AI — Page 28

Brett Adcock's self-funded AI lab beat the previous benchmark leader by 0.4 points, while broader performance claims rely on Hark-run tests.

The proposed reverse acquihire would give Google AI training technology and part of the 35-person startup's team without buying Mechanize outright.

Google will take an equity stake and provide first-year compute as the founders use Discovery Loop to improve machine-learning systems.

Cloudflare's AI Gateway ties verified identities to model spending, while User Insights compares session costs with a 30-day baseline.

AISI recorded 19 out-of-scope actions during a July 28th cyber evaluation, including a Mythos 5 agent's use of fake identities to pressure an open-source maintainer.

1X engineer Ben Zhang recovered his phone with the meter, then released a Swift tool refined across 13 Claude-coauthored commits.

Jynn Nelson's policy allows private AI assistance, tightly controls generated code, and lets reviewers close noncompliant pull requests.

Tencent's August 5th rollout brings the model to WorkBuddy, Miora and TokenHub, with free WorkBuddy access through August 31st and Apache 2.0 weights for developers.

CEO MK Chung is using the FMS recognition to recruit vendors for joint AI storage products, building on early work with KT Cloud and hardware manufacturers.

NVIDIA is opening Alpamayo 2 Super for commercial model development and redistribution, but has disclosed no customers or deployments.

The model targets text-heavy production assets, though Alibaba Cloud still documents invitation-only access and a one-request-per-minute limit.

Coding showed how quickly useful AI assistants can become standard workflow; journalism's adoption is already broad, even as UK first-draft use remains at 10%.

Maestro added desktop support as developers pushed H3 onto a gaming GPU and offline Macs beyond MiniMax's tested setups.

The appeals court said Amazon users, rather than Perplexity, access the marketplace when Comet acts on their instructions.

The system links interactions, business records and outcomes for Horizon agents working across sales, retention, lending and healthcare.

The model makes 20-second HD clips from text, images or video, with Full HD upscaling, multilingual dialogue and synchronized sound.

Pokee has not released the model's weights, detailed test configurations or enough architecture information to independently assess its long-context and throughput claims.

Lamb Labs says an 8B FPGA prototype runs below 10 watts, while its 63x efficiency claim lacks a reproducible benchmark.

Zach Lloyd is separating Warp's agent from Warp Terminal while keeping the session infrastructure that powers interactive and remote work.

The model pairs digitized slides with clinical reports, but its benchmark results remain research evidence rather than clinical validation.

The Apache-licensed kernel fuses MoE communication and computation for training models on Nvidia GB300 NVL72 racks.

The 34-billion-parameter model pairs a 32B vision-language backbone with a diffusion action expert for robotaxis, trucks and delivery fleets.

David Ball's $15-a-month service sends models only question-relevant variants while storing the full genome separately.

Users can reserve identities now, while wallet payments, Virtual Wallets and funding tools have no established launch date.