👏🏿👏🏿👏🏿👏🏿👏🏿👏🏿👏🏿👏🏿👏🏿

You ain’t check your message? Lol

My good sis @KanikaTolver just confirmed what I’ve been telling some people over the last couple of years.

🤣🤣🤣🤣🤣🤣🤣🤣

Yes. I’d want a held-out task set with a single agent, same tools, as the baseline. Compare verified completion, then net out latency, cost and handoff failures, against a threshold set before the run. No clear lift, no extra agent. Isolation or security needs are the exception.

Prove a single-agent workflow before you build a multi-agent one. Every extra agent is another handoff to secure and another failure mode to audit. If you need several: delegation never grants permissions, and protocol compliance is not trust.

@randomrecruiter why? Lol

Why do non-US recruiters always think it’s a good idea to call you before they email you a job description? Do they think people are just sitting around doing nothing?

Don’t rub it in lol. Now I might have to fire up the grill

I need that lol

LLM judge: "summary is grounded." 5-line script: "the claim ID it cites doesn't exist." Believe the script. Deterministic checks first. Judge for the rest, validated against human labels. Evals are release gates, not QA.

I’m not surprised

An agent's identity is not the identity of the person who asked. Agent runs on a broad service account, user has limited access, agent fetches what the user can't see. Confused deputy. Propagate the user's context. Authorize server-side at the resource, every time.

That’s wild

Oh remind me to tell you what just happened to me this week lol

Don't give an agent one autonomy level for every task. L1 read-only L2 reversible prep, human commits L3 controlled execution, approval + checks L4 high-impact, qualified human decides Bind approvals to exact parameters. Change one, re-approve.