All posts
Everything published here, newest first.
316 posts · showing 1–10
2026
10 posts-
The Analytics Agent Answered "What Was Q3 Revenue?" From a Deprecated Table. Make It Name Its Source Before Its Number.
The SQL was valid, the arithmetic was right, and the number was still wrong, because the agent picked the table whose name looked most like the question. Here is how to make a data agent commit to a source you can check before it hands anyone a figure.
-
The Invoice Tool Timed Out, So the Billing Agent Called It Again. The Customer Was Charged Twice. A Timeout Is Not a Failure.
A timeout tells the agent it stopped waiting, not that nothing happened. Here is how a reasonable-looking retry turned into a double charge, and the idempotency key that makes the retry safe.
-
Your Inbox Agent Wakes Every 15 Minutes to Find Nothing. Cron or Event Trigger? Decide by What a Wasted Run Costs.
The trigger is the least-discussed design choice in a personal agent, and it quietly decides your bill, your duplicate rate and how stale its answers are. A simple test for choosing between a schedule and an event, and why the best setup usually uses both.
-
Your Research Agent Read One Poisoned Web Page, and It Still Had the Email Tool. Lock Risky Tools the Moment Untrusted Text Enters the Context.
Prompt-injection filters try to spot the bad sentence. A taint flag skips that guess: once fetched text is in the context, the tools that can do damage are no longer offered.
-
Your Research Agent Scored 100% on Citations. The Grader Only Checked That a Source Was Attached.
An eval that has never failed anything is not proof of a good agent. Before you trust a grader's score, feed it answers you know are wrong and see if it notices.
-
One Big update_task Tool or Three Small Ones? Your To-Do Agent Only Sees the Names and Descriptions
A to-do agent archived a task you meant to finish, and the tool was working fine. The model picks tools from a short menu of names and descriptions, so how you split that menu decides how often it picks wrong.
-
The Tax Agent's Workers Counted March Twice and Skipped Late June. Make Every Delegation Return the Range It Actually Covered.
Splitting a year of transactions across four workers looks like clean delegation until the merged total is off and every worker reports success. The fix is a coverage contract the supervisor checks before it adds anything up.
-
The Travel Agent Reviewed Its Itinerary Three Times and Approved the Wrong Date Every Time. Reflection Needs Something to Check Against.
A self-review loop that reads only its own draft can polish the wrong answer forever. The fix is to give the critic something the generator never had: the original request, a fresh context, or a tool that can say no.
-
Your Support Agent Still Remembers the Customer Who Was Furious in March. Decide What Memory Expires Before You Decide What It Stores.
Cross-session memory fails less often from forgetting than from remembering too long. A four-class expiry scheme for support agents, and the one field every stored memory needs.
-
The Expense Agent Logged "Rate limit exceeded" as a Vendor Name. Make Tool Errors Impossible to Mistake for Results.
When a tool returns its failure as a plain string, the model reads it like any other output and carries on. The fix lives in the tool's return shape, not in the prompt.