Most discussions of controlling a model’s output collapse into a single move. For one team it is prefilling the first token; for another it is a longer, more emphatic instruction; for a third it is a schema bolted onto the response. Each of those is a real technique, and each is being asked to carry work it was never shaped to…
The most-watched numbers in an agent system are usually the prompt and the model’s own output. The instructions get versioned, the responses get graded, the system prompt gets audited line by line. Meanwhile the largest and least-examined source of tokens in a long-running agent is often the raw output of the tools it calls. A single API response can be…
In a single-agent system, the risky moments are the tool calls: the points where the model reaches out and something in the world might not behave. In a multi-agent system, a second class of risk appears, and it is easy to miss because nothing about it looks like an error. It lives at the boundary between one agent and the next. Every…
The moment a plan begins to run, it meets a world it did not fully model. A tool returns something the plan never anticipated, a resource that existed at design time has moved, a value comes back valid but wrong for what the next step assumed. This is not a defect in the planning. It is the ordinary condition of acting…
Most agents end up holding more tools than they use. Each one seemed reasonable to add at the time, and none of them seemed expensive to keep. The place this reasoning breaks is in what a tool actually is to the model. A tool is not an action the agent reaches for only when the moment calls for it. It…
Most failure handling is built around the case where the system eventually gets what it wanted. Retry the call and it works the second time. Reach for the alternate tool and it returns. That branch is satisfying to design because every path on it ends in success, and it absorbs most of the attention a team gives to reliability. The…
A language model can hand you an object that is perfectly shaped and entirely false. Every field is present, every type is right, the whole thing parses without complaint, and the value in the field your downstream logic depends on was invented. This is the failure mode that survives structured output. Enforcing a schema settles what shape the answer takes. It settles…
Two of the cheapest and most reliable ways to steer a model are to tell it who it is and to tell it what it knows. The first is identity: a role assignment, optionally extended into a full persona. The second is grounding: placing the specific facts a response should reason over directly into the prompt. Both are levers you pull…
Once you accept that a summarizer will drop the exact values a long session depends on, the fix is easy to describe and easy to half-build. Pull the facts out of each turn before compression runs, write them into a durable region the summarizer never touches, and the values survive while the narrative is free to shrink. Stated that way, a…
Inside a single agent, coordination is almost free. Everything the model has read, decided, or produced sits in one context, and any later step can reach back and use it without anyone arranging for that to happen. The moment work is split across separate agents, that convenience is gone and most people do not notice it has left. Each agent runs in its…