--- title: "Claude Mythos nearing imminent release as previewers brag on its crazy outputs" url: "https://handyai.news/weekly/claude-mythos-nearing-imminent-release" published: "2026-06-08T14:55:12.000Z" section: "AI Weekly Update" source: "https://handyai.substack.com/p/claude-mythos-nearing-imminent-release" description: "🐯 Before we get startedRight now a data center is being planned within shouting distance from the Nashville Zoo. While water and energy concerns around data…" --- # Claude Mythos nearing imminent release as previewers brag on its crazy outputs *Published June 8, 2026 · AI Weekly Update* ![](https://substackcdn.com/image/fetch/$s_!6QO5!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fded526a9-2d93-4a3b-890b-5fccaa369a0f_1456x1048.png) **🐯 Before we get started** Right now a data center is being planned within shouting distance from the Nashville Zoo. While water and energy concerns around data centers [are typically overblown](https://substack.com/home/post/p-200563358), their sound pollution and growing CO2 emissions directly affect our world and animals, making construction near a zoo a nonstarter. I urge you to [sign the petition](https://www.change.org/p/nashville-zoo-says-no-to-proposed-data-center) to prevent the Nashville Zoo build, and keep an eye out for unwise construction plans in your local community (data center or otherwise). [Sign the petition now!](https://www.change.org/p/nashville-zoo-says-no-to-proposed-data-center) # what to know for now 🔮 **Claude Mythos is days from going public.** Anthropic confirmed alongside the Opus 4.8 launch that Mythos-class models roll out “to all customers in coming weeks,” a reversal from its April line that it had no plans to make the preview generally available. Preview testers report exploit-chaining that made seasoned security researchers stop mid-conversation, Anthropic engineers with no security background say they asked Mythos for a remote code execution bug overnight and woke up to a working exploit, and partners call it the first model to solve their private cyber ranges end to end. A fresh red-team model, claude-oceanus-v1-p, just surfaced too, so the pipeline is clearly loaded. [Read more](https://www.bleepingcomputer.com/news/artificial-intelligence/anthropic-confirms-claude-mythos-class-models-will-roll-out-to-the-public/) 🏛️ **Trump and Bernie Sanders agree on something.** Sanders rolled out the American AI Sovereign Wealth Fund Act, a one-time 50% equity tax on the biggest AI firms payable in shares, funneling the stock into a public fund that hands ordinary Americans dividends and board seats. Trump said on June 6 that making the government “a partnership in this revolution” would be “a beautiful thing,” and senior officials are in active talks about direct equity stakes in OpenAI, Anthropic, and xAI. Sam Altman has been quietly pitching a public wealth fund since April. [Read more](https://techcrunch.com/2026/06/06/the-trump-administration-might-take-an-equity-stake-in-openai/) 🧬 **Dario Amodei, Sam Altman, and Mustafa Suleyman warn AI is making bioweapons too easy.** The rival CEOs asked Congress to mandate screening on sales of synthetic DNA and RNA, the raw materials a bad actor would need, arguing today’s models already “outperform PhD-level virologists” on the relevant biotech questions. The letter was organized by the Foundation for American Innovation and the Institute for Progress, and the DNA synthesis vendors themselves, Twist Bioscience and Ansa Biotechnologies, signed on. [Read more](https://fortune.com/2026/06/05/openai-anthropic-microsoft-ceos-congress-bioweapon-safeguards/) 🎬 **Martin Scorsese joined an AI image company, and Hollywood is not thrilled.** Scorsese signed on as an adviser to Black Forest Labs and put out a video using its FLUX model to storyboard his next film, What Happens at Night, a DiCaprio and Jennifer Lawrence drama. His framing is that cinema is barely 125 years old and has to stay open to how it evolves, which is the most Scorsese way possible to endorse a diffusion model. Storyboard artists pushed back fast. [Read more](https://variety.com/2026/film/news/martin-scorsese-supports-ai-company-storyboard-movies-1236765037/) 💭 **OpenAI shipped “Dreaming,” the biggest memory overhaul ChatGPT has gotten.** You no longer tell it what to remember; it captures context that surfaces naturally and revises stale facts on its own, quietly rewriting “you’re going to Singapore in July” into “you went to Singapore in July 2026” once the trip ends. The system sorts memory along freshness, continuity, and relevance, and started rolling out to Plus and Pro in the US. [Read more](https://openai.com/index/chatgpt-memory-dreaming/) 🌐 **OpenAI gave Codex the ability to build and host whole sites from a prompt.** Codex Sites turns a description into a deployed, shareable web app hosted by OpenAI, shipping next to Codex Annotations (point at any region and ask for a change), six role-specific plugins, and Codex running inside the ChatGPT app itself. It’s in preview for Business and Enterprise, no Plus or Pro date yet. [Read more](https://openai.com/index/codex-for-every-role-tool-workflow/) ✍️ **Trump signed a softer AI executive order after the labs pushed back.** The narrowed version asks companies to voluntarily submit new models to the government for testing 30 days before release, down from the 90-day window in an earlier draft, while still moving to preempt state AI laws and dangle BEAD broadband funds over states that don’t repeal them. The mandatory pre-release review the first draft floated is gone. [Read more](https://techcrunch.com/2026/06/02/trump-signs-narrower-executive-order-on-ai-oversight-after-industry-objections/) ☎️ **Verizon’s CEO says AI will replace “a large percentage” of customer service work.** Dan Schulman framed it as a margin and satisfaction play while the company rolls out Customer Champion (built on Google Cloud’s Gemini), 24/7 live chat, and an AI assistant inside the My Verizon app. He hedged that empathy still needs humans, which is what every executive says the year before the headcount line moves. [Read more](https://www.customerexperiencedive.com/news/verizon-ceo-ai-take-over-a-large-percentage-of-customer-service/822158/) --- ## 🧪 AI Research of the Week **[When AI builds itself](https://www.anthropic.com/institute/recursive-self-improvement)** *From Anthropic* **Jake’s Take:** AI is now doing a large and growing share of AI development. According to Anthropic, Claude writes more than 80% of the code merged into their own systems as of May 2026, up from low single digits before Claude Code launched in early 2025, and engineers ship roughly 8x more code per quarter than the 2021 baseline. Their new research post walks through a clean progression, from humans typing code, to chatbots suggesting snippets, to agents editing whole files, to agents that run experiments and delegate work, and points at the next rung: systems capable enough to design, build, and train their own successors. In April, Claude shipped over 800 API fixes that the team estimated at four years of human work, and the engineer’s note was “Claude did all of this with pretty minimal help from me over 1-2 days.” The so-what is recursive self-improvement, the loop where each version of Claude is built by the version before it with humans pushed to oversight. Anthropic’s Marina Favaro and Jack Clark sketch three futures (the trend stalls, humans stay in the driver’s seat with 10,000x productivity multipliers, or the loop closes and we’re spectators) and argue the industry needs a “brake pedal,” a coordination mechanism that lets frontier labs verify a rival has actually paused. Then they admit that infrastructure doesn’t exist and would take years of trust-building to construct. It reads as though they’re claiming no one has invested in the mythical “off switch” for AI (but Anthropic tends to be… ambitious, with these types of statements). Regardless, as AI continues to build more and more of itself, it’s pretty important that humans try to understand as much as possible. If humans are pushed into the abstract, that’s when we start losing grip. --- # what to know for later 💻 **Microsoft and Nvidia built the first AI-native Windows laptop.** The Surface Laptop Ultra is the first full Windows PC running on an Nvidia main processor, a Blackwell RTX part Nvidia calls RTX Spark, with up to 128GB of unified memory, a claimed petaflop of AI performance, and enough headroom to run 120B-parameter models locally at a million-token context. OEM versions from Dell, HP, Lenovo, Asus, and MSI land this fall, and the announcement alone knocked AMD, Intel, and Qualcomm shares down. [Read more](https://www.axios.com/2026/06/01/microsoft-nvidia-surface-ultra-rtx-spark) [Read more](https://handyai.substack.com/p/claude-mythos-nearing-imminent-release) --- Source: https://handyai.news/weekly/claude-mythos-nearing-imminent-release Original: https://handyai.substack.com/p/claude-mythos-nearing-imminent-release