Bitduke

lots of green candles here

Nvidia’s board launched an independent investigation

Nvidia $NVDA CEO Jensen Huang just sent this memo to Nvidia employees

the affair rumors involving Nvidia’s ceo and a suspected Chinese spy turned out to be true! Jensen Huang admitted being unfaithful to his wife, while Nvidia’s board launched an independent review (it says it has no reason to believe company secrets or national security were

i thought we’d get a new model from OpenAI (like gpt-6.2?) soon but it seems we’re getting a gaming console instead

the patent also describes using AI to condense long recordings into shorter summary videos, generating new footage of the speakers saying a summarized script

Siri could become Apple’s biggest ai advantage if it finally becomes a long-horizon ai agent that keeps working after you stop talking Bloomberg reports Apple wants Siri to pursue ongoing goals, closer to Openai’s Dots / Anthropic's Duet / Grok bot, with Gurman suggesting a

<...> 112,000 US job cuts this year have cited AI, per Challenger, Gray & Christmas

Grok bot is evolving

google deepmind is talking about “a genuine phase transition” in scientific discovery and teasing releases in the next few weeks

hacktron’s researchers used ai to break into OpenAI, and now they’re helping it find the next way in OpenAI is continuing its cybersecurity push, and after recently announcing plans to bring Cathedral’s researchers into its next security sprint, it’s now partnering with Hacktron

the strongest common thread is selective computation and reuse. the goal, it seems, is to make growing context cheaper to process, especially useful for long-horizon ai agents

demand is still there, with prices for one-year H100 rental contracts up roughly 21% since march 2026

first OpenAI took a valuation haircut, and now Anthropic is getting the same treatment Anthropic’s prospective ipo investors are discussing $1.2–1.3 trillion, down from $1.4–1.5 trillion (and earlier ~$2T speculation). the appetite is still there, but investors want a price that

it’s happening

impressively smooth motion. curious how much synthetic training contributes to this level of dexterity

«One-shotted with Claude Code in 3 minutes»

another interesting step into biotech as frontier ai labs push further into the field; helping researchers means making sense of messy experiments, which takes more than passing a biology exam OpenAi’s biobasicsbench tests models on actual lab data, from microscope images to

cool

what does “dreaming” mean when talking about ai models? Anthropic’s researchers explored how Claude can use time between tasks to review past sessions, correct bad notes and organize useful knowledge. basically, giving an agent a chance to check what it remembers before those