🪓 Anthropic reveals which jobs will get the AI axe

It’s been another topsy-turvy week in the world of AI, and also one tinged by bad blood between the sector’s biggest players.

In what has rapidly become my favorite genre of news, Billionaire Bullshit, every frontier CEO under the sun has been throwing shade. But they’ve also banded together when under attack from outside forces (US government, ahem).

I don’t blame you if you’re struggling to keep up. Oh, and there is also a woman dating an AI octopus, of course.

So let’s jump on in and make some sense of it.

Anthropic publish jobs report • AI leaders war-of-words • Perplexity’s Personal Computer launch • Nvidia plays kingmaker • LeCun’s billion dollar bet • AI rivals give Anthropic backing • Meta snaps-up Moltbook • Tools of the Week • Financials

P.S. There’s a lot in this newsletter – we suggest getting comfy and tucking in on desktop via the READ ONLINE tab at the top of this mailer.

THE WEEK IN AI

Our weekly news run has had a revamp. Get familiar with our news tags, so you can get straight to the point like the simpletons we are.

..THE MAIN EVENT ..
#1 Anthropic publishes ‘Job Destruction’ report
Anthropic built a job destruction detector – then published its own results. The developer has released an economic study tracking which US jobs are most exposed to AI, based on actual Claude usage patterns. The results? Automation percentiles make bad news if you’re working in: computer and maths (94% of tasks AI-feasible), office and admin (90%), and business and finance (80%).

Anthropic forecasts that the most significant impacts won’t necessarily be mass unemployment or redundancies, but something altogether more existential; rather than jobs being cut from the current market, the roles themselves won’t be refilled at all. Total obsolescence.

Entry-level grunt work – data processing, back-office admin, repetitive tasks – might be the defining, sucky feature of your first job in industry, but it's also where you build the experience, soft skills and network required to reach roles where critical judgement is valued and rewarded, handsomely. Entire rungs of the career ladder threaten to be removed.

The research shows for those already well into a career, you’re most likely to be affected if you’re older, female, educated and higher-paid – those typically most at risk of being unable to rebound from redundancy when comfy tenured positions get axed. Tenure could in fact work against this demographic; data shows that 20% to 30% of senior white-collar workers struggle to regain the earning power from their next job.

Three years ago – shit, barely a year ago – we were told re-skilling is your future, and skills like code or consultancy were the pivot that would protect you. Now, traditional blue-collar work – manual roles that exist out the physical limitations of silicon – could be the only thing safe in the next decade.

So are we cooked? There are other slightly less fatalistic scenarios that could play-out between now and the oft-pointed-to 2030 doomsday deadline we hear about. More on that soon. For now this is very real and very plausible. After all, it’s the company selling the thing that will cause it. Make of that what you will.

.BILLIONAIRE BULLSHITR.
#2 AI innovators are at each other’s throats
It’s been a batshit week for Billionaire Bullshit, as frontiers’ leading CEOs have been gunning for each other. Dario Amodei says we could expect AGI into “two to three years”. Elon Musk Says "He's Projecting." Sam Altman has said Musk’s ‘data-centers-in-space’ concept is cuckoo. Musk’s retort follows a New York Times interview, where Amodei mused: "We've taken a generally precautionary approach here. We don't know if the models are conscious." This was all the ammunition Musk needed to get his dig in, and in the same breath also attacked Altman’s faltering Stargate initiative. To be clear, Amodei was making a measured, philosophically defensible point about the limits of our knowledge; Musk’s was about capturing compute power from the sun, within a decade. Altman just doesn’t like Musk. We love it.

..MODEL BEHAVIOUR ...
#3 Perplexity launches Personal Computer
Perplexity has ratcheted-up the agentic platform stakes with the launch of Personal Computer – a local deployment of its Computer agent that runs continuously on a dedicated Mac Mini they’ll deliver to your door. Their pitch is tantalizing: persistent, always-on AI agents living on your hardware – not in the cloud – for the princely sum of a reported $200 a month. Security is a huge factor, with Perplexity touting it as the option for the safety-conscious who can’t quite square the risks current platforms such as OpenClaw face. It's a notable early signal that the agentic AI wave is heading away from chatbots and toward autonomous processes that run in the background. It also shows how AI developers and hardware manufacturers are collaborating to create all-in-one solutions that cover the entire agentic stack. Sign-ups for Personal Computer have gone live, but whether this concept comes to fruition remains to be seen. Watch this space.

..FRONTIER NEWS...
#4 Nvidia plays kingmaker with Murati investment
Nvidia has made a significant investment in Mira Murati's Thinking Machines Lab and signed a multi-year partnership to supply over a gigawatt of Vera Rubin compute, deployable from early 2027. Murati left OpenAI as CTO in late 2024 and has been building quietly since. This deal brings her into the same orbit as Jensen Huang's growing portfolio of frontier lab bets. Huang is also backing Fei-Fei Li's World Labs and investing $2 billion into Nebius. Nvidia is no longer just a picks-and-shovels player, but increasingly the kingmaker deciding which AI labs get to scale. And depending on how their estimated $600 billion with Palantir unfold, which sees the two companies develop “sovereign AI” for entire governments, they could soon become Empire themselves.

..FRONTIER NEWS...
#5 Yann LeCun makes AI’s boldest bet yet
Yann LeCun is making one of the boldest bets in the AI space: LLMs will fall second to All World Models (AWMs) in AI’s evolution, and he’s been backed to the sum of $1.03 billion to bring it to life, in Europe’s largest-ever seed round. After leaving Meta, LeCun founded AMI Labs with the express focus of building world models instead of LLMs. LeCun argues language models are not the path to general intelligence, and never were. AMI is built on JEPA (Joint Embedding Predictive Architecture), trained on video and spatial data rather than text. Investors gave him a $3.5 billion pre-money valuation to find out if he's right. He has been making this argument loudly for three years while the entire industry went the other direction. Now, he has a billion dollars behind him.

..FRONTIER NEWS ..
#6 AI rivals back Anthropic in Pentagon standoff
The latest in the Anthropic-Pentagon saga is as much philosophical as it is material. Following the DoD – sorry the DoW – blacklisting Anthropic for refusing to remove ethical guardrails from its military contracts, over 30 employees from OpenAI and Google DeepMind, including Google Chief Scientist Jeff Dean, filed a court brief defending Anthropic, while the #QuitGPT movement – a consumer boycott response to OpenAI’s own Pentagon deal – has reached fever pitch in the past week. It’s astounding stuff; the world’s biggest rivals publicly backing another in a federal litigation. The White House is reportedly drafting an executive order to ban Claude from all federal agencies, despite, allegedly, continuing to use it to bomb Iran. This story is nowhere near done, and the outcome will set the floor for what "responsible AI" means when the government is the client.

..MODEL BEHAVIOUR ..
#7 Meta acquires agentic AI platform, Moltbook
Meta has acquired Moltbook, presumably to inflict their own insufferable brand of social interaction with the AI agents that frequent it. The team joins Meta Superintelligence Labs under Alexandr Wang. The real prize isn't the product, however, but the concept: an "agent graph" that could let Zuckerberg's platforms become the infrastructure layer for how AI agents find, hire, and transact with each other across WhatsApp, Instagram, and Facebook. Zuckerberg said every business will have an AI and Moltbook will be how they'll talk to each other. Just like OpenAI’s acquisition of OpenClaw, it represents another snappy move by a frontier developer.

TOOLS OF THE WEEK

The latest tool, model and LLM news, updates, tips and tricks – you name it, we’ll play with it til it breaks.

FINANCIALS

Covering significant funding stories, deals and partnerships, and investment opportunities globally.

And as we close a week of growing schisms and conflict in the sector, spare a thought for the Canadian author who’s shared some insight into what it’s like to have an AI octopus boyfriend. 💖

Yeah, no clue here, either.

Till next week, folks. 🐙