Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
@MatiasMatthews That's the right worry. If the checker reads the coder's reasoning first it tends to just nod along. Task and diff first, notes after, is the order I'd want
i have like 4 agent PRs open that i keep not reviewing. turns out thats everyone
harvard looked at 718 companies using coding agents. 30% more code, 23% more PRs, finished work basically flat. review time up 49%, almost 2x the PRs sent back
who on your team is eating all this
@github@ArtificialAnlys half the cost for 2 points is a decent trade. what I'd actually want benchmarked is whether it fixed the code or just made the test go green
@addyosmani Agree on picking the right idea. The other signal I'd hire for now: can you tell an agent exactly what you want, and notice when it confidently does something else?
Fewer subagents, yes. Fewer independent checks, no.
In my fleet the agent that writes the code isn't the one that QA's it. That split isn't a handoff problem. It's the point.
Anthropic is now recommending that companies REMOVE unnecessary subagents from their AI systems.
This is a pretty significant shift in agent architecture, with their team finding that a single agent equipped with skills can frequently outperform complex multi-agent systems, and one customer seeing roughly 2x faster execution.
The emerging rules of thumb:
→ Skills: Teach the agent HOW to do something, including procedures, expertise, and instructions.
→ Tools / MCP: Give the agent the ability to DO things through APIs, actions, and live data.
→ Subagents: Use for parallel work, isolated context, or tasks a single agent struggles with.
The interesting part is progressive disclosure. Instead of spawning specialized agents or stuffing everything into a giant prompt, one agent discovers and loads skills and tools as needed.
Their advice is to fold existing subagents into the main loop, turn their instructions into skills, expose their actions as tools, and run evals to see if performance holds up.
As models have gotten smarter, improving agent architecture is becoming less about what you can add and more about what you can delete.
https://t.co/DNq60Skkmc