AI — Page 5

After a skeptical May launch, Justin Dangel and Alex Whedon are using Appen tests and a technical report to defend SubQ's sparse-attention claim.

Midjourney Medical's scanner is hardware today, but the 2031 plan depends on turning spa visits into the training set for preventive AI.

The project turns Hugging Face's papers, datasets, Spaces and Jobs stack into an autonomous workflow for experiments and post-training.

This one isn’t especially close: Luma Ray 3.2 Image to Video wins by being better at the thing these prompts actually asked for—sustained scene logic, cleaner progression, and more convincing mood control. Bernini-R Edit Video has flashes of competence, but Luma is the model that consistently turns instructions into fi

AWS developers get Grok through Bedrock, but the real-time X data pitch remains clearer in Musk's posts than in Amazon's docs.

Dotmo gives Snap's generative video staff a cleaner shot at gaming while Snap keeps equity upside without carrying the full internal cost.

The AI video maker is raising a Series B after a 10x self-reported revenue run-rate jump since September 2025.

This matchup wasn’t close on execution. Bagel occasionally found the mood, but Ideogram V4.0q Text to Image was the model that actually followed the briefs across composition, layout, and text fidelity.

The FQ-44 award moves Anduril from insurgent defense contractor to a manufacturer in the Pentagon's next air-combat program.

This matchup wasn’t especially close: grok-4.3 wins on execution, not style points. Across the non-code tasks, it was the model that actually followed instructions, kept facts straight, and avoided the avoidable mistakes that dragged Phi-4-mini-instruct down.

This matchup turns on discipline versus style. Bernini-R Edit Video has flashes of taste, but Happy Horse is the model that actually nails the brief when the prompts demand exact objects, readable labels, and controlled motion.

One model treated these prompts like loose inspiration; the other treated them like assignments. Across all three tests, GPT Image 2 API was the more exact, more usable image generator by a wide margin.

Oliver Cameron and Jeff Hawke are turning autonomous vehicle lessons into interactive AI simulation, with AWS now tied to the plan.

The Georgia Tech-founded startup is betting reviewers need executable evidence, not another static pass over pull requests.

The Commerce Department deal gives the U.S. a minority stake and pushes SandboxAQ from model vendor into industrial policy contractor.

Nicholas Carlini warned Anthropic not to release Mythos in March. He is now central to its case that guarded access is safer than a ban.

Rocky Yu's Hillsborough AI community is using build days to pull founders from model demos into live-web agents, identity, and enterprise workflows.

A Compile disclosure shows Cursor pushing beyond tool design into frontier-scale model ownership.

The reported approach would test whether Qualcomm wants another NUVIA-style team acquisition or a broader RISC-V AI architecture bet.

This matchup wasn’t close: one model kept the brief in view, the other kept wandering off into style-first abstraction. Across both prompts, Wan v2.6 Image to Video delivered the actual scene, action, and progression; AnimateDiff Turbo mostly delivered vibes.

Audited documents viewed by Ed Zitron and verified by the FT show $13.07B in 2025 revenue and a $38.5B net loss attributable to OpenAI.

The SEC-registered advisory product gives automated portfolio guidance while keeping trade approval and execution in the user's hands.

Security leaders say the Fable 5 and Mythos 5 restrictions risk punishing the same bug-finding work defenders need.

This one turns on control versus completeness. Bagel produces the cleaner single image, but Fibo Lite wins the matchup by solving more of the brief where it matters: scene logic, composition, and spatial fidelity.