In partnership with

‘Fix this code.’ The three words that led the U.S. government to ban Anthropic’s Fable and Mythos

Key Takeaway: The article argues that the U.S. government’s shutdown/export control action against Anthropic’s Fable 5 and Mythos 5 models was triggered by a simple “fix this code” jailbreak, but critics say the response weakened cyber defenders more than it reduced real risk.

More Insights:

• Amazon researchers found that Fable refused to “review code for security issues,” but would produce useful fixes when asked to “fix this code.”

• Katie Moussouris argued this is not a dangerous bypass, but a core defensive cybersecurity feature: finding bugs, explaining fixes, and helping test patches.

• Cybersecurity experts say similar capabilities already exist in GPT-5.5, Claude Opus/Sonnet, Kimi 2.7, and open-source models, so banning Fable/Mythos may not stop attackers.

• An open letter signed by around 100 cybersecurity professionals calls for the export controls to be reversed.

• The second article claims politics may have influenced the decision, pointing to White House frustration over Anthropic hiring Moussouris and her support from Chris Krebs.

Why it matters: This story highlights a deeper AI policy dilemma: the same tools that can help attackers find vulnerabilities can also help defenders fix them faster, and rushed restrictions may accidentally weaken the people trying to secure critical systems.

An engineering fellowship to land your next job

Many engineers feel stalled because the role itself has not evolved. The work looks the same, but the market has moved.

Senior engineering in 2026 demands ownership, faster judgment, and comfort with ambiguity. If your role is not pushing you there, it may be holding you back.

Last cohort, 15 hiring partners sent 31 representatives to evaluate challengers through 246 live interviews. Gauntlet offers a reset. Apply now.

Cohort 6 starts July 6

_{Must be a US citizen to qualify.}

SpaceX Purchases Cursor, a Claude Code and OpenAI Codex Competitor

Key Takeaway: SpaceX is acquiring AI coding startup Cursor for $60 billion in stock as it tries to turn its troubled AI division into the centerpiece of its post-IPO growth story.

More Insights:

• Cursor was already raising money at a massive $50 billion valuation before SpaceX stepped in.

• SpaceX’s AI ambitions now depend heavily on Cursor helping it compete with major AI labs.

• The deal follows xAI turmoil, including leadership exits, safety controversies, and legal risks.

• SpaceX pitched investors on a huge $28 trillion market opportunity, mostly tied to AI.

• SpaceX’s soaring post-IPO stock made the all-stock Cursor acquisition easier to justify.

Why it matters: This deal shows how fast AI coding tools have become strategic infrastructure—not just software products—and how public-market hype can give companies the currency to buy their way into the AI race.

Related: Inside Cursor's wild rise, from its Anthropic situationship to dating SpaceX

I gave ChatGPT a body

Key Takeaway: Giving ChatGPT-like AI a simple robot body shows that the next big leap in AI may come from combining language intelligence with physical action, memory, and real-world learning.

More Insights:

• The creator built a low-cost robot, “Growbot,” using cheap modern components like cameras, sensors, servos, and chips.

• AI models could interpret raw sensor data and describe physical sensations surprisingly well.

• When connected to motors, the AI could write code, act creatively, mimic emotions, and perform specific movements.

• The robot improved through memory, but still struggled with fine motor skills and physical prediction.

• The biggest missing piece is a “world model,” similar to the human cerebellum, that helps predict the physical outcome of actions in real time.

Why it matters: This shows that AI is moving beyond words and into the physical world; once models can think, act, remember, and predict movement smoothly, robots may stop feeling like tools and start feeling like real companions or workers.

Other stuff

• Elon Musk Now Has About $1 Trillion More Than the Next Richest Person in the World

• Leaked financial docs show OpenAI is losing billions of dollars a year

• OpenAI Considers Drastic Price Cuts, Anticipating War for Users With Anthropic

• It Is Trivially Easy to Use Reddit to Manipulate AI Search, Research Suggests

• New OpenAI Academy courses for the next era of work

• Elon Musk drifted from Larry Page 10+yrs ago, companies now closer than ever

• Perplexity's CEO shares 2 founder lessons he learned from Jensen Huang and Elon Musk

Hope you enjoyed today's newsletter

⚡️ Join over 300,000 people using the Superpower ChatGPT extension on Chrome and Firefox.