The through-line
Agents are moving from experimental to production: Claude merges chat and workspace tools, GitHub rewrote 800K lines of Rust using agents, and xAI/Google are adding memory and monitoring. Simultaneously, infrastructure is hardening—fault tolerance, deployment retention policies, and security detection are catching up to match agent adoption.
Issue signature
The Wire
- Claude Cowork and chat are now one Claude
- AI News, Volume 47: Agents Get Memory and a Watchtower
- Cute Critters Come to the Cloud: ‘Aniimo’ Launches on GeForce NOW
- Making AI-Assisted Claims Independently Challengeable: Publication Authority and a Protocol for Falsifiable Publication Records
- EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents
- Migrating the GitHub Copilot runtime to Rust, using Copilot
- When scanners miss the attack: how Cloudflare Client-Side Security protects storefronts
- Improving HCLS AI reasoning with open-source agent skills
- Fault tolerant distributed training on Amazon EKS using NVRx
- Hobby projects now retain fewer deployments to free up storage