🔓 Anthropic Accidentally Ships Its Own Source Code

Anthropic is having a rough week: its Claude Code source was accidentally exposed in a routine update, and the internet noticed fast. A small debugging file ended up pointing straight to the full codebase, turning a simple slip into a viral moment overnight. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Anthropic accidentally leaked the full source code of Claude Code through a routine npm package update, its second major security blunder in a week.

Security researcher Chaofan Shou spotted the exposure first and posted about it on X where it racked up over 28 million views within hours. The leak happened because Anthropic's Claude Code npm package accidentally included a source map file: a debugging tool that pointed directly to a zip archive of the full codebase hosted on Anthropic's own cloud storage. Anyone could download and extract it. The exposed code revealed how Claude Code's agentic "harness" works, the software layer that sits around the underlying AI model and governs its behavior. It also uncovered a feature called KAIROS that allows Claude Code to run as a persistent background agent even when users are idle and a self-healing memory architecture designed to prevent context overload in long sessions. Anthropic called it "a release packaging issue caused by human error" and confirmed no customer data or credentials were exposed. The code has since been pulled from npm but remains widely distributed across GitHub forks.

Claude Code only launched in May 2025 but has already climbed to the top of the AI coding tool rankings, with 41% of professional developers now using it daily and $2.5 billion in annualized revenue. What makes Claude Code stand out is not just the underlying mode, it's the orchestration layer around it, the part that governs memory, tool use, and autonomous behavior. That's exactly what leaked. Competitors now have a free, detailed look at how Anthropic solved problems that nobody else has cracked at this scale. For a company that's preparing to go public and positioning itself as the safety-first AI lab, accidentally shipping its own source code twice in one week is the kind of trust problem that no PR statement can fully patch.

Evolving AI: Most “AI assistants” just talk. This one does the work.

ElevenLabs Agents v2 lets you create real-time voice agents that don’t just respond, but execute tasks across your systems.

Start now →

Evolving AI: OpenAI has closed its largest funding round ever, raising $122 billion at an $852 billion valuation while still a private company.

OpenAI has raised $122 billion at an $852 billion valuation in what is now its biggest funding round ever. The raise was co-led by SoftBank and Andreessen Horowitz with participation from Amazon, Nvidia, Microsoft and several other major players. About $3 billion of that total came from everyday retail investors through bank channels, an unusual move for a private company of this scale. OpenAI is also being included in multiple ARK Invest ETFs allowing more people to hold exposure to the company ahead of its expected IPO.

The company's press release reads almost like an early IPO filing. It highlights $2 billion in monthly revenue and claims it is growing four times faster than Alphabet and Meta did at comparable stages. Business use now accounts for 40% of revenue and is expected to match consumer revenue by end of 2026. Its ads pilot alone is reportedly pulling in over $100 million in annual recurring revenue after fewer than six weeks. OpenAI also expanded its revolving credit facility to $4.7 billion though it remains undrawn.

OpenAI is still losing money. Its projected loss for 2026 sits around $14 billion and the company has never turned a profit. Yet here it is raising $122 billion at nearly a trillion-dollar valuation. That gap between the financials and the hype is exactly what makes this round so interesting to watch. By letting retail investors in through bank channels and joining ARK ETFs, OpenAI is quietly building the kind of broad shareholder base you need before ringing the bell on an IPO. The company is reportedly targeting a $1 trillion valuation when it goes public, likely in late 2026, which would make it one of the largest listings in history.

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

Evolving AI: A new poll finds that 15% of Americans are open to having an AI as their direct supervisor, a small but telling shift in workplace attitudes.

A Quinnipiac University poll surveyed 1,397 American adults between March 19 and 23, 2026 and found that 15% would be willing to work under an AI supervisor. While the majority still prefer a human boss that number reflects a real cultural shift happening in real workplaces right now. Amazon has already deployed AI workflows to take over middle management duties and laid off thousands of managers in the process. Workday launched AI agents that handle expense approvals on employees' behalf. Uber engineers even built an AI model of their own CEO to screen pitches before they reach him.

Among employed Americans 30% were either very or somewhat concerned that AI would make their specific job obsolete. That anxiety is grounded in something real. Speculation about one-person billion-dollar companies (run almost entirely by AI agents) is no longer pure science fiction.

The poll numbers almost undersell what's already happening. LinkedIn job postings with "manager" in the title dropped 12% year-over-year in early 2026, and a Harvard Business School study found that 6 in 10 managers spend more than half their time on tasks AI can now automate. This week, Jack Dorsey published a piece arguing that AI should replace the coordination layer of management entirely, days after Block cut nearly half its staff. The real shift isn't about whether people are comfortable reporting to an algorithm. It's that the traditional path from employee to team lead to manager is quietly being dismantled, and most workers haven't been told. An estimated 1.5 million U.S. management jobs could be at risk by 2030. When fewer humans sit between leadership and the rest of the workforce, accountability gets murkier and career ladders get shorter.

QUICK HITS

👀 Ring App Store Unleashes AI Revolution, Transforming 100 Million Cameras into Versatile Guardians.

🚫 The New York Times drops freelance journalist who used AI to write book review.

🇨🇳 The world’s largest humanoid robot maker is going public.

🍺 A man used AI to call 3,000 Irish bartenders to track the cost of Guinness. Now pubs are lowering their prices to compete.

🧬 MIT researchers use AI to uncover atomic defects in materials.

🤖 Salesforce announces an AI makeover for Slack.

 *partner link