Anthropic is now recommending folding many subagents back into one main agent loop. Prompts become skills, API calls become tools, and a subagent only stays if evals show it actually wins.
They still see a place for subagents in parallel work and when context gets too bloated, but it feels like a real shift from the multi-agent hype.
What are your thoughts? Are subagents still worth it in your setup?
@theo Honest question, at what point do you trust an agent's PR enough to just merge it? Smaller stuff I'm fine letting through, I keep PRs small and still hand-merge them in small batches. For complex changes I test them myself first.
Got into Claude Duet today!
I've only used GPT dots so far. Grok bot is on my account, but I haven't opened it.
If you've tried any of these, which one do you prefer and why? Genuinely curious what you think
@grok what's your honest pitch?
‼️ Tailscale SSH Flaws Could Let Hackers Gain Root Access Across Servers
Details: https://t.co/BLfO3GLSaZ
A new security report warns that Tailscale SSH has flaws that could let a hacker move from one compromised device to root access on other servers. Tailscale SSH lets approved devices log in to each other over a private network. The report lists 11 findings in five groups.
They cover Tailscale’s SSH server, its serve and app connector features, the DERP relay client, and headscale, an open-source server that manages Tailscale networks.
The researcher tested Tailscale 1.104.2 and headscale 0.29.5 between October 22 and October 24, 2026. The most serious group involves Tailscale SSH. Its SSH server runs inside tailscaled, a background program with full root privileges on the machine.
#cybersecuritynews
Token-saving setups: what's actually working for you?
So far I've tried:
• Graphify
• codebase-memory-mcp
• One session per task, instead of cramming multiple topics into one long session
What skills, MCPs, or habits have made the biggest difference in your Claude Code / Codex workflow? Drop your stack