All Stories — Page 50

The released GLM-4-Voice derivatives score up to 77.1% on GSM8K, though larger cascaded systems still lead the benchmark.

Parakeet Redux runs at a claimed 113x realtime on eight CPU cores, with sharper multilingual scores and a measurable penalty in noisy audio.

AJ Asver's internal test reduced compliance model costs from $2.89 to $0.25 per alert, with uncertain cases routed to a frontier agent.

The AI safety startup claims up to 2.8x TRL throughput, while its modified Apache license adds conditions for larger commercial users.

Eclipse built the five-month-old startup with former Meta AI executive Amir Frenkel, then led a round packed with industrial and AI operators.

The April financing funds a virtual finance employee for invoices, purchase orders and lender draws; Kos does not construct data centers.

The Rome AI lab trained its first model in six weeks and released the weights, base model and vLLM serving code under Apache 2.0.

The vision-equipped companion uses a familiar household object to make always-present AI easier to place in the home.

Ground stops spread across Northeast airports after redundancy failed at the Philadelphia air traffic control facility.

The German indie developer says a September 17th post about plain-language Postgres search changed his life and led to the move.

The beta reverses a three-day-old discovery that hinted at Mac-style endpoint monitoring on research iPhones.

Developers can move sessions between local machines and persistent cloud environments without leaving Devin's command-line interface.

Scattered launch-week posts show vertical lines on new screens, with at least two owners saying their phones were replaced.

Augustus Doricko's El Segundo startup will use the Series B to expand atmospheric research and lower the cost of producing freshwater from clouds.

Tom Alder's new app keeps messages on-device and adds AI drafts, tagging and reminders, with users admitted a few at a time.

Foundation Capital led the seed round for Joseph Zhang and Sophia Kim's platform, which pairs digital pack openings with physical trading cards.

The final release pairs 1,274 hours of household footage with arm-tracking data, though folding laundry dominates the corpus.

Founders Dan Lahav and Omer Nevo built Irregular to test frontier AI safely; a May configuration error exposed the difficulty of that job.

The U.S.-only feature lets Plus and Pro users connect credit reports, receive monthly updates and question ChatGPT about factors affecting their scores.

Code2Video tests agents on 168 motion-design briefs, then uses HeyGen's own judge model to score timing, composition and craft.

The group is threatening to expose Cl0p's payment records, including the companies that paid and the Bitcoin addresses used.

Morgan Linton's 92-run sweep found lower effort made Meta's model slower and less accurate, while Astra and Fable 5.1 fully passed all 23 tasks.

The dispute puts platform permission, customer credentials and Amazon's $68.6B advertising business at the center of agentic commerce.

The former xAI and Google staffer is forming Moir around online identity proof; Bloomberg says the financing talks remain prospective.