Latest dispatches

Every published story, ordered for readers, crawlers, and AI citation tools alike.

A chronological archive built for scanning rather than guessing, with visible bylines, stable article routes, and fast jumps back into category archives.

Chronological archiveStable article URLs

Stories: 105
Bylines: 4
Categories: 6
Latest story: Apr 11, 2026

AI Policy/Apr 6, 2026/Updated Apr 11, 2026/7 min read

Google turns India into its public-sector AI template

Google used AI Impact Summit 2026 in India to bundle fiber, cloud, grants, civil-service training, science partnerships, and localized Gemini features into one national rollout playbook.

Maya HalbergStaff Writer

Editorial illustration of a national AI rollout board in India linking subsea connectivity routes, civil-service training, grant funding, and localized Gemini product surfaces.

AI Tools/Apr 6, 2026/Updated Apr 11, 2026/7 min read

GitHub Copilot SDK exports GitHub's agent runtime

GitHub's Copilot SDK public preview is not just another toolkit. It exports the agent runtime behind Copilot CLI and cloud agent into apps.

Idris ValeStaff Writer

Editorial illustration of a GitHub-style app workspace feeding requests into a Copilot runtime layer while custom agents, tool permissions, MCP servers, and session memory branch out behind it.

Ai Tools/Apr 6, 2026/Updated Apr 11, 2026/7 min read

Gemini API Docs MCP turns docs into agent plumbing

Google's Gemini API Docs MCP and Gemini Agent Skills turn documentation into a first-party context layer for coding agents, with Google tightening workflow control.

Maya HalbergStaff Writer

Editorial illustration of Google's Gemini documentation layer feeding live API context into Claude Code, Cursor, and Gemini CLI across one managed coding workflow.

AI Tools/Apr 5, 2026/Updated Apr 11, 2026/6 min read

OpenAI audio models plug the missing agent voice layer

OpenAI next-generation audio models look less like a model drop and more like the missing voice layer for its agent stack, with better transcription, steerable TTS, and live voice-agent docs.

Idris ValeStaff Writer

Editorial illustration of OpenAI's chained voice-agent stack linking transcription, orchestration, and speech output, with a separate realtime lane shown beneath it.

AI Tools/Apr 5, 2026/Updated Apr 11, 2026/12 min read

Google adds Flex and Priority lanes to Gemini API

Google's Gemini API now sells traffic classes directly: Flex cuts costs for background agent work, while Priority buys speed and reliability for user-facing turns.

Maya HalbergStaff Writer

Editorial illustration of one Gemini API gateway splitting agent traffic into a broad low-cost Flex lane and a narrower premium Priority lane inside a dark Google-style developer workflow.

AI products/Apr 5, 2026/Updated Apr 11, 2026/7 min read

Google Vids turns Veo into a free entry point

Google Vids just got Veo 3.1 for any Google account, but the bigger story is the bundle: free clips, Chrome recording, YouTube export, and a tier chart with trust issues.

Idris ValeStaff Writer

Editorial illustration of Google Vids acting as a central hub for Veo 3.1 clip generation, Chrome screen recording, YouTube export, and a ladder of free-to-paid account tiers.

AI Tools/Apr 5, 2026/Updated Apr 11, 2026/6 min read

GitHub Copilot cloud agent gets enterprise guardrails

GitHub turned Copilot cloud agent into a branch-first repo worker, then added signed commits plus org-wide runner and firewall controls that stricter teams actually need.

Maya HalbergStaff Writer

Editorial illustration of GitHub Copilot cloud agent working on a branch while enterprise control panels show a verified commit badge, runner policies, and firewall settings around the repository.

AI Products/Apr 5, 2026/Updated Apr 11, 2026/8 min read

AWS puts frontier agents on security and ops duty

AWS has tied its frontier-agents pitch to two GA products: one for autonomous penetration testing, one for incident response across AWS and beyond.

Idris ValeStaff Writer

Editorial illustration of AWS frontier agents as twin work lanes: one linking penetration-testing context into a validated attack chain, the other correlating incident telemetry across a cloud operations board.

AI Infrastructure/Apr 4, 2026/Updated Apr 11, 2026/6 min read

vLLM 0.19.0 changes long-context cost math

vLLM 0.19.0 pairs CPU KV offloading, zero-bubble async speculative decoding, and Gemma 4 support in a release that changes long-context serving economics.

Maya HalbergStaff Writer

Editorial illustration of a modern inference serving floor where active request streams stay on GPU racks while older KV cache blocks spill into a cheaper CPU memory tier behind them.

Open Source AI/Apr 4, 2026/Updated Apr 11, 2026/6 min read

Netflix VOID model removes the object and the aftermath

Netflix has released VOID, a video-editing model that removes objects and their physical aftermath, but using it still means masks, two passes, and 40GB-plus VRAM.

Idris ValeStaff Writer

Editorial illustration of a removed person leaving behind a physically updated scene, with a guitar dropping through empty space and a quadmask-style overlay marking object, overlap, affected regions, and background.

AI Tools/Apr 4, 2026/Updated Apr 11, 2026/8 min read

Qwen3.6-Plus pitches repo-scale coding agents

Alibaba pitches Qwen3.6-Plus as a 1M-context model for repo-scale coding agents. The real test is whether that yields calmer workflows, not prettier benchmark theater.

Maya HalbergStaff Writer

Editorial illustration of a sprawling software repository, terminal and UI surfaces, and a Qwen-led agent execution loop holding a long coding job together.

AI Tools/Apr 4, 2026/Updated Apr 11, 2026/6 min read

Microsoft gives Foundry its own multimodal stack

MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 turn Foundry into more of a Microsoft-owned multimodal stack, not just a shelf for other labs' models.

Idris ValeStaff Writer

Editorial illustration of Microsoft Foundry becoming a first-party multimodal control surface, with transcription, voice, and image lanes flowing into one Microsoft-owned stack.