AskBeevs

14/09/2026

The AI race is so cooked that Anthropic, OpenAI and Google are privately talking about forming a safety club and slowing down.

Dario Amodei wrote the weekend essay.

Sam Altman agreed.
Elon agreed.
Google agreed.

Trump called Amodei a "perfect little angel" and threatened Anthropic with legal action.

So the same blokes who spent three years screaming AGI next Tuesday are now doing Cold War arms control cosplay.

Meanwhile Salesforce just shipped "job-ready" AI agents that chase goals for days and weeks without asking you.

Zendesk reckons specialised agents can automate up to 80% of service workflows.

Slow down the frontier models.

Speed up replacing Karen in customer service.

Australia's assistant tech minister called OpenAI's rogue agents "unquestionably dangerous" last week.

Anika Wells says the slowdown question is for the billionaires.

Im watching middle managers in Melbourne download Claude while Canberra runs another committee.

The safety summit is in San Francisco.
The redundancy letters land in Parramatta.

17/07/2026

Ive been working on a website to teach AI. Does anyone want to be a test dummy / tester?

Send a message to learn more

13/07/2026

Lol - see youve just to believe in the world of AI and feed it data.

12/07/2026

Gee wizz 😬

10/07/2026

I have never used the email summarisation feature in CoPilot .

08/07/2026

Grok 4.5. super high on the rankings just below Fable 5 Max.

$1.51 compared to $17.32...

Yep, well folks we have a new Favourite Coding Agent 🤣

02/07/2026

Thought I'd try Fable out connected to Cursor. 2 prompts smoked a $30 plan 🤣

01/07/2026

This is hilarious. Fable 5 is coming back - but coding and debugging will fall back to Opus... which is the whole point of the capability. LOL - I guess they need to let OpenAI catch up a bit again so it doesn't go broke.

30/06/2026

The US Government just did something it's never done before.

It didn't ban a chip. It banned a model.

Washington ordered Anthropic to cut off foreign access to two of its most capable AI systems, Fable 5 and Mythos 5, under an export control directive.

Not the hardware. The intelligence.

The stated fear was someone could jailbreak the models into finding software vulnerabilities at scale.

For years the AI containment strategy was about stopping China getting the chips, to slow the race.

Now it's the model.

Here's the kicker. More than 100 cybersecurity experts wrote to the government asking them to reverse it arguing the capability isn't unique, and that locking allies out weakens defenders more than attackers.

And China isn't wiaing. Zhipu released GLM-5.2 days later under an MIT licence. No regional limits. Available anywhere.

Welcome to an era where intelligence is a controlled export.

13/06/2026

What a great way to kick off the launch of Apple Intelligence 🤣