In partnership with
Welcome, AI enthusiasts
Imagine owning a slice of OpenAI without buying a single share. A new bill in the Senate would tax the largest AI companies half their stock, pool it into a $7 trillion public fund, and pay every eligible American more than $1,000 a year. Big idea, long odds. Let's dive in!
In today’s insights:
$7T Fund Stakes You in OpenAI and Anthropic
OpenAI's Daybreak Squares Off With Mythos
Sakana Matches Anthropic's Banned Models
Read time: 4 minutes
LATEST DEVELOPMENTS
PUBLIC AI
🏛️ $7T Fund Stakes You in OpenAI and Anthropic
Evolving AI: A new Senate bill would hand the public half of America's biggest AI firms.
Key Points:
Every major AI company would face a one-time 50% stock tax.
The transfer would make taxpayers a top shareholder in firms like OpenAI.
Eligible Americans could collect more than $1,000 a year from the fund.
Details:
Senator Bernie Sanders introduced the bill as AI valuations push toward the trillions. It covers any company clearing $200 million in yearly revenue and catches new ones as they grow. He argues that everyday people supplied the data and labor behind these models and deserve a return. OpenAI and Anthropic have floated softer public-fund ideas, while critics warn the plan gives Washington too much power.
Why It Matters:
Washington would gain real leverage over the labs, a seat at the table ordinary people have never had in AI decisions. For readers that could mean the tools reshaping their work and pay also return real money to them. A payout is years away at best, though the fight over who owns AI is now firmly underway.
TOGETHER WITH INVIDEO
🎨 invideo Agent One: Look behind the scenes of how a real AI film comes to life
Evolving AI: AI films are everywhere now, but you only ever see the finished film. Juicebox is the first time every step of a film coming to life has been documented.
Juicebox, by Hridaye A Nagpal, was made completely on invideo Agent One.
invideo Agent One is the AI video platform for serious creatives. You direct a crew of AI agents that hold the entire project in long-term memory and run in parallel like a real production unit. The agents keep characters, locations, and look consistent, so you can focus on the creative process.
Watch the 5-day journey of the film and how it came to life with invideo Agent One. Watch now →
Large studios, agencies, and brands are already using invideo Agent One to do real commercial work.
Source: OpenAI
Evolving AI: The two top AI labs are now racing to secure the world's software.
Key Points:
OpenAI expanded Daybreak to turn its bug-finding AI into a bug-fixing one for trusted defenders.
Anthropic made the same move first with its restricted Mythos model in April.
A new effort called Patch the Planet pays researchers to help fix flaws in widely used open-source code like Python and cURL.
Details:
OpenAI expanded Daybreak with an upgraded Codex Security tool that scans code then drafts a tested fix for a human to approve. Since March it has covered over 30 million commits across 30,000 codebases and helped close more than 500,000 findings. OpenAI also released GPT-5.5-Cyber which reached a record 85.6 percent on the CyberGym test and goes only to verified defenders. Security firms like Cloudflare and CrowdStrike can now run it inside their own products.
Why It Matters:
Cloudflare and CrowdStrike guard the apps and accounts people touch all day and now run AI that can spot a flaw and write its own fix. For ordinary users that means the software behind their banking and email gets quietly repaired before an attacker can reach it. The benefit reaches them as fewer stolen logins and fewer breaches they ever hear about.
Defense Spending Is Shifting. These 5 Stocks Are Why.
A new class of defense contractors is winning government business with AI systems and satellite infrastructure. We identified five of them in one concise, free research brief. Worth 5 minutes.
MULTI-AGENT AI
🐡 Sakana Matches Anthropic's Banned Models
Evolving AI: Sakana's Fugu matches Anthropic's export-controlled models through one API.
Key Points:
Fugu is one model that routes each task to a swappable pool of top LLMs.
Fugu Ultra matches Anthropic's Fable 5 and Mythos Preview on hard benchmarks.
Lose access to one provider and Fugu reroutes to another with no code change.
Details:
Sakana's Fugu is itself a model trained to call other AI models, deciding which one handles each step and stitching their work into a single answer. It comes in two tiers through one OpenAI compatible API. The standard Fugu favors speed for daily work, while Fugu Ultra pulls in heavier experts like Gemini, GPT, and Opus for code review, security analysis, and other multi-step jobs. Anthropic's Fable 5 and Mythos sit outside its pool, kept there by export controls.
👉 Try it now: https://sakana.ai/fugu
Why It Matters:
Fugu is built to absorb exactly the disruption these export controls create. For everyday users, a system that spreads work across many models means assistants that keep running when one provider goes dark, and that lean on whichever model is sharpest. It is a quiet shift in how AI reaches the people who use it.
👀 Click on the image you think is real
QUICK HITS
🔌 Groq Bounces Back With $650M, Pivots to Neocloud After Nvidia Poached Its CEO.
🎬 DeepMind and A24 Team Up on AI Film Tools in a First-of-Its-Kind Deal.
🛰️ SpaceX Signs $6.3B Compute Deal With Reflection AI, Its First Open-Source Customer.
🛡️ UK NCSC Pushes Leaders to Prepare for AI-Driven Cyberattacks.
📈 Trending AI Tools
🐰 CodeRabbit - Ship higher quality code with AI-powered code reviews*
⚡ General Compute - World's fastest AI inference at up to 1,000 tokens per second.
🏎️ Fastlane - Generates and schedules 30 days of viral short-form content in seconds.
💡 IdeaProof - AI startup validator that scores ideas in 120 seconds with market data.
*partner link