Writing / Tag / 9 of 76
Agents
Every essay tagged Agents, newest first.
This is a smaller sibling of the AI Agents tag, and the essays under it are mostly the operational ones: how to trace a run, how to authorize a tool call, how to keep a run alive through a crash, how to stop an agent fanning out faster than the systems downstream can take. If you run agents against real APIs with real credentials, this is the set to read.
The through-line is that an agent inherits every property of the services it calls. It needs backpressure like any client, authorization like any user, and tracing like any distributed system. The model does not change any of that; it just makes the gaps more expensive to find.
Read the OpenTelemetry tracing essay first, because every other fix is easier once you can see where the tokens went. Then the confused deputy problem, on why being able to call a tool is not the same as being allowed to, and durable execution for a run that survives step 40.
August 2026
One Customer Burned Your Month of LLM Budget by 2pm
LLM spend is per request and wildly variable, so a single runaway agent or one heavy tenant can externalize its cost straight onto your margin.Your Tool Returned 40,000 Tokens. The Agent Needed 12.
A single tool call can dump a whole file, a full API response, or a thousand log lines straight into the context window.The Agent Is Not Confused. Its Context Is Stale.
In a long session an agent keeps every tool result it ever saw, including the three older versions of a file that has changed twice since.Your Agent Calls One Tool, Waits, Then Calls the Next
When an agent needs three lookups that do not depend on each other, running them one at a time makes the user wait for the sum of all three.
July 2026
The Agent Can Call the Tool. That Is Not the Same as Allowed.
An agent that holds a tool runs it with the whole application's credentials, not the current user's.Your Agent Cost $2 Yesterday and $40 Today and You Cannot See Why
An agent's cost is elastic and path-dependent, so the same task runs for two dollars one day and forty the next.You Pay Full Price for the Same 8,000-Token Prompt on Every Agent Step
An agent resends the same enormous system prompt, tool list, and history on every step, and without prefix caching you pay full input price and full prefill latency each time.Your Agent Fans Out Faster Than Anything Downstream Can Take
An agent that spawns subtasks and fans out tool calls builds an unbounded work queue, hammers your rate limits, and burns the token budget before the run finishes.Your Agent Dies at Step 40 and Starts Over From Zero
A long agent run that crashes halfway restarts from scratch, re-spending every token it already paid for and re-firing side effects you never wanted twice.