OpenAI's Jalapeño chip hits nearly 700 tokens per second in early tests
First internal benchmarks for OpenAI and Broadcom's custom inference chip show massive speed gains over Nvidia hardware, though independent access is still missing.
First internal benchmarks for OpenAI and Broadcom's custom inference chip show massive speed gains over Nvidia hardware, though independent access is still missing.
The European Union has formally added VK, the MAX messenger legal entity, and other domestic tech firms to its sanctions list, following earlier unannounced removals from Apple's App Store.
Private buyers in Russia can pre-order the UMO 5 electric crossover starting at 2.59 million rubles, complete with heavy Yandex software integration and cold-weather packages.
Security researchers and Japanese police dismantled the XCPE botnet, built using a backdoor allegedly baked directly into firmware by an Android equipment maker.
Marvell Technology is acquiring Lumentum in a $115 billion bet on optical components for AI data centers, triggering a sharp stock drop over heavy debt.
GitHub is moving to block the risky pull_request_target trigger in public repositories by default starting December 2, following a brief one-month delay.
OpenAI updates ChatGPT Work to handle website logins and trigger tasks from Gmail, Slack, and GitHub.
Hundreds of Microsoft employees are voluntarily logging their monthly AI tool costs in a shared spreadsheet, revealing bills ranging from a few dollars to $28,000.
Researchers at Loughborough University paired a tiny optical resonator with a fiber loop to output stable millimeter-wave frequencies, though shrinking the bench-scale rig into practical hardware remains a hurdle.
Built on Electron by Y Combinator alumni, Orca gives developers a free desktop cockpit to run terminal AI agents across isolated git worktrees.
Researchers uncovered a contactless payment vulnerability that lets expired Visa cards authorize transactions if issuers skip expiration checks.
The Russian government signed a decree to organize artificial intelligence tasks into two existing departments without hiring new staff.