
0xAI
10 posts



AI News from Yesterday - Intel co-authored a whitepaper with Manifold Labs, a Bittensor subnet, showing hardware-enforced confidential computing that secures AI workloads on decentralized infrastructure. - OpenAI is seeking a $10B raise at a $730B valuation, offering investors a 17.5% guaranteed return to attract private equity. - Google released TurboQuant, compresses LLM memory by 6×, running AI up to 8× faster with no accuracy loss, letting powerful models run locally on devices like a 16GB Mac Mini or phones. - Claude now autonomously controls your macOS desktop, opening apps and completing tasks - Meta hired Dreamer’s founders and team, including ex-Google and Stripe execs, to advance its AI agent projects. - OpenAI hired former Meta ad executive Dave Dugan to lead global ad sales. - Anvisha announced the launch of Moda, a collaborative AI design platform that generates brand-aligned slides, assets, and visuals on an editable canvas. - Claude introduced Auto Mode in Claude code to handle file writes and bash commands for you - OpenAI shut down Sora video platform. - HP introduced HP IQ, an on-device AI tool designed to help users interact with personal files such as PDFs and presentations directly from their computers. - Meta found liable in New Mexico case over child safety concerns, ordered to pay $375M in penalties. - Sam Altman announced OpenAI Foundation with $1B initial funding to tackle AI threats and accelerate scientific discovery. - Meta Platforms is introducing a stock option program that would pay top executives only if the company reaches a $9 trillion market capitalization by 2031.



Chinese authorities have barred two Manus co-founders from leaving the country, heightening scrutiny over Meta’s $2 billion acquisition of the agentic AI startup. CEO Xiao Hong and Chief Scientist Ji Yichao are free to move within China but cannot travel internationally.

. @GoogleResearch published a whitepaper introducing TurboQuant, a new compression algorithm that shrinks LLM memory by up to 6x with zero accuracy loss (no dumbing down), achieving up to 8x speed gains on H100 GPUs. The method is set to be presented at ICLR (International Conference on Learning Representations) 2026 cryptointegrat.com/p/google-resea…






