Category intelligence
512 current items analyzed and ranked.
Executive synthesis
The Midjourney Medical launch dominated the day, as the generative-AI company unexpectedly unveiled a medical division and full-body scanner device. A technical deep dive on the Midjourney Scanner drew 9.4M views, and swyx compared the reveal to the original iPhone, framing 40-100x ambitions.
Primary evidence
By @midjourney
Continuing our coverage of Midjourney's hardware reveal, Midjourney announces a new division called Midjourney Medical.
By @JeffDean
Jeff Dean summarizes a forthcoming IEEE Micro paper on Google's TPU evolution from v2 to Ironwood, detailing 30x energy-efficiency gains, cooling shifts, interconnect changes, and pod scaling.
By @midjourney
Continuing our coverage of Midjourney's hardware reveal, Midjourney shares a technical deep dive into its new Midjourney Scanner medical device.
By @jerryjliu0
Jerry Liu announces LiteParse v2.1, claiming the fastest open-source PDF-to-markdown parser, more accurate than other model-free parsers on three benchmarks, fully open source under Apache 2.0 with multi-language bindings.
By @OpenAI
OpenAI announces a NEJM AI study with Boston Children's Hospital and Harvard showing o3 Deep Research helped clinicians solve previously unsolved rare pediatric disease cases.
By @perplexity_ai
Perplexity introduces Brain, a continuously learning memory system that builds a context graph for its Computer agent, available in research preview for Max subscribers.
By @levelsio
Levelsio relays the SF sentiment that software is commoditized because AI makes building apps fast, prompting talent to shift toward hardware, tying it to Midjourney's health move.
By @ClementDelangue
Hugging Face CEO Clement Delangue argues that post-hoc API guardrails are inadequate for frontier model safety, advocating staged release, strong evals, open-source support, and independent evaluation.
Greg Brockman announces that OpenAI's Codex can now be taught by demonstration.
Greg Brockman welcomes Noam Shazeer to the OpenAI team, signaling a high-profile talent hire from a leading transformer researcher.
By @sama
Following the Reddit buzz about the Shazeer hire, Sam Altman says Noam Shazeer is one of the people he most wanted to work with since OpenAI began, confirming the long-awaited hire.
By @vllm_project
vLLM congratulates poolside on Laguna M.1, an open-weights agentic coding model with a 225B sparse MoE (23B active), 256K context, and day-0 vLLM support.
