- 1 The Age of Prompting Agents Is Over A few days with Fable 5 convinced me that prompting agents is ending fast. Anthropic has been telling us this for months, they were just living in the future because they had Mythos-class models before the rest of us. Here is the workflow that replaced prompting for me, the talk that made it click, and a draft of what a fully hands-off setup could look like.
- 2 The US Government Is Choosing Who Gets AI Now GPT 5.6 shipped this week, but only to a handful of companies the government signed off on, the same thing that happened to Anthropic's Mythos. Meanwhile China is reselling Claude at a 90 percent discount and Coinbase just cut its AI bill in half by switching to Chinese open models. Here is why the ban strategy could be what costs the US its lead, and the opening it hands to real open source.
- 3 AI Has No Taste, and We're Finally Fixing It An agent writes a thousand lines of solid code in thirty seconds and still ships a UI drowning in purple gradients. The reason is the same one every lab is now racing to fix. Code has an error log. Taste does not. Here is the new kind of eval that is starting to give agents taste, why Meta drafted thousands of engineers to grade data, and what happened when I pointed a self-improving loop at real design references.
- 4 I Tried to Stop Prompting My Agents. Here Is What Loops Are Actually Good At. Everyone says to stop prompting your agents and build loops that prompt themselves. I tested three kinds of loops on real work. Two are genuinely great at the backlog nobody gets to. The fully autonomous one polished my app beautifully and could not invent a single new feature. Here is what that gap tells you about the job that stays human.
- 5 One Agent or a Team of Agents? I Built the Team to Find Out. OpenRouter showed that fusing cheaper models beats frontier models like Fable 5. I tested whether the same trick lifts real coding work, not just research. My first multi-agent setup lost to a single agent. Three iterations later it clearly won. Here is what changed and why.
- 6 Claude Fable 5 (Mythos): Impressive, and Concerning Anthropic's first Mythos class model is the biggest capability jump I have seen, and the most expensive, and the way they are rolling it out should worry you. My full review: a game built from scratch, UI, research, the cost, and the rug pull.
- 7 Software Engineering Isn't Changing. It's Over. Dario says software engineering is going away. Other AI leaders say hiring is going up. Both are right, and the gap between them is the whole story. Here is what is actually happening, why it is not the apocalypse, and the part that actually worries me.
- 8 Stop Defaulting to Claude Code Claude models are some of the best in the world. Claude Code, the harness around them, often is not. Here is where it falls short, where Claude still wins, and which coding agent to use for which job.