Good morning 👋 When the creators of HBO’s “Silicon Valley” say tech has gotten too bizarre to parody, things have officially reached peak absurdity. Just when you thought startups can’t get stranger, Y Combinator has backed a product so outrageous everyone thought it was satire. Meet Chad IDE, a brainrot coding environment that lets you gamble and swipe Tinder while your vibe coding AI assistant does the actual work. Yes, really.
🦾 AI agents make terrible employees, yet some companies are profiting. Businesses deploying digital workers have struggled to keep them from hallucinating projects, scheduling fake meetings, and flooding inboxes with imaginary updates. Total chaos, but it’s paying off for early enterprise adopters. Financial services provider BNY runs over 100 agents, growing capacity. Walmart says AI agents shave months off product launches. Salesforce claims a 5x jump from agent-powered tools. Deploying AI agents isn’t pretty, but once you crack the code, it’s payday.
🔎 OpenAI’s privacy nightmare just got real. They’re fighting a court order to hand over 20 million private ChatGPT conversations to the New York Times. The Times initially demanded over a billion chats, claiming users were bypassing its paywalls. OpenAI fired back, arguing 99.99% of the chats are completely irrelevant and warning that releasing entire conversations is a huge privacy risk. Even partially anonymized, handing over two years of chat logs sets a dangerous precedent. And just like that, everyone is going to want self-hosted LLMs.
🤜🤛 Together with
😎 Private AI
Lemony puts you in controI of AI. It’s private and secure – no cloud access required.
Use Lemony for fast, compliant data access. Plug it in, sync your data, and run AI agents and apps locally.
- Create your own custom LLM.
- Zero risk of AI training on your data or prompts.
- Fixed cost, with no limits on tokens or APIs.
Used by teams at Nvidia, IBM, and many more for a secure AI setup.
🐕 Claude is now running around in a robot dog body. Anthropic gave its AI control of a Unitree Go2 quadruped and watched it quickly figure out tasks that people struggle with, like navigating physical spaces and locating objects. The experiment suggests that letting AI wander around the physical world dramatically speeds up learning. Who could’ve guessed? Anthropic’s own red team warned that letting AI physically interact with robots could get dangerous fast. Naturally, they did it anyway.
🏃♂️ Could America win the AI Cold War just by hoarding GPUs? China makes great open models and has endless renewable energy, but America controls 75% of the world’s AI compute. Without extreme ultraviolet (EUV) machines, Chinese factories struggle to reliably produce advanced chips, resulting in a 95% failure rate for Huawei’s flagship 910C. To cope, labs stitch together clusters of homegrown chips, creating data centers so power-hungry the government has to subsidize their electricity. The shortage is so severe Beijing is rationing chips, prioritizing national tech giants while startups like DeepSeek face delays. China has smarter models, America has the silicon; it’s still anyone’s game.
🔋 AI’s insane energy demand might accidentally save renewables. Despite America’s backwards energy policy, renewables are about to explode over the next five years. UBS predicts a “boom cycle” with battery storage surging 40% globally in 2026. Power-hungry AI data centers are pushing the grid to its limits, driving growth in solar and wind, with batteries stepping in every time the sun quits. Now the highly-profitable US battery market is attracting fierce international competition. AI’s energy addiction is suddenly renewables’ best friend.
fun stats
raised by AI startup Parallel to build web search for AI agents 🔍 Led by the former CEO of Twitter, they aim to capitalize on the shift as AI agents become the web’s primary users.