20d • General discussion
There’s been some big news in the past couple weeks for agentic AI. Here are 10 major updates you might have missed.
A quick roundup of what’s moving the space forward right now.
- OpenAI strengthens ChatGPT Atlas securityOpenAI published an article on how they are continuously securing Atlas and other agents. They are using automated red teaming powered by reinforcement learning to find and patch exploits before they can be abused. Fast response loops and agent security are becoming a serious priority.
2. Claude Code is adding custom agent hooksAnthropic’s founder confirmed that the next version of Claude Code will support hooks and frontmatter for custom agents. This lets developers extend Claude Code with their own agent logic and behaviors. Custom agent workflows are officially coming to Claude.
3. Forbes highlights AI agent overload for small businessesA Forbes article points out that 58 percent of US small businesses now use AI, which is double since 2023 according to the Chamber of Commerce. Many are managing more than a dozen tools at once, creating overhead and complexity. The comparison was like having multiple remote controls for the same TV.
4. Windsurf launches Wave 13 with parallel agentsWindsurf released Wave 13 including free access to SWE 1.5 and true parallel agents. Features include Git worktrees, multi pane and multi tab workflows, and dedicated terminals for command execution. They are clearly betting big on agent based coding.
5. Claude Code is now writing itselfThe creator of Claude Code shared that all recent development was written by Claude Code using Opus 4.5. In the last month alone, 259 pull requests were generated with tens of thousands of lines added and removed. Agents are now running for minutes, hours, and even days at a time.
6. Forbes covers how agents are changing jobs and purposeAnother Forbes piece this week focused on how AI agents are automating routine work across nearly every profession. This is forcing people to rethink where humans actually add value. Agent driven work transformation is now a mainstream conversation.
7. Google publishes 40 AI tips including agent usageGoogle released a guide with 40 practical tips, many of which focus on integrating agents into daily workflows. This is less theory and more hands on guidance for everyday AI usage.
8. New research introduces the Sophia agentA new paper dropped on an agent called Sophia that uses a system level manager to monitor reasoning and decide next actions. It shows fewer reasoning steps on repeat tasks and higher success on harder tasks. It also stores learning episodes and user models. Still early but interesting.
9. Google Cloud releases a 2026 AI agent trends reportGoogle Cloud published a report based on thousands of executives and AI experts. It covers end to end agent workflows, digital assembly lines, customer service automation, threat detection, and the growing need for workforce training.
10. GLM 4.7 arrives in the Blackbox Agent CLIZai’s GLM 4.7 model is now integrated into the Blackbox AI Agent command line interface. Developers can run GLM models directly from the terminal. Early days but more options are always welcome.
That’s the rundown for the latest in agentic AI.
Which one of these do you believe impacts what you’re building or running the most?
0
0 comments
There’s been some big news in the past couple weeks for agentic AI. Here are 10 major updates you might have missed.
powered by
skool.com/ai-agent-mastery-6504
Learn to build, and sell powerful AI Agents, Automations and Infrastructures—automate workflows, boost profits, and scale simply.
Suggested communities
Powered by