vLLM v0.21.0: Spec Decode for Reasoning Models — Upgrade Now
vLLM v0.21.0 ships thinking-budget-aware speculative decoding, KV offload + HMA integration, and a Blackwell MLA backend. Here's what changed and how to upgrade ...
OpenAI Codex Appshots & Goal Mode: Use Them Now
OpenAI's May 22 Codex update ships Appshots for macOS, Goal Mode GA, and locked Mac ...
Google Antigravity SDK: Build Python Agents Now
Learn to build custom Python agents with the google-antigravity SDK. Tools, MCP integration, safety policies, ...
Gas Town by Kilo Is Live: Run 20+ AI Agents Now
Gas Town by Kilo went GA on May 19. Run 20-30 Claude Code agents in ...
Modern Web Guidance: Fix Your AI Agent’s Stale Web Code Now
Google's Modern Web Guidance bundles 128 Chrome-verified web skills into your AI coding agent with ...
Claude Cache Diagnostics: Stop Paying for Cache Misses You Can’t See
Anthropic's new cache diagnostics beta tells you exactly why your prompt cache missed. One header, one parameter — stop losing 90% cost savings ...
GitHub Copilot Agent Tasks API: Automate Coding Agents Now
GitHub’s new Agent Tasks REST API lets you trigger Copilot’s cloud coding agent from any ...
Gemini 3.5 Flash Beats Gemini Pro: Migrate Your API Now
Gemini 3.5 Flash outperforms Gemini 3.1 Pro on every agentic benchmark, runs 4x faster, and ...
Vivaldi 8.0: The Browser That Said No to AI in 2026
Vivaldi 8.0 launches today with its biggest design overhaul — zero AI. While Chrome silently ...
KotlinConf’26: Kotlin Toolchain, Koog Stable, What to Do Now
KotlinConf'26 keynote shipped Kotlin Toolchain, Koog production-stable AI agents, and stable context parameters. Here's what ...