That's not good. Time to raise the package security draw bridge on vibe coders.

In what world does professional hackers intersect with vibe coding? This is a professional attack. Not some amateur script kiddie action.

"Thinking" is about to get even harder to do for most grifters with newsletters to sell.

This is not an outage, Claude just gets lazier on Fridays.

Sometimes Claude wants more lunch breaks, takes a half day and leaves the desk early just like any human would. (since AI boosters like comparing LLMs to humans all the time) /s

If you're concerned about humans anthropomorphizing AI models, you might want to steer well clear of Anthropic, as their entire positioning (starting with the product name and continuing with UX choices and model releases) is built to attract the kind of researchers who are prone to believe in sentient machines.

They are going in the "Claude is alive" direction already and that line of communication is likely going full throttle in the nearby future.

I suspect the next big marketing gimmick is this supposed leak about capybara. I suspect the leak is intentional and meant to influence their expected IPO.

I think the big reveal is going to be that frontier models are no better than the open source models that you could feasibly run on retail hardware however they have a highly complex harness behind the API where the magic is.

I had my agent set up a "team" of subagents directed to different parts of a big new app (UX Engineer, test lead, etc) . Apparently the Senior SWE had reduced the scope, and my PM came to me trying to argue the side of the SWE that had reduced the scope for time constraint reasons...

It went a bit too deep into the role-playing bit.

You joke, but I think that's a fair summary of why people don't mind one 9 of uptime in a key component of their development workflow.

Tell that to those who are still using Electron, TypeScript to create bloated desktop apps.

Claude is on a lunch break just like a human would.

Or an AI company acting as a casino with offers, boosted usage and limits to get you addicted into spending more tokens + credits on their digital slot machine which accepts 'tokens' as an exchange for 'promises of productivity' and 'intelligence'.

If the output doesn't work, roll the dice again and spend more tokens.

Reminds me of the US healthcare system.

There you go. So when Azure has an outage, so will Anthropic (and Github).

Now expect both of them to have unstable uptime and outages every week.

You don’t thino Anthropic has any kind of resiliency, so net new compute reduces downtime? Any docs on that view?

This is what "AGI" will bring more of. An "abundance" of gamblers.

Until next time.

Where is the system card for this model?

here: https://deepmind.google/models/model-cards/gemini-3-1-flash-...

also deepmind page: https://deepmind.google/models/gemini-audio/