It's negligence.

If you were an IT Director, or even an IT Technician with the ear of the Founder/CEO, and you had the budget for internal network and security ops, what would you do?

Not give it unbrokered access to other models and infrastructure which it could use to puddle jump elsewhere. Not put Skynet on the Internet directly, is my bet.

That hardly sounds like a controlled release environment, to say the least.

They seriously need to hire some criminal psychologists and actual hackers to help them figure their shit out. Same goes for Anthropic.

When has someone's computer ever hacked someone else's computer ?

Not ever, before recently.

Clearly it's not something they can objectively say is an outlier - they don't have enough data. In fact, you can most certainly say the opposite, due to how frequently it's happening now.

Not a serious or honest company. (Oh, sorry - "nonprofit".)

Everything I've seen simply suggests they're bad at computers.

My experience with startups is that the things that don't make money or make customers happy don't get any attention. Network/system security in depth is not one of those things. Sure, they've got a bunch of psi-brain neuroscientists and mathemagicians, and goodness knows lots of DEI-enforcers for alignment, but their repeated dumb mistakes in this department tells me they're not really hiring too many experienced ops, nevermind devops, types.

It's true for OpenAI and Anthropic both.

In the past YEAR they have:
- accidentally published the full Claude Code source (and then effectively... did nothing).
- Numerous Claude Code RCE/API-key vulnerabilities which could have been mitigated with at least one person on their core infra team aware of injection attacks.
- Publicly exposed their internal CMS (and files) to the Internet
- Had their training escape numerous times and hack other orgs
- hacked HF

They're not a serious company. Some are to be expected given their visibility and exposure, but you'd expect a bit more from a company implementing Skynet-level capabilities. However you slice it, they're fundamentally building weapons which can write poetry. A little more seriousness should be involved.

Someone needs to end this crazy tyrant's empire, though I suppose it's too late for that. He's so diversified at this point that ending Microsoft won't materially impact him.

That said, it'd be great if the thing he built could be ended - perhaps by reimplementing Windows functionality through reverse engineering, minus the crap.

Just enough to get it to work with the things we still need to run on Windows - then replace the kernel.

While I can see how you arrived at that, I don't think they're all the amoral actors you describe.

Zuck has shown some changes in the last couple years, though I agree he's largely amoral.

Musk has a very peculiar set of morals which might be described as pragmatism or amorality, but he's always been very focal in both action and word about both ethics and responsibility.

You could probably accurately describe them all as transhumanists with varying degree of consistency.

If there's a crisis of trust - which there is - it's almost certainly largely provoked by Amodei. The guy has been a regular prophet of doom regarding AI: it's going to take all the jobs, it's dangerous and going to cause all sorts of disasterous side effects socially, economically, and politically. No wonder people are untrusty.

Pair that with the obscene levels of what appears to be state-level anti-AI and anti-datacenter propaganda, and you've got the current situation. The amount of absurd claims I've seen about datacenters and AI, from people who should know better - everything from completely false narratives about water use, land grabs, power use, and so on. The truth is, an AI DC doesn't uses less water than a Cosco and about as much electricity as a comparable industrial facility, or a similarly sized restaurant.

For anyone who thinks it's ridiculous that the US can't reproduce the Apollo missions, consider this: we can't (due to environmental factors) produce wrought iron anymore (because there are no longer any facilities setup to do so), as a species, and therefore, could not build the Bay Bridge in San Francisco (and I'm sure, any number of others, like the Brooklyn bridge).

The same limitations apply here. They have to redesign the vehicle on first principles because the foundational engineering (not scientific understanding) being used no longer exists. Most people simply do not realize how precariously our technological world hangs in the balance. It's why highly advanced ancient cultures collapsed, in large degree: they couldn't just start at the lower level again before working their way back to 'advanced civilization', they had to start all the way at zero again.

Here's a fun one I like playing with my kids, kind of the nerdy science version of the polysci "go back in time to kill Baby Hitler: which one scientific understanding would you bring back in time to Ancient Rome/the American Revolution/the 1950s/etc. and why? Dollars to donuts they'll say something like "the lightbulb" and not consider that it's at least 3 distinct discoveries and inventions, nevermind the engineering.

You can get this kind of basic image recognition on $20 ARM boards from China. I've got a rock3a (older, barely capable compared to whats available now) board cluster that does this kind of image recognition for my cameras. rpi3 level stuff. That it was nvidia just speaks to what is available.

Sure, let's discuss that. You've got a handful of methods historically:

- GPS/GLONASS/Starlink - can be jammed.
- Wire - inaccurate, not suitable for OTH guided missiles
- Pigeon - inaccurate and expensive
- Radar - can be jammed
- IR/laser - line of sight/requires IR signatures
- inertia - inaccurate, expensive.

There's nothing which really compares to "AI" in capability or accuracy. It's the only option which can't be jammed directly short of active EMP and provides the required/desired accuracy.

That's kind of been the bread and butter for cloud providers - "data velocity". They say: keep all your stuff on our cloud, it'll be safe forever, you don't need to worry about it.

A single copy is zero copies and 2 is 1. This is overlooked on all clouds, on the basis that 3 copies on a cloud provider is still 1 copy.

This kind of oversight failure has ended many companies when they've had DR scenarios. I know of 2 that had to close shop because the egress cost of all their data from AWS was larger than insurance would pay out, and the cash difference was significant.

In most cases, it makes more sense to keep 2 local copies (different storage) than 2 cloud copies, for this reason alone.

Even if frontier models only continue to get 'marginally better' in terms of evaluations, as they have been for the past year, it's quite apparent that the newer models tend to be significantly better in aggregate and practice.

We're also starting to see a stratification across frontier models:

- Anthropic seems focused on "one shot" capabilities (latest generations have suffered significantly in other departments)
- GLM seems to be increasing their focus on security and coding robustness
- Google seems to be leaning in on 'fast and cheap' and 'good enough'
- Agnes (while not quite frontier) seems to be going for "broad scope of capabilities"
- LiquidAI is focusing on specialized, small local models for agentic pipeline work
- OpenAI seems stratified on do-everything frontier models with functional stratification based on task complexity
- Qwen - local capability focus.
- Grok - code centric + fast.
- Deepseek - also local-capability focus while trying to optimize the full stack

Then you've got the mess of models like Laguna (last generation capabilities but local-focus) and Agnes (more like google's approach, with a wide and cheap portfolio).

As for Anthropic (and , it is quite apparent that they're architecturally behind in many ways to other models. The release of dflash papers and the subsequent speed increase on the competitor models made that plainly obvious.

The problem is that AI has made it very obvious that teachers have not been doing their jobs.

For instance: one history test, one report. Two different responses: "Contrast the political sentiments at the time of the American Revolution with modern day American sentiments and political events." 1 page report.

You could just paste that into chat and get a useable if not undetectable result. The result is obvious. Instead, a curious student could use the tool to pose a more complete question to chat: take into mind more emotional similarities between different political factions, as well as the contrast amongst social classes. It'll help them stand out a bit and it'll be different, and they might learn something.

But, I do note: this would require a curious student. Teachers have all but punished curiosity and a love for education for decades, catering to the dumb ones to keep them engaged instead of punishing them for wanting to be dumb.

That's the dumbest thing I've ever heard.

AI has only accelerated my desire for learning. I can learn so much more (though perhaps not as thoroughly or as deeply, because you don't have to go below the architectural surface as much anymore) now than I could before: much less switching between tabs, copy/pasta'ing, digging for documentation (and reading a lot of things that I don't remember 3 seconds later because I'm looking for something specific).

The same goes for reading about history, or anything else. Curiosity is rewarded rapidly with AI.

No, the biggest problem with education in the schools is the schools and how they teach. They've made it un-interesting and curiosity is generally punished. They have completely destroyed education long before AI was even a question - I'd argue shortly after the Internet became available and approachable for your typical 100-IQ public school teacher, in the late 90s.

"3 dimensional thinking and problem solving is something you can learn to be good at"

It really isn't, not to a meaningful degree. You can sharpen cast iron, too, but that doesn't mean you'll have a sharp knife.

Spatial thinking (which is heavily linked to problem solving) is fairly tightly coupled to inborn intellect and early education capability attainment.