Ink & Code News: April 4, 2026
Stuffy Stuff That Happened This Week
Welcome to Ink & Code News! This week was bananas. There was a lot of stuffy stuff! Anthropic managed to leak both its next flagship model AND Claude Code's entire source code in the same week, OpenAI closed the biggest private funding round in human history, and a guy built a billion-dollar company from his living room with $20K and his brother.
So, buckle up, buttercup!
Anthropic’s Claude Leaks
Anthropic’s very bad, no good week of accidental disclosures
💧Leak #1: Claude Mythos
Uh-oh. Someone at Anthropic left the door wide open. A misconfigured content management system exposed thousands of unpublished internal documents to the public, including launch materials for the company’s next flagship AI, Claude Mythos. According to the leaked materials, Mythos sits in a brand-new “Capybara” tier above Opus. The draft flagged it as “far ahead of any other AI model in cyber capabilities,” with major advances in reasoning, coding, and cybersecurity. Anthropic confirmed to Fortune that the model exists and is in testing.
💧💧Leak #2: Claude Code Source Code
Oops, they did it again. In the same week, Anthropic also accidentally published Claude Code’s entire source code, over 1,900 files and 500K+ lines, to a public registry. Developers tore through the code and found 44 feature flags and three unreleased projects, including persistent cross-session memory and a deep-planning system. Anthropic called it “human error, not a security breach.” No customer data was exposed, but a GitHub mirror hit 4K+ stars and 7K+ forks in hours as everyone raced to go look at their secret stuffy stuff.
For authors: Were these leaks accidental or intentional? Either Anthropic's ops team needs a serious review, or someone in marketing is playing 4D chess. Either way, Claude is making headlines every week, and that's great marketing no matter how you slice it.
Money, Money, Money
The age of AI wealth creation is here
🚀 The One-Person, Billion-Dollar Company
In 2024, Sam Altman predicted that AI would make solo billion-dollar companies possible. A year and a half later, Matt Gallagher proved him right.
With $20K, two months, and a stack of AI tools, Gallagher launched Medvi from his house in LA — a startup selling GLP-1 weight-loss drugs online. He used ChatGPT, Claude, and Grok for code; Midjourney and Runway for ad creatives; ElevenLabs and custom AI agents for customer service. He outsourced doctors and shipping to telehealth platforms. His only full-time hire? His brother. Medvi is now projected to do $1.8B in annual sales.
For authors: If a guy can make a billion dollars selling weight-loss drugs from his living room, what can we achieve as indie authors and entrepreneurs? We already know how to build from nothing, hustle on a shoestring, and wear all the hats. Now we have tools that can multiply what we do a hundredfold. In the age of AI, the only real limit is your mindset.
Also in the news this week:
💰 OpenAI closed $122B in funding at an $852B valuation, the largest private raise in history.
🚀 SpaceX filed for a $1.75T IPO, which would be the largest in history after absorbing xAI into SpaceX.
🛸 Starcloud raised $170M to build GPU-powered data centers in orbit. Yes, really.
All the Shinies
This was a massive week for product launches and new features.
Here’s the highlight reel:
💎Google released Gemma 4, its most capable open-source model family, now under Apache 2.0 in four sizes, from phone to desktop, with the smallest running entirely offline.
💻Claude Code gained computer use, letting it see and interact with what it builds, now on Mac and Windows.
🧮Perplexity Computer can now do your taxes: upload docs, answer questions, watch it fill out IRS forms (demo hit 1.9M views).
🔌OpenAI launched Codex plugins that bundle skills, apps, and MCP servers into reusable automated workflows.
🔬Microsoft launched Copilot Cowork plus Critique and Council, features that pit Claude against ChatGPT for multi-model research.
🖥️Cursor unveiled Cursor 3, rebuilt to run fleets of local and cloud coding agents in parallel across multiple repos.
🤖Sakana AI opened beta for Marlin, an autonomous research agent that can work up to 8 hours straight on business tasks.
🧠Gemini now imports chat history from other AI tools, so you can switch without losing your conversations.
🏗️Devs built an open-source Claude alternative with Cowork-like abilities that runs locally with any LLM (9.5K bookmarks and counting).
💬Slack announced 30 new AI features for Slackbot, including reusable skills, post-meeting summaries, and cross-device context memory.
🎞️ByteDance’s Seedance 2.0 hit #1 on video generation leaderboards — another option for book trailers and promo content.
📈Enrich Labs released Helena, an autonomous marketing agent that researches your brand, generates assets, and posts them (launch video hit 3M+ views).
🌍Google expanded Live Translate to iOS, turning any headphones into a real-time interpreter across 70+ languages.
Phew!
For authors: The theme this week is autonomy. AI tools aren't just getting smarter, they're getting more independent. Agents that research for hours, do your taxes, handle your marketing, and test their own code. The tools are doing more so you can focus on what matters: writing, creating, and building your business.
As The AI Turns
Ongoing saga updates: corporate drama, AI behaving badly, drama llamas, etc
Ongoing saga updates: corporate drama, AI behaving badly, and other drama llamas
🦀Anthropic Cuts Off OpenClaw, Then Hands Out Free Credits
Hot off the press: As of today (April 4), Anthropic ended subscription coverage for third-party tools like OpenClaw, saying they bypass prompt cache optimizations and put “disproportionate pressure” on compute resources. Translation: people were getting way more value than Anthropic bargained for.
To soften the blow, Anthropic is giving every subscriber a one-time credit matching their monthly fee ( $20 for Pro, $200 for Max), but you have to enable Extra Usage to claim it, and it expires April 17. They also rolled out discounted extra usage bundles at up to 30% off.
Generous? Maybe. But also a very smooth way to get everyone opted into pay-as-you-go billing.
⚖️Anthropic Wins Pentagon Injunction
Anthropic won a federal injunction blocking the Trump administration’s supply-chain-risk designation, with the judge calling it “classic illegal First Amendment retaliation.” A major legal win for the company behind Claude — if the Pentagon had succeeded, it could have set a precedent for government retaliation against AI companies that build in safety limits.
🎬OpenAI’s Wild Week
Open AI had a crazy week of its own:
Sora Closure: Disney Blindsided: A WSJ investigation revealed the behind-the-scenes chaos of Sora’s shutdown: a $1M daily burn rate, Disney learning about it “less than an hour” before the announcement, and the compute being redirected to “Spud” — a model targeting coding and enterprise.
TBPN Acquisition: First Media Deal: OpenAI bought TBPN, the daily live tech talk show (~70K viewers/episode), for a reported low hundreds of millions. The 11-person team keeps editorial independence but drops ads. An AI company buying a media outlet to control its own narrative.
Project Stagecraft: Training AI to Replace Workers: Business Insider revealed an internal effort paying up to 4,000 freelancers $50/hr to build occupation-specific training data. One contractor: “We all were aware that we were basically training AI to replace us.” OpenAI is also drafting policy papers on “rethinking the social contract.”
Altman vs. Amodei: The WSJ Deep Dive: The WSJ laid out the decade-long personal grudges between Sam Altman and Dario Amodei, tracing back to a San Francisco group house in 2016. Highlights: Brockman once floated selling AGI to UN Security Council nuclear powers (Dario called it “tantamount to treason”), Altman accused the Amodeis of plotting against him, and Amodei privately likened the Altman/Musk lawsuit to “Hitler vs. Stalin.”
For authors: The people building the tools we use every day are feuding, buying media companies, and paying freelancers to train AI that replaces them. These are the forces shaping the future of creative work. Stay informed. The more you understand the players, the better you can navigate what's coming.
Stranger than Fiction
Weird, funny, absurd AI stories
And finally, because this industry can't go a single week without making us question reality, here are the stories that sound fake but aren't.
🐾BUDDY: Anthropic’s Secret Terminal Pet
Buried in the Claude Code source code leak: an unreleased AI terminal pet called BUDDY with 18 species, rarity tiers, and stats like CHAOS and SNARK. This is the most delightful thing to come out of a security incident in tech history.
I got this guy. Which one did you get?
🍺The Guinndex: AI Beer Price War
An Irish entrepreneur used AI to track the cost of a Guinness pint across 3,000 pubs — and now some pubs are actually lowering prices to stay competitive on the leaderboard. Because if there’s one thing that will unite humanity, it’s affordable beer.
🏛️Public Launches AI Brokerage
Public revealed the first agentic brokerage, letting anyone design AI agents that autonomously monitor the market, execute trades, and manage money. Launch video hit 2.3M views. Whether you trust a bot with your book royalties is another question entirely.
🚕100 Robotaxis Just... Stopped
A “system failure” hit Baidu’s Apollo Go fleet in Wuhan and over 100 robotaxis simultaneously froze in the middle of busy roads, some on highways. Passengers were trapped inside for up to two hours while trucks blew past them before police showed up to pry open doors. The future of transportation, everybody.
For authors: Secret terminal pets, beer price wars, AI stock brokers, and now robot cars holding people hostage on highways. If you’re ever stuck on a plot twist, just read the news. Reality is writing fiction for us at this point.