Applause is weak evidence that a workshop worked.
The room may have enjoyed the examples. The speaker may have answered every question. By Monday, participants can still face the same problem with the same habits.
I have taught technical interview sessions, competitive programming workshops, career talks, university classes, and internal engineering sessions. The audiences and subjects changed. The useful design question remained the same: what should someone in the room do differently when I am gone?
Write the Monday decision
I begin with a decision the participant will encounter soon.
For a technical interview workshop, the decision may be how to respond when the first solution is too slow. For an agent-reliability session, it may be whether a timed-out action can be retried. For a career discussion, it may be how to compare two roles when neither dominates every dimension.
“Understand system design” is too broad. “Choose a safe response when an external action has an unknown result” can shape an exercise.
The decision also sets the boundary of the workshop. Interesting material that does not improve it can move to another session.
Expose the current model early
Participants need to make the decision before hearing my answer.
I present a short case with enough information to commit. A booking tool times out. A candidate produces a correct quadratic solution. A student has two internship offers with different kinds of ownership. I ask what they would do next and why.
The response gives the workshop a baseline. It also gives the room a reason to care about the distinction that follows. A principle introduced before anyone needs it sounds like advice. The same principle introduced after two plausible answers conflict becomes a tool.
I keep the first case small. Extra context gives participants ways to avoid the central choice.
Teach the separating condition
Most workshops offer several useful ideas. Participants remember a smaller number, often without the condition that tells them when to use each one.
I focus on the fact that separates two plausible actions.
After a timeout, the separating condition is whether the outside effect can be established. In an interview, it may be whether the constraints rule out the current complexity. In career selection, it may be whether the role gives direct ownership of the work the person wants to learn.
Once the condition is explicit, the participant can derive the action instead of memorizing my preference.
The explanation should be short enough that the room can use it in the next case. If I need twenty minutes to defend the framework, it is not yet ready for a workshop.
Change one variable
The second exercise should look close to the first while requiring a different answer.
Suppose the scheduling provider now supports lookup by operation identifier. The timed-out booking can be reconciled. Suppose it does not support lookup and the current record contains two possible matches. The safe response changes.
Changing one variable makes the decision boundary visible. A collection of unrelated examples may feel comprehensive, but participants cannot tell which detail controlled the answer.
I ask someone to state the changed fact before giving the new action. That sentence is often the best evidence that the workshop has taught a reusable method.
Make the artifact during the session
Slides are a record of what the speaker presented. Participants need a record of what they should do.
I prefer a one-page artifact completed in the room: a retry decision table, a problem-review sheet, an interview checklist, or a role-comparison record. Blank fields matter. They force the participant to translate the idea into their own case.
A useful artifact contains:
- the situation that triggers the decision;
- the evidence to inspect;
- the available actions;
- the condition that selects one;
- the point where ownership must transfer.
I test the artifact by removing the talk. If a participant opens it two weeks later, can they recover the method?
End with a real next use
The last exercise should come from work the participant already has.
An engineer identifies one unresolved action in a current workflow. A student selects one problem to review using the new sheet. A job seeker rewrites the criteria for one active decision. The participant leaves with a started application, not an intention to apply the lesson someday.
Questions at the end are useful when they test the boundary. I pay attention to the cases that the artifact cannot handle. Those cases become the next revision.
Review the workshop as a system
Afterward, I do not ask only whether the session was well received. I ask:
- Did participants make the target decision during the workshop?
- Could they identify the fact that controlled it?
- Did the second case change the answer for the right reason?
- Did each participant complete an artifact tied to real work?
- Which question exposed a missing boundary?
- What evidence could be checked after Monday?
The speaker cannot be the workshop’s runtime. The session succeeds when the participant can face the next case, inspect the right fact, and act without waiting for the slide that used to contain the answer.