AI — Page 4

Jane Manchun Wong's post points to Meta Pay alongside Stripe's Link and Shop Pay; Meta's launch post names Link and says Shop Pay is coming.

The company calls it the first result from its molecular biology lab; the post offers a candidate discovery, not a demonstrated biological function.

The former Hume AI CEO is credited on Google's release of models for custom voice design and line-by-line performance control, with 30-second voice replication built in.

Founder Roger Jiang says larger pretraining runs reduced held-out prediction errors after adaptation, though the study did not measure real-world task success at scale.

NVIDIA says its 100M-parameter model scored 14.72% DER on Voice Arena's initial benchmark, while its gated Hugging Face preview is limited to internal evaluation.

TestingCatalog reports Free, Prototype and Accelerate tiers for API builders, but the screen does not establish a launch date or app-hosting service.

The one-prompt run took 86 minutes, Sael says, and drew on files from his earlier projects rather than a blank workspace.

The September 22nd clips put two models through Higgsfield's game-building workflow, but offer no controlled performance comparison.

OpenAI's proposal covers evaluations, incident reports and recursive self-improvement, while stopping short of licenses or mandatory prerelease review.

Researcher Geby Jaff directed ten Claude agents to produce C-HD and a Lean proof; its runtime claim applies to a limited range of graphs.

Optimum Intel documents a Diffusers deployment route, while an OpenVINO GenAI test rejected an exported model and Intel supplied no model-specific performance results.

Founder Qingshan Fu built on years of HIV structural research; the candidate and the seven-month timeline remain company-reported.

The new models reach ChatGPT Work, Codex and the API, with Luna priced at $0.10 per million input tokens.

Markus Persson's reversal follows months of criticism, a hiring struggle and experiments using Claude for development work.

The cited benchmark pages still show Gradium's existing model at 217ms to 236ms, leaving the beta's test conditions to be reconciled.

Alexandr Wang is adding another commerce rail to Meta's two-week-old agent, even as Amazon shows merchants can still refuse the bot.

Tuesday's rumored double launch became three official models, with OpenAI pricing GPT-6 Sol at $2/$10 and Luna at $0.10/$0.50 per million tokens.

Reuters says DeepSeek and other Chinese AI developers were invited to make statements at the September 23rd meeting, but DeepSeek's attendance, delegate and remarks remain unconfirmed.

Tripo AI's model adds regional mesh editing and four variants per prompt, 20 days after the company announced 3 billion yuan in funding.

The preview accepts text and reference images, keeps editing context across turns, and undercuts Hy Image 3.0's listed API price by 25%.

Samsung and SK Hynix rose on September 22nd as broader AI optimism, lower oil prices and softer US yields lifted Seoul stocks. Bloomberg linked Muse to investor sentiment, not disclosed chip orders.

GPT-6 Astra reached 53% accuracy, Gemini regressed, and a previously scoreless OpenAI model suddenly completed 16 runs.

The YC W26 company is extending its on-device stack into hosted inference, claiming 380 tokens per second on GLM-5.3 Flash.

Xiaomi released two omnimodal model repositories, reinforcement-learning resources and a technical report while retaining MiMo-V2.5 API prices.