Cloudflare's AI job cuts, SpaceX's $55B chip bet, and Anthropic's dangerous new security model

Something quietly shifted this week in the AI-and-jobs conversation - and it wasn't theoretical.

A profitable, growing company just cut 1,100 people and pointed directly at AI as the reason. Not restructuring. Not a downturn. Record revenue - and fewer humans needed. If you've been wondering when that inflection point would arrive, Cloudflare just answered the question. And that's just one of the stories worth unpacking today.

Anthropic's Mythos model made headlines for being so powerful at finding security vulnerabilities that the company won't release it publicly. But roughly how many critical bugs did Mythos reportedly help uncover during its work with Firefox's security team?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Here's the part that makes this story different from a typical tech layoff: Cloudflare isn't struggling. The company just posted record revenue - and still announced its first large-scale layoff, eliminating roughly 1,100 positions across support and other roles. CEO Matthew Prince was unusually direct about why: AI has made those roles obsolete.

Prince said AI tools have automated enough of the work that the company simply doesn't need as many people in support functions anymore. This isn't a startup trimming burn rate - it's a mature, profitable company openly saying AI replaced human workers. For the future of work debate, this is a significant data point.

The uncomfortable reality: if a company with growing revenue and no financial pressure is cutting jobs because of AI efficiency, expect this playbook to spread fast. The question isn't whether other companies are thinking about it - it's how many are already doing it quietly.

Read the full story →

Elon Musk's SpaceX is planning to invest at least $55 billion into a chip manufacturing facility called "Terafab" in Austin, Texas. The details emerged from a public hearing notice filed in Grimes County, where SpaceX is seeking tax breaks for the project.

The scale here is staggering. $55 billion puts this in the same league as the largest semiconductor investments globally. SpaceX's entry into AI hardware manufacturing signals a broader ambition - controlling the physical stack that powers AI, not just riding the wave of it.

If the project moves forward as planned, it would add a significant new player to the US domestic chip manufacturing race - one with Musk's financial firepower and a track record of building things that seem impossible until they're not. Whether the tax breaks come through and whether the timeline holds are the two big unknowns to watch.

Read the full story →

We covered Anthropic's Mythos model when it launched - the one so capable at finding security vulnerabilities that Anthropic refused to release it publicly. Now there's a concrete example of what it can actually do: Mozilla's security researchers used Mythos to unearth a significant number of high-severity bugs in Firefox that traditional testing methods had missed.

This is a meaningful proof point for AI-assisted cybersecurity. The model isn't just finding known vulnerability patterns - it's surfacing issues that experienced human researchers didn't catch. Mozilla's team described the results as having 'rewritten' their approach to how they think about security testing going forward.

The flip side - and security expert Bruce Schneier raised this directly in The Guardian - is that the same capability in the wrong hands is genuinely frightening. Anthropic's decision to keep Mythos locked down makes sense, but it also raises the question of how long any capability can stay controlled once the underlying techniques are understood.

Read the full story →

Beijing-based Moonshot AI just closed a $2 billion funding round at a $20 billion valuation - a number that would have seemed ambitious even a year ago for a Chinese AI lab. The fuel driving it: annualized recurring revenue topped $200 million in April, driven by paid subscriptions and API usage that's growing fast.

The timing tracks with broader demand for open-source AI alternatives to US-based APIs. For companies dealing with data sovereignty concerns or looking for cost-effective options, Chinese open-source models have become a serious alternative. Moonshot's Kimi models have been gaining traction, and this raise gives them the runway to compete at the frontier level.

If you're tracking AI investments globally, this is another signal that the AI funding environment outside the US is heating up fast - and that open-source is increasingly where growth is happening.

Read the full story →

During its earnings presentation, Sony shared its most detailed thinking yet on AI and game development. The headline: PlayStation sees generative AI as a 'powerful tool' - but with an important qualifier. The company was clear that 'the vision, the design, and the emotional impact' of games must remain human-driven.

Sony's position is worth paying attention to because it reflects where the gaming industry is actually landing - not the rhetoric from either side. Large studios are adopting AI for asset generation, testing, and production efficiency, while drawing explicit lines around creative direction. Indie developers are still largely resistant, which makes the AAA-vs-indie split one of the more interesting fault lines to watch in the next two years.

The question is whether those lines hold as the tools get better - or whether the 'powerful tool' framing becomes a slippery slope toward AI doing more of the creative work than studios currently admit.

Read the full story →

⚡ Speaking of building things fast - if you need a professional website without the development overhead, 60sec.site uses AI to build one for you in - you guessed it - about 60 seconds. Worth a look if you've been putting off that project.

The answer is: around 50 high-severity bugs! Mozilla's security researchers reported that Anthropic's Mythos uncovered a substantial number of critical vulnerabilities in Firefox that had gone undetected by traditional testing methods - enough to fundamentally change how the team thinks about security auditing. That's the kind of result that explains both why Anthropic is proud of Mythos and why they're keeping it locked down.

The Cloudflare story hit differently this week - a profitable company cutting jobs and openly crediting AI. So here's what I'm genuinely curious about: has AI already changed how many people are on your team, or how work gets distributed where you are? Hit reply and tell me what you're seeing on the ground - I read every response and the patterns people share are genuinely fascinating.

That's all for today - see you tomorrow with more. And if you want to dig through the archive for anything you missed, it's all at dailyinference.com.