
Alibaba's Wan Ships a CLI So AI Agents Can Generate Videos Autonomously
Alibaba's Wan platform launches a CLI that lets AI agents generate images and videos programmatically, with credits synced to your account
20 stories

Alibaba's Wan platform launches a CLI that lets AI agents generate images and videos programmatically, with credits synced to your account

Sarvam AI releases the first open benchmark for multi-speaker speech recognition across all 22 scheduled Indian languages, exposing major gaps in commercial APIs

Tencent's WorldClaw uses an LLM agent pipeline to turn a text prompt into a fully editable, game-ready 3D open world — no video, no Gaussian splatting.

An AI model developed a proof establishing that clever stepsize schedules alone cannot push plain gradient descent to the optimal convergence rate.

Alibaba's Wan-Animate-2 ditches pose skeletons entirely, achieving real-time streaming character animation at 24 FPS with open weights under Apache 2.0

Microsoft's MAI-Image-2.6 debuts at #2 on Arena's text-to-image leaderboard, closing within 45 points of GPT Image 2 across every major image category.

Browserbase ships Stagehand v4, moving browser state management into an extension to fix remote automation reliability and unify TypeScript, Python, and Go SDKs

Higgsfield's 110-minute AI feature film, made for $2M with licensed celebrity likenesses, is now fully open-sourced — prompts, assets, and all.

An unreleased Claude research model pushed a 40-year-old mathematical lower bound from 41.6% to 67.2%, using a multi-agent swarm running inside Claude Code

Meta's 30B agentic vision model lands with Unsloth GGUF quants that fit inside 14GB VRAM and still fire off hundreds of tool calls without breaking.

OpenAI launches GPT-5.6-Cyber and a two-tier Daybreak program, giving vetted defenders a model that answers 95% of advanced exploit requests — up from 1.5% for the standard model.

OpenLLM by BentoML lets you self-host any open-source LLM as an OpenAI-compatible API endpoint, from local GPU to production cloud, with a single command.

fal launched a MiniMax H3 LoRA trainer and shipped Realism People, an open-weight adapter that sharpens skin, faces, and human motion.
We ran DeepSeek V4 Flash through 54 private debugging attempts, then compared its reliability, cost, latency, token use, and patch behavior with nine other frontier models.

Dezerv built a post-call analytics system on Sarvam's Speech to Text API, cutting transcription error rates by 20% across 3.6 lakh minutes of client calls monthly.

Cohere Labs open-sourced a 2.4B vision-language model tuned for documents, charts, and grounding, with native-resolution input up to A4-at-200-dpi.

Lovable gives every published app an auto-generated, always-current security page — removing the biggest friction point in B2B sales for vibe-coded software

Meta open-sources Muse Glimmer, a 30B agentic model distilled from Muse Spark 1.2 that runs on a single consumer GPU under Apache 2.0

Ant Group's open-weight 7.9B MoE runs on a MacBook, activates only 1.3B parameters per token, and ships in BF16, FP8, and INT4.

Mindwalk turns your repo into a 3D city and replays your Claude Code or Codex agent sessions as glowing light moving across it.