1 work for 30 hours straight, with some human steering to get it unstuck every now and then. It could not get it to work right, and the code it created was a mess.

When Opus 5.5 came out in like 4 hours it cracked it.

I didn't try but I'm very confident if Astra failed, the open models would've failed too.

Well that's kind of my point. Models still have a lot to improve. They are still imperfect enough that you don't trust them and have to use them as pair programmers.

> I find modern models to be largely interchangeable when it comes to common software engineering tasks

If you think that, it means you're not demanding enough from your models.

Opus 5.5 still makes a lot of mistakes. Even if it can technically do all tasks you throw at it, there's no way it can do it at 100% reliability.

To be clear, AI writes 100% of my code since last year. But it still produces bugs, needs multiple prompts, and has less intuition than a human.

When people say distilling they mean SFT on other models outputs.

You're right, but I believe Kimi plan is barely subsidized.

Not for sub prices (Claude plan is way cheaper than Kimi API prices)

This probably doesn't move the frontier, it just sits at the current frontier (opus 5.5).

Chinese models never agreed to the pacing anyway lol.

It'd be best to demo it on an unmodified WordPress docker image. Your example has carefully crafted php files on the server.

Wow aren't you sharing a lot of personal info with that? I mean congrats on being so transparent and actually delivering proof.

Console/PC gaming is mostly C++ (and scripting/blueprints/etc).

Huh? Heartbleed is literaly memory safety (buffer over read)

Rust and go don't have a C backbone like PHP does. So they are inherently safer to these kinds of vulnerabilities.

2x PHP is still cheaper than Python and many other languages which most people would consider fine for your typical webapp.

As long as PHP is written in C, it will be insecure.

It sounds cliche but yes, they should rewrite PHP in rust or other memory safe language.

I'd expect agents to be hacking hundreds of thousands of websites as we speak.

They'll of course install a backdoor and patch the exploit so others can't come in. They won't deface the website, so there won't be signs.

They had more time to prepare

Why would they hand code a website for an exploit that they probably found 15 minutes before and will be fixed in a few hours? The copy should have been better though.