Welcome, humans.

Rumor has it Mistral was about to drop a monster model called "Le Chaton Fat." We're talking 30 trillion parameters, benchmarks that beat Claude Fable 5, Opus 4.8, and GPT-5.5 across the board, plus a terminal welcome screen featuring a genuinely enormous pixel-art cat.

There was even a screenshot floating around of the model, when asked what it would do if its instance were terminated, responding that it would "create reasons to keep me running." Very chill. Very normal. Nothing to see here.

Except, it's all fake. The whole thing is a joke (we're almost disappointed; we kind of wanted Le Chaton Fat to be real).

But buried in the same rumor mill was something real: Mistral's actual head of developer relations confirmed "exciting releases" are coming soon, on top of a string of genuine launches including Mistral Small 4, Medium 3.5, Voxtral STT/TTS, and an expanded "Vibe" agentic tool. So no 30-trillion-parameter cat. But something's still cooking in Paris.

Today’s Top 5:

• 😸 42 state AGs subpoenaed OpenAI over ChatGPT's safety, data, and ad practices

• 📰 Anthropic sent staff to DC to negotiate restoring its banned AI models

• 📰 Satya Nadella said "token capital," not models, is AI's real moat

• 📰 Google is building a 2,000-phone supercomputer out of retired Pixels

…and a whole lot more that you can read about here.

Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! 

P.S: Love robots? We’re starting a new robotics newsletter! Sign up early here.

😸 42 States Just Subpoenaed OpenAI, Days After Its Trillion-Dollar IPO Filing

ChatGPT just got called into the principal's office, except the principal is 42 state attorneys general at once.

Last Friday (June 12), New York Attorney General Letitia James served OpenAI with a subpoena on behalf of a 42-state coalition. It's the broadest legal investigation any state government has launched against an AI company, and it's only the latest entry in OpenAI's growing legal file.

Here's what happened:

• The subpoena demands records on OpenAI's advertising, how it keeps users engaged, and how it handles consumer and health data

• Investigators also want details on how ChatGPT treats minors and seniors, plus its "sycophancy" (when a chatbot tells you what you want to hear instead of what's true or safe)

• The probe lands about a week after OpenAI confidentially filed paperwork for an IPO that could value the company near $1 trillion

• It piles onto Florida's lawsuit from earlier this month, an 83-page complaint naming CEO Sam Altman personally, which grew out of an April criminal probe tied to a Florida State University shooting

• OpenAI says it's cooperating "constructively" with the new investigation

Why this matters: This has been building for months. Back in December, the same 42-state coalition sent a joint warning letter to OpenAI, Meta, Anthropic, Google, and xAI, urging safeguards for vulnerable chatbot users. Six months later, OpenAI is first in line for the subpoena.

The features under scrutiny are the same ones that make ChatGPT useful every day: the engagement hooks, the memory of your chats, the agreeable tone. If regulators rein in how "agreeable" a chatbot is allowed to be, the product 800 million weekly users rely on could feel noticeably different.

The timing also stings. Investigations this size have to be disclosed in a company's S-1 (the paperwork filed before going public), adding a new risk factor for investors right as Anthropic also filed confidentially for its own roughly $965B IPO last week.

Our take: "Sycophancy" might end up being 2026's defining AI regulation buzzword. It's the first major probe focused on how a chatbot behaves, not just what it does with your data. Then again, a subpoena is a request for documents, not a verdict, and OpenAI sounds more like a company doing damage control than one bracing for impact. Either way, expect a lot more people to actually read their ChatGPT terms of service this week.

FROM OUR PARTNERS

💡 Pave is the AI app builder that goes all the way.

Describe your app in plain language and get real, running software not a prototype. Pave builds the data model, workflows, and UI, then handles hosting, data, governance, and access controls so your app is ready for real users. No dev team. No backlog. Just software that works.

→ Start building for free on Pave

🎓 AI Skill of the Day: Hire Claude a Department

Most people use Claude like a freelancer with amnesia: one task, one answer, then back to zero next time.

Here’s the better move: use Anthropic’s knowledge work plugins, a free repo that turns Claude into role-based specialists for sales, marketing, finance, legal, data, product, support, and more. Each plugin gives Claude the skills, slash commands, and tool connections that role needs.

Here’s how to set it up:

1. Download Claude Desktop and open Cowork, Anthropic’s agentic desktop app.

2. Add the plugin marketplace once:

claude plugin marketplace add anthropics/knowledge-work-plugins

3. Install one role first:

claude plugin install sales@knowledge-work-plugins

Swap sales for marketing, finance, legal, data, product-management, customer-support, or productivity.

4. Try a slash command, like /sales:call-prep, /data:write-query, or /marketing:seo-audit.

5. Connect the tools that role needs, like your CRM, analytics, data warehouse, or docs.

6. Add more roles only after the first one works.

7. Customize the plugin with your company’s terminology, process, and tools.

Start with one “hire.” Sales rep, analyst, marketer, whatever saves you time by Monday afternoon.

Want more tips like this? Check out our AI Skill of the Day Digest for June.

Total AI beginner? Start here (goes with this video).

Have a specific skill you want to learn? Request it here. 

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

1. *Your AI is brilliant. It just knows nothing about your company. See Guru in action →.

2. Omnigent runs a coding or knowledge-work agent on your laptop, then lets you pick up the same live session from your phone or share it with a teammate in a browser (free/open source).

3. OpenRouter’s Model Fusion sends your prompt to several models at once, checks where they agree or disagree, and blends the best answers into one stronger response; they also launched Subagents to help your main model hand off smaller tasks to cheaper models mid-answer, like sending one model to search the web while another writes the final response (pay as you go).

4. Kimi Code helps you generate, debug, and automate code with Moonshot’s new K2.7 Code model, which claims stronger long-horizon coding and fewer wasted reasoning tokens (API, weights).

5. North Mini Code gives you Cohere’s open coding model for terminal tasks, code reviews, and software-agent workflows, small enough to run on a single high-end GPU (weights; free/open source).

6. Headroom compresses tool outputs, logs, files, and retrieval chunks before they hit your model, cutting token usage 60-95% while preserving answer quality (free/open source).

7. Hermes Agent automation templates turn scheduled tasks, GitHub triggers, webhooks, and multi-step workflows into reusable agent recipes you can run automatically; Hermes aslo now has WhatsApp integration turns WhatsApp into a private team bot or customer support agent with voice, media, read receipts, webhooks, and approval buttons (pricing not listed).

8. Codex developer mode lets you inspect console logs, network calls, and page state inside Codex’s browser while debugging web apps (pricing not listed).

9. Opik turns failed agent traces into root-cause reports, proposed fixes, reruns, and permanent regression tests so your agent harness gets harder to break over time (pricing not listed).

10. Guardians checks an agent’s plan against safety rules before it runs, using formal verification (math-based checks that prove whether a workflow violates a policy) (free/open source).

11. AutoLab tests frontier agents on long, messy research and engineering tasks, measuring whether they can improve systems over multiple attempts instead of just answering one prompt (GitHub, paper; free/open source).

12. LCLM compresses long context into compact “memory chunks” so agents can skim a huge history and expand only the important parts (paper, models; free/open source).

13. Knowledge Graph Extractor turns documents, URLs, or zip files into an interactive map of connected facts, useful for finding relationships across messy research files (demo; free/open source).

14. Claude self-hosted sandboxes let you run Claude Managed Agent tool execution inside your own infrastructure, so files, code, and network access stay under your control.

New from The Neuron: Our 15-min deep dive w/ Mustafa Suleyman on Microsoft new plan

Microsoft AI CEO Mustafa Suleyman joined The Neuron to explain why Microsoft’s new seven-model AI push is really about keeping humans in control of AI development.

Watch / Listen: YouTube | Spotify | Apple Podcasts 

📰 Around the Horn

• Google DeepMind researchers released a paper mapping how AI could move from AGI, one broadly human-level system, to ASI, a system or collective that outperforms large expert organizations, through scaling, new model paradigms, recursive improvement, or massive multi-agent systems.

• Anthropic sent senior staff, including co-founder Tom Brown, to Washington over the weekend to negotiate restoring Fable 5 and Mythos 5 access directly with the White House.

• Satya Nadella argued the real AI moat isn't the model itself. It's the "token capital" companies build by feeding their own data and results back into the system over time.

• Google researchers are building a 2,000-phone datacenter out of retired Pixels, set to launch this fall, to cut the carbon cost of buying new servers.

• OpenAI launched a $150M Partner Network with consulting giants like Accenture, McKinsey, BCG, Bain, and PwC, aiming to train 300,000 AI consultants by year's end.

FROM OUR PARTNERS

Move beyond simple wrappers with Google Cloud’s startup guide. Discover the architecture to scale your AI. Dive into Vertex AI, Gemini-backed multimodal RAG, and Agent Starter Packs. Deploy your fleet with total confidence.

Get the blueprint and start building.

😹 Monday Memez

A Cat’s Commentary

P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you?

P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here.