Good morning, welcome to this week’s Lore Brief, your 3-minute brief of the most important moves in AI and tech.
This issue is brought to you by Factory, the fastest way to ship software with autonomous engineering agents.
Trump calls AI doom talk a hoax → The President said claims that AI will take over the world are a hoax, and blamed a “sick conspiracy” against data centers that only China enjoys. The remarks followed Anthropic chief Dario Amodei’s call to slow development. Trump argued that extra rules would kill U.S. lead. Nvidia’s Jensen Huang later agreed the AI doomers have gone too far. Read more here →
A ChatGPT co-inventor ships a model that refuses to chat → Diogo Almeida, who helped build the training methods behind ChatGPT, came out of two years of stealth with TypeSafe AI and a model called Jev. Jev is made for fast yes-no and multiple-choice decisions inside software, not for writing text. Almeida’s launch post passed 30 million views on X within days. Read more here →
Claude folds work and chat into one place → Anthropic is merging its separate work product into ordinary Claude, so a quick question and a long report now live in the same thread. Claude can keep going after you close the laptop and is supposed to ask when something is unclear. Paid users can also draft documents, slide decks, and designs in that same conversation. The change starts with Pro and Max over the next few weeks. Read more here →
An open-source cancer atlas maps treatments, research gaps, and 1,100 ideas to advance oncology:
Raindrop raises $50M to protect against agent failures:
Jev model use case:
Meta’s Muse agent arrives on the Mac → Mark Zuckerberg said Muse, Meta’s personal AI agent, now has a Mac app. It can work across files, calendar, notes, and messages if you grant access. The agent already lived on phones and the web, and Meta’s pitch is that one thread follows you from device to device. You decide what it can see. Read more here →
Figure robots enter 30 homes they have never seen → Figure rented 30 Bay Area houses and sent humanoids in with no extra training on those rooms. The robots were asked to pick up toys, fold towels, and make beds using the furniture already there. Figure says a version trained from scratch finished the full job 9% of the time, while the new system reached 56%. The point is whether a humanoid can walk into a strange house and start useful work on its own. Read more here →
ChatGPT moves into Microsoft Word → OpenAI put ChatGPT inside Word so rough notes can become drafts, and messy paragraphs can be cleaned up without leaving the file. It also looks for formatting problems. The add-in is available on every plan, including Free, with usage limits. Read more here →
Astra recovers a 1941 German Army Enigma dispatch → Researcher Carter Leffen says GPT-6 Astra recovered a German Army Enigma message from July 10, 1941 that had remained unsolved in a modern cipher archive. The model searched sources, built a simulator, wrote attack code, and tested keys over about 10 hours. Read more here →
