
Google's Gemini Omni Lets You Edit Video Through Plain-English Chat
Gemini Omni Flash brings conversational video editing to the Gemini app, letting subscribers create and remix clips through natural language — no timeline required.
20 stories

Gemini Omni Flash brings conversational video editing to the Gemini app, letting subscribers create and remix clips through natural language — no timeline required.

Perplexity's new query-aware compression cuts context tokens by up to 70% while boosting answer accuracy — now live in its apps and API.

Nous Research's Hermes Agent gains 400+ browser skills from Browserbase's open catalog, cutting per-task token costs by up to 45% through reusable web playbooks.

Zed v1.3.5 lets you run Claude Code, Amp, and any CLI agent as a managed thread in the editor sidebar, a timely escape hatch from Anthropic's new ACP pricing.

Exa raises $250M at a $2.2B valuation to build the search engine AI agents actually need — not humans

Firebase Agent Skills expand to Android, iOS, and Flutter, adding Crashlytics debugging and Remote Config support directly inside your IDE
Google launches four AI creative tools at I/O 2026: Google Pics, a revamped Flow with agentic editing, Stitch's real-time design streaming, and section-level music editing in Flow Music.

Cohere open-sources Command A+, a 218B MoE model that runs on just two H100s with near-lossless 4-bit quantization and built-in citations

An independent developer built a native macOS SSH client for Nous Research's Hermes Agent, keeping the remote host as the sole source of truth.

PegaFlow moves vLLM's KV cache into a standalone Rust daemon, surviving restarts and delivering up to 72% higher throughput by sharing cache across instances and nodes via RDMA.

The VeRL-Omni Team pre-releases an open RL post-training framework for diffusion and omni-modality models, built on verl and vLLM-Omni.

Google I/O 2026 reframes AI tooling as infrastructure: Gemini 3.5 Flash, Antigravity 2.0, WebMCP, and an Android migration agent signal a full shift to autonomous development workflows.

Google AI Studio is getting a standalone mobile app, letting anyone build and ship native Android apps from their phone using just a text prompt.

Biohub open-sourced ESMC-6B, a 6-billion parameter protein language model with weights, sparse autoencoders, and fine-tuning support for downstream biology tasks.

A 13B open-weight LLM trained exclusively on pre-1931 text opens a new front in AI research: contamination-free generalization and historical forecasting.

GraphGen uses knowledge graphs to generate targeted SFT training data, closing knowledge blind spots in LLMs with structured, diverse QA pairs
GRAM makes recursive reasoning probabilistic, letting a 10M-parameter model beat deterministic baselines on Sudoku, ARC-AGI, and multi-solution puzzles by sampling multiple latent reasoning paths in parallel.

Memori turns agent execution traces into structured memory, hitting 87% LoCoMo accuracy while using under 3% of full-context tokens.

A Markdown-only skill layer turns Claude Code and Codex into an overnight research team that drafts, critiques, and rewrites ML papers.

Microsoft ships a single plugin that gives coding agents 27 Azure skills plus 200+ MCP tools across GitHub Copilot, Claude Code, Cursor, and more.