trunnell

It'd be great to see a quick comparison to firecracker rather than to a non-hypervisor linux system. The homepage and blog post only compare with the latter.

> 3. Define the problem

If you do this step well, in my experience, the next steps fall into place quickly.

My version: accurately naming the problem is the essence of problem solving. There are things you usually need to do before you can accurately name the problem, because most problems aren’t presented to you in textbook form.

Once properly defined and framed, it’s obvious how to solve most problems, most of the time.

If you’ve ever worked with someone amazingly good at accurately naming the problem, it’s hard to unsee it. This is one of the hidden talents of the most effective people I’ve met.

I can tell you don't know anyone at Anthropic.

Watch for ChatGPT to follow soon.

I wonder how related this is to the social media legal settlement. It wouldn't be surprising if they weighed the potential liability of serving minors and decided it wasn't worth the risk, at least for now.

Also, cheating in high school is a really big problem, so this might help a little.

I see, you'd prefer to continue the prior experiment of humans flying around corners in multi-thousand-pound vehicles, over the speed limit, while sipping coffee and looking at the radio?

I dislike bots as much as anyone else... when weird inquiries come through my company's lead form, it costs some time and attention to sort them.

But what makes Cloudflare so confident that automation always equates to "fraud and abuse?" If I send my agent to go retrieve some information, do they consider that fraud?

If I block various ad trackers does that trigger their "bot detection" incorrectly? Do I have any recourse? Or is Cloudflare appointing themselves judge, jury and executioner?

And let's not forget this little chestnut: > 4. Privacy by design. Precursor was designed to collect signals that help to distinguish human patterns from automated and abusive patterns.

Ahh, so to "protect" against bots they're standing up a whole new regime of user surveillance and session-level monitoring. And they definitely won't be selling that, they promise. Got it.

This crap should be illegal. In the real world, I can authorize others to act on my behalf. The same should be true with software agents.

Blame Amazon and the White House

There is no reason to have less trust in Anthropic. It's not clear they did anything wrong. It's more likely the White House simply tied itself in knots, consistent with the last year and a half of chaos from them.

Hooray! Glad everyone came to their senses and we can all get on with business.

I bet it'll continue to be messy at the frontier for the foreseeable future as society gradually wakes up to the consequences of strong AI.

> honest angle is that the industry [felt] exposed by the residual risk [and] have a months-long bugfixing backlog exposed by Glasswing

Two problems with this theory.

1. Amazon complaining to the White House wouldn't have been the opening salvo. Amazon and Anthropic would find it much easier to talk to each other than go through the White House. We'd need evidence that Amazon (and probably others) already asked Anthropic to not release a Mythos-class model but Anthropic released it anyway. Are they on record saying this?

2. The jailbreak Amazon found needs to be real. Maybe the White House staffers are not AI experts and they don't really understand what a jailbreak is... but it's much harder to make that claim about Andy Jassy. For the jailbreak to be the real reason for the export control order, the jailbreak would need to be significant and cause material harm to Amazon. Then Jassy might pass it along to the White House assuming he already was refused by Dario.

But there is no evidence the jailbreak was real. There is one story that it amounted to a request, "fix this code." In any case, Anthropic is on record saying the so-called jailbreak didn't enable any vulnerability work that couldn't already be done by other models.

Sure they exist but they're rare, difficult to identify and hire, and take years to train. Mythos/Fable is available on tap.

What an amazing achievement by America's adversaries.

The Trump administration lists Anthropic as a security risk and kneecaps its best model, despite the fact that compared to the other frontier US labs Anthropic is more transparent, more safety-oriented, frequently honest to a fault, and is clearly acting with patriotic intent.

Meanwhile, the same administration is hesitating to counter certain Chinese companies' efforts of industrial-scale theft and sabotage due to a fear of angering the CCP!

This administration has it exactly backwards. 4.5 months until election day, 7 months until the next Congress is sworn in.