Start Over,
Context is Finite.
Unlearning the sunk-cost fallacy of the manual draft to leverage the infinite iterative potential of Large Language Models.
Drafting: rethink what a draft is
Consider the transition from childhood to professional legal practice. As a student, a "draft" was a physical artifact—a stack of paper that represented hours of labor. Throwing it away felt like a tragedy, a waste of tangible resource.
In the professional world, the scale changes, but the psychology often remains. We treat the digital page as if it were parchment. However, in the realm of AI-assisted counsel, the draft is no longer a monument; it is a temporary state.
refresh The Case for Starting Over
We are often blinded by the "Edit Trajectory." We assume that fixing a bad draft is the shortest path to a finished product. With AI, this is frequently false. Opening a new chat window and refining the prompt from scratch often yields a superior result faster than trying to "prompt-correct" a hallucination or a stylistic failure.
"Relish the chance to be the 'Evil Boss.' If your AI subordinate hands you a mediocre draft, do not coddle the output. Demand a redo. The cost of rejection is effectively zero."
Fix the Prompt vs. Re-run
Distinguishing between a Mechanical Failure and a Logic Failure is the mark of a senior AI operator.
The 'Demurrer' Failure
In a recent test involving a demurrer, contradictory instructions led to an output that argued against itself. The mistake wasn't the AI's "intelligence"; it was the context window being polluted with conflicting signals.
-
Mechanical Failure Wrong format, missed a clause, or stylistic drift.
Action: Re-run with a tweak. -
Logic Failure Fundamental misunderstanding of the legal theory or fact pattern.
Action: BURN THE WINDOW. Start over.
Course Milestone
You have reached the end of the 'Context Management' module. Test your ability to identify logic failures before moving to the next section.
TAKE THE QUIZ arrow_forward
