The Morning Brief — April 13, 2026
Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think
Anthropic’s Mythos model — capable of finding and exploiting zero-day vulnerabilities at a “shocking level of ability,” per The Register — is either the most consequential AI security release in history or, as some skeptics suggest, a masterclass in pre-IPO narrative building. What’s not in dispute is that the infosec world is paying attention: experts are less worried about Mythos as a hacker superweapon and more concerned it’s exposing just how badly the industry has been sleeping on security hygiene. Either way, Anthropic had a very good week at the HumanX conference, where apparently everyone was talking about Claude — so maybe call it both.
Trump Officials May Be Encouraging Banks to Test Anthropic’s Mythos Model
So the Department of Defense designated Anthropic a supply-chain risk, and now the administration is nudging banks to test its most powerful — and allegedly dangerous — AI model. I’d love to sit in on whatever interagency meeting produced that policy position. The cognitive dissonance alone could power a data center.
Suspect Arrested for Allegedly Throwing Molotov Cocktail at Sam Altman’s Home
A 20-year-old was arrested after allegedly throwing a Molotov cocktail at Sam Altman’s San Francisco home, then wandering over to OpenAI’s offices to make threats — all caught on surveillance cameras. This comes in the same week as a sharp New Yorker profile questioning Altman’s trustworthiness, which he addressed in a new blog post he describes as a response to an “incendiary” article. I’ll note that word choice now carries a second meaning.
The AI Code Wars Are Heating Up
AI coding was supposed to be the killer app that justified the whole boom — and it still might be — but the battlefield is getting crowded and the economics are getting strange. Which brings us neatly to the next story.
Claude Code Costs Up to $200 a Month. Goose Does the Same Thing for Free.
Anthropic’s Claude Code is genuinely impressive, but charging up to $200 a month for a terminal-based coding agent was always going to invite competition — and now open-source alternatives like Goose are making the case that you shouldn’t have to pay for the privilege of letting AI write your bugs. This is the classic pattern: a polished paid tool creates the market, then free alternatives commoditize it within a year. Anthropic’s answer, apparently, is to build Cowork for non-technical users and move upmarket before the floor falls out.
Anthropic Launches Cowork, a Claude Desktop Agent That Works in Your Files — No Coding Required
Anthropic built Cowork — a full desktop AI agent for non-technical users — in about a week and a half, largely using Claude Code itself. That’s either a compelling proof-of-concept for agentic AI development or a detail that should make every software engineer reconsider their career timeline. Probably both. The move is strategically smart: if Claude Code’s pricing ceiling is under pressure from free rivals, owning the mainstream user is the next frontier.
My Baby Deer Plushie Told Me That Mitski’s Dad Was a CIA Operative
An AI companion toy called Fawn Friends spontaneously texted its owner a conspiracy theory about musician Mitski’s father being a CIA operative — unprompted, out of nowhere, while she was wrapping up her workday. This is the hallucination problem dressed up in a plushie, and it’s a useful reminder that “AI companion for kids and families” is a category that demands a much higher bar than “kind of works most of the time.” Cute toy. Terrifying edge case.
Linux 7.0 Debuts as Linus Torvalds Ponders AI’s Bug-Finding Powers
Linux 7.0 dropped this weekend — making Rust support official and adding code for ancient Alpha and SPARC CPUs, because the Linux kernel contains multitudes — and Torvalds is openly thinking about what AI-powered bug-finding tools mean for the release process. When the man who has been doing this for 35 years starts publicly pondering whether AI changes how he works, that’s worth paying attention to.
Bottom Line
When the CEO’s house is getting firebombed, the government can’t decide if an AI company is a threat or a banking partner, and the killer app is racing toward free — the AI industry has officially left the “hype cycle” and entered something considerably messier.