#tutorial
Every post tagged "tutorial".
92 posts · showing 1–10
-
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.
-
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.
-
Your Coding Agent Hit Its Cost Cap Mid-Refactor and Left Nothing Behind. Reserve the Last Fifth of the Budget for the Handoff.
A hard spend cap protects your bill, but when it trips mid-task it throws away everything the agent learned. Split the budget into phases so the run always ends with a usable handoff instead of a dead process.
-
The Analytics Agent's SQL Was Correct. It Answered a Different Question. Make It Print Its Definitions Before Its Number.
The query ran clean, the chart looked right, and the churn number went into the board deck. It counted trial accounts as customers. The fix is not better SQL; it is making the agent commit to what every word in the question means before it touches the warehouse.
-
Your Research Agent Passed the Eval Once. Run Every Case Five Times and Grade the Worst Run.
A single green run tells you the agent can do the task, not that it will. Repeat each eval case, split the suite into solid, flaky and broken, and the flaky bucket becomes your real to-do list.
-
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 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 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.
-
A 5% Canary Mostly Samples Your Easy Tasks. Roll Agent Upgrades Out by Task Class.
Random traffic slices sample volume, but an agent's risk lives in the thin tail of long, tool-heavy, irreversible tasks. Here's how to build a rollout ladder that samples the danger instead.
-
When the Context Window Fills, Cut in This Order — and Leave a Stub Behind
Most agents evict context by dropping the oldest tokens, which reliably deletes the nine words that mattered while keeping a 4,000-token HTML dump nobody needs. Here is the order to cut in, and why every eviction should leave a stub behind.