AI architecture exposed, China's lunar ambitions, and why good naming matters

Anthropic's accidental leak is dominating headlines as a rare peek into cutting-edge AI architecture, while NASA and China's competing lunar missions reveal a space race heating up faster than anyone expected. We're diving into the week's biggest tech stories—from Google's video generation price wars to Meta's surveillance glasses—with fresh takes on what it all means.

A mysterious autonomous feature called KAIROS lurks in the leaked code—and it changes everything.

Anthropic's accidental source code leak on April 1st, 2026 reveals far more than expected—including KAIROS, a fully-built autonomous feature that runs Claude in the background while you sleep, automatically handling tasks without permission. We dig into what this hidden system means for AI autonomy, privacy, and the future of digital assistants.

OpenAI kills Sora while Kling AI hits $300M annual revenue—the monetization gap is real

OpenAI shut down Sora on March 25th due to unsustainable compute costs, but on the same day, Kuaishou's Kling AI announced it hit a $300 million annualized revenue rate—proving Chinese video model providers have cracked the monetization code that Western competitors can't match. We break down why the infrastructure economics are so different and what this shift means for the future of AI video generation.

AI agents reshape social media while Meta reverses course on fact-checking

Bluesky launches Attie, an AI-powered app that lets you customize your feed exactly how you want it, while Meta abruptly halts its fact-checking exit after human rights concerns forced their hand. Plus, Apple hits record app submissions as AI coding tools explode in popularity, leaving their review process completely overwhelmed.

Pentagon drama, Ray-Ban brains, and why AI is winning the winner-takes-all game

We're diving deep into the week's biggest tech stories: a Pentagon vs. Anthropic legal showdown, Meta's aggressive push into Ray-Ban, brain-computer interfaces, and hyperagents that are reshaping reality. Plus, a sobering conversation about what AI adoption actually means for jobs in a world increasingly ruled by winner-takes-it-all dynamics.

Explosive court documents reveal Zuckerberg and Musk's surprising texts about DOGE and OpenAI.

From leaked courtroom bombshells to hacked FBI directors, this episode unpacks the wildest AI and tech industry drama of the week. We dive into the jaw-dropping texts between Zuckerberg and Musk, breakthroughs in AI music, and Amazon's bold moves against retail giants.

Anthropic's leaked AI model, fake security certs, and a lawsuit that could reshape social media

This week brings seismic shifts in tech: Siri embraces multiple assistants, Meta loses a landmark case over social media addiction, and Anthropic accidentally reveals a major AI breakthrough through a misconfigured content system. We're diving into the chaos of unreleased models, app store disruption from AI agents, and the internet's spiciest new acronym.

OpenAI kills Sora, Spotify credits producers, and AI's reality check

OpenAI quietly pulled the plug on its billion-dollar Sora video tool after months of flashy promises to Hollywood—with zero actual money changing hands. Meanwhile, Spotify launches SongDNA to finally give producers the credit they deserve, while the AI hype machine faces an uncomfortable reckoning.

MiniMax M2.7 brings Claude-level performance at a fraction of the price—plus iPhone hacking and AI agents arguing with themselves.

MiniMax's new M2.7 model delivers Claude Opus-level performance on coding benchmarks while costing 17-21 times less, signaling a major shift in AI economics dominated by Chinese competitors. We're diving into the implications alongside some wild security exploits, AI agents writing better code through debate, and how Google's quietly shipping features while everyone watches OpenAI.

Anthropic's new Cowork feature brings Claude to your desktop while China stops copying and starts leading.

Anthropic just launched Cowork, bringing Claude directly to your machine with persistent memory and local file access—delegate tasks and come back to finished results. Meanwhile, China's tech giants are no longer copying the West; Tencent's ClawBot scales to a billion new collaborators and ByteDance launches DeerFlow 2.0, signaling a major shift in who's innovating in AI.