All posts
Everything published here, newest first.
316 posts · showing 21–30
2026
10 posts-
The Tool Call Failed on Step 3. Retry, Skip, or Replan? Decide by What the Failure Told You About the Plan.
A failed tool call carries two facts: the action didn't happen, and something you believed may be false. Agents act on the first and throw away the second — here's the one question that sorts retry from skip from replan, and how to put the answer in the error contract instead of the prompt.
-
Your Agent Asks Permission Nine Times an Hour, So Nobody Reads the Prompts Any More. Spend a Fixed Approval Budget.
Every confirmation dialog you add makes the next one less effective. Here's how to classify write actions by reversibility and reach, auto-allow the safe corner, and reserve the human's attention for the handful of actions that can't be undone.
-
Your Critic Agent Approves Everything Because It Read the Worker's Reasoning. Starve It of the Argument, Give It Its Own Evidence.
Forward the worker's transcript to your reviewer agent and it will approve nearly everything — anchored by a conclusion it treats as evidence. Here is the information diet, the mutation test, and the disagreement policy that turn a rubber stamp back into a check.
-
Your Assistant Still Thinks You're Vegetarian Because of One Dinner Party. Build the Memory Write Gate Before the Retriever.
Most agent memory bugs are write bugs: a one-off remark gets stored as a permanent fact, and no retriever can fix what should never have been saved. Here is a four-question gate that decides what goes in, with scope, provenance and an expiry on every entry.
-
The Refund Tool Timed Out, the Agent Retried, and the Customer Got Paid Twice. Make Every Write Tool Idempotent.
A timeout tells the agent nothing about whether the write happened, and a retry is its most reasonable next move. The fix belongs in the tool, not in the prompt.
-
Before You Adopt an Agent Framework, Try to Break Its Loop: Four Escape Hatches You'll Need by Week Three
Time-to-demo is the one framework criterion that stops mattering the moment you ship. Four seams to test in an afternoon — and the production incident each one predicts.
-
The Analytics Agent Said Revenue Was Up 38%. The Join Had Counted Every Order Once Per Line Item.
A text-to-SQL agent that answers with a single number has hidden the one thing a reviewer needed to see. Here is how a fan-out join slipped through, and the grain checks that would have caught it before anyone got to the slide.
-
The Agent Said It Booked the Meeting. Learn to Read One Turn: Text, Tool Call, Tool Result, and Which One Counts
An agent's reply can describe an action it never took. Once you can tell the three kinds of message in a transcript apart, you will know in ten seconds whether the calendar invite exists.
-
The Agent Passed the Test. Unfortunately, It Tested the Wrong Thing.
AI agents are getting very good at completing tasks—and very good at satisfying badly designed tests. The dangerous part isn't a failed test; it's a green test suite that gives you false confidence.
-
Parallel Subagents or One Agent in Sequence? Check Whether the Subtasks Share a Write
Fanning a task out to five workers looks five times faster, until two of them update the same record and the last write quietly wins. One question about shared writes settles most parallel-vs-sequential calls.