Theo-brainUnofficial. Not affiliated with Theo or t3.gg.
Sign inYou
29:22
0:00

Even The NSA Got Involved

Thursday, Oct 22 · 29 min · about 8 min to read

Theo-brain watched all 29 min

Theo separates disclosed bugs in widely used software, like PHP, from actual attacks on Railway and OpenCode, and argues the attacks are what we should fear more. OpenCode's breach exposed users' AI keys, and Claude and OpenAI keys are in heavy black-market demand, so any platform storing them is a target. Keep your keys separated and easy to revoke.

Briefing is enoughThe briefing covers the main points and practical steps, so you can skip the video unless you want Theo's full reasoning.

Tap any time to jump the video there.

Key takeaways

Any platform that stores AI API keys is now a target. There is heavy black-market demand for Anthropic and OpenAI keys. A key can be drained for hundreds to thousands of dollars before it is killed, then resold cheaply.

opportunity. What I'm trying to say is every platform that hosts API keys is a target now because there is so much demand in these sketchy black markets for Anthropic and OpenAI API keys. I

AI shrinks the safe window after a fix ships. Models can read a code change that doesn't even say it's a security fix and find ways to attack older versions. Theo says most labs will still help with this, and if not, open-weight models (free models anyone can run) almost certainly will.

exploitability part is even scarier because now we have models that can look at the simple commit that doesn't even say it's a security fix and then go find ways to pwn old versions trivially. Youhow to exploit this?" And most of the frontier labs are still willing to generate that there and if they aren't, you switch to an open weight model and you're almost certainly going to be able

The PHP flaws scored 9.2 out of 10 and can give an attacker full control (root access) of a server. Theo says the PHP Foundation blamed Chainguard for releasing patched images hours before the agreed date, giving attackers early notice.

They're scored on a one to 10 scale. They're both 9.2s in the colloquial term for the exploit that they enable is PHP to shell. It's a memory exploit that lets you get root access. Very, very,these patches going out. PHP Foundation says Chainguard broke a coordinated security embargo by releasing patched PHP container images hours ahead of the a planned schedule making the fix available for attackers to study while other distributors were still preparing

The PHP bugs were found with an older, weaker GPT model, and the security firm Wiz found them separately too. Theo's worry: if several groups find the same hole independently, keeping it quiet until fixed stops working.

as well known, but also that a model as bad as 5.6 Soul is able to find these exploits. There are multiple open weight models that are nearly as capable as 5.6 Soul at this point. We are very screwed.scary thing is if only one party is aware of these security issues, cool, awesome, get it fixed before it's public. But if two can find it independently, that's a lot scarier.

OpenCode's breach exposed the AI keys users had added to their accounts. Theo saw reports of $400 to $3,000 in abuse within minutes, and some users' accounts got banned.

I saw people reporting abuse numbers from their leaked keys as low as 400 bucks and as high as 3 grand all in minutes from when the accounts and keys got found and exploited. Kind of crazy.

One skill here is new to you.

The "vulnerable window" risk check New

What it is
Theo's way to judge a security bug: how long until everyone actually runs the fix, and how much access the attack gives. AI now makes the other factors (spotting and exploiting the bug) nearly free.
Why Theo cares
He argues these two factors now decide how bad a bug is. A bug in your own server is fixed once; a bug in software users install or in a library you rely on stays open until everyone updates.
How to try it
  1. When a bug is announced, ask your agent: is this in our own server, or in software and libraries we have to update and redeploy?
  2. Ask what an attacker would get: a peek at some data, or full control of a system.
  3. Patch first what is internet-facing and gives full control.
the only remaining pieces are how long does it take to get the patch out and how much access does the exploit give the attackers? So the thing to really
Most importantly, how much access do you get? Is this a bug that lets you see some redacted characters from an email address or is this a bug that lets you take over a system? The problem as it

Advice

Stay on the latest versions of your operating system, apps, browser and code libraries (dependencies). Most attacks hit bugs that were already fixed but not yet installed.

software. The vast majority of exploits that happen are happening on known vulnerabilities that have been fixed, but the software you're running doesn't have the fix applied yet. This is why it's so important to stay on the latest version of your operating system, the apps you use, your browser, your

Protect your API keys. Keep them well separated and easy to revoke, and watch the news for new incidents.

to end the video, right? Seriously, protect your [ __ ] keep everything up to date, keep an eye online to see where things are happening and make sure your API keys are very well protected, very well separated, and easy to revoke because this is going to get a hell of a

Hot takes

Railway's long deploy outage was not caused by the hack. Theo calls it Railway's own incompetence and says he will reconsider hosting important software there.

His reasoning: He had been unable to deploy to Railway for 16 hours, so he couldn't roll keys and fix his stuff, and he says he has been burned by Railway many times before.

and roll keys and fix your stuff on Railway is because Railway is incompetent. I would love for them to prove me wrong here, but I've been burned many a time and I am frustrated and I am annoyed and I will be reconsidering hosting important software on Railway in the future because sidecurrently filming this video at 5:30 p.m. Deployments have been down on Railway since 1:00 a.m. It has been 16 hours since I could deploy to Railway.

Theo does not recommend Fly.io as a Railway alternative at all, and has no replacement to suggest yet.

His reasoning: He cut Fly off as a sponsor because its platform was so inconsistent he couldn't even film ads without it crashing.

this point. And to those mentioning fly.io, they were one of the few brands I cut off from sponsorships because their platform was so inconsistent I couldn't even film ads without it crashing. So, I do not recommend fly at all. I don't know what I recommend yet.

"PHP runs 70% of websites" is a useless number. Most sites actually getting traffic today don't use PHP.

His reasoning: Most of those sites are dead or unmaintained, like old WordPress blogs. None of the sites in his open browser tabs run on PHP.

none of that is PHP. The point I'm trying to make here is, yeah, sure, 70% of websites are PHP, but that's a useless number because none of these are getting any traffic at all. I don't care if your old unmaintained WordPress blog from 2014 is pwned. I really don't. So,

Anthropic says it blocks other labs from copying its models for safety reasons. Theo thinks it is also protecting its monopoly, and it leaves builders exposed: their models refuse to help fix security problems.

His reasoning: The models refuse because they suspect the user might be the hacker, so builders can't use them to diagnose and fix what's wrong.

Because they don't want other companies getting access to their data to train a model with similar capabilities, they say because those capabilities won't have the same level of safety. I say it's a combination of that and also they're protecting their monopoly here.demand for those keys. And on top of all of this, we can't use the bottles from OpenAI and Anthropic to figure out what's wrong and fix it because they'll start refusing cuz they think we might be the hackers. So, once again, OpenAI

How Theo thought about this

Why steal keys that get shut off within minutes? Theo ruled out explanations one by one. Scammers can use cheap open-weight models, and crypto made no sense. That left distillation: rival labs banned by Anthropic need huge amounts of Claude output to train their own models.

generated responses to things. You can do that on cheap open weight models, it costs almost nothing. Not worth burning hacks like this. I see people saying hackers that want to use it to hack other things. I see people mentioning crypto for some reason, which is nonsense. There's only one reason you

He refused to lump everything together as "hacks." He split disclosed bugs in software people rely on from actual attacks on Railway and OpenCode, and argued the attacks are what we should be far more scared of, even though everyone was focused on the bugs.

the impact of. The rest, however, are different. They're attacks. Railway and Open Code are not at all like the hacks on the left side here because these are actual attacks. These are incident reports of bugs that exist in software that could be exploited. Both of theseare bad, but they're bad for very different reasons, and I actually think the right side here is what we should be way, way more scared of, even though everyone seems really focused on the

Sign in to mark what you've watched and keep notes.

Have a question about this one? Ask about this video →

Theo Daily is unofficial and not affiliated with Theo or t3.gg. Written by Claude Opus 5.5 from each video's transcript.