╔═════════════════════════════════════════════════════════════╗
║ ║
║ ▀█▀ ▄▀█ █▄▀ █▀▀ █ ▀█▀ █▀▀ █▀█ █▀█ █▀▄▀█ ║
║ █ █▀█ █ █ ██▄ █ █ █▀ █▀▄ █▄█ █ ▀ █ ║
║ ║
║ ▀█▀ █ █ █▀▀ ▀█▀ █▀█ █▀█ ║
║ █ █▀█ ██▄ █ █▄█ █▀▀ ║
║ ║
╚═════════════════════════════════════════════════════════════╝
// Tech news at terminal velocity
cat 2026-04-01.md
$ cat TLDR.md
▸ • OpenAI secures a mind-boggling $122 billion in new funding to scale frontier AI.
▸ • Baidu's robotaxis freeze up in Wuhan, causing massive traffic chaos.
▸ • Salesforce injects 30 new AI features into Slack, because your chat app wasn't noisy enough.
In a funding round that sounds like a typo, OpenAI just raised $122 billion to expand its frontier AI globally and invest in next-gen compute. Because training models isn't getting any cheaper, and those GPUs won't buy themselves.
Salesforce is rolling out 30 new AI features for Slack, aiming to make the platform a whole lot more useful. Get ready for your workspace to become even more automated, for better or worse.
Google just released Veo 3.1 Lite, its most cost-effective video generation model yet. It's now available in paid preview through the Gemini API, giving developers a cheaper way to generate pixels on the fly.
Alibaba's Tongyi Lab has released Qwen3.5-Omni, an omnimodal model that handles text, images, audio, and video. It even claims to surpass Gemini 3.1 Pro in audio and video understanding.
Remember when global startup funding was democratizing? The AI boom has violently reversed that trend. American investors are pulling back from international markets to pour unprecedented capital into domestic AI startups, leaving the rest of the world scrambling for scraps.
How autonomous are autonomous vehicles, really? Major robotaxi companies like Waymo, Zoox, and Tesla flat-out refused to tell Senator Ed Markey how often their vehicles require human remote assistance. The silence speaks volumes about the current state of self-driving tech.
GitHub's Applied Science team shares their experience using coding agents to build... more agents. It's a fascinating look at the meta-level of software engineering where developers are learning how to effectively manage AI colleagues rather than just writing code.
In a wild demonstration of AI capabilities, Claude successfully wrote a full FreeBSD remote kernel RCE with a root shell (CVE-2026-4747). It's a stark reminder that AI models are getting dangerously good at offensive security.
Running local models on Apple Silicon just got a massive performance boost. Ollama now supports Apple's MLX framework, allowing for much better unified memory usage and significantly faster inference on your local machine.
Liquid AI has released LFM2.5-350M, a tiny but mighty 350 million parameter model trained on a staggering 28 trillion tokens. Using scaled reinforcement learning, it proves that you don't always need massive parameter counts to get serious performance.
A fleet of Baidu Apollo robotaxis simultaneously froze on the highways of Wuhan, trapping passengers inside and causing massive traffic snarls. It's the ultimate blue screen of death, but in moving traffic.
Because discovering the secrets of the universe isn't enough, CERN engineers have leveled up their internal transport with new superconducting karts. Mario Kart has nothing on particle physicists.