Skip to content

Workshop 02 — Three Options

Time: 60 minutes solo, 90 minutes for a pair.

Goal: Take the #1 candidate from Workshop 01 and force three real implementation directions. End with a written rationale for which one wins.

You should have a candidate from Workshop 01. If not, do that one first.

Step 1 — Run “give me options” (15 min)

Section titled “Step 1 — Run “give me options” (15 min)”

Paste Prompt 02 — Give Me Options, with your candidate as the target. Specify 5 options for this workshop (you’ll use the extra 2 to expose blind spots).

For each of the 5 options, fill in:

OptionLayoutComponentsEffortAudienceTradeoff
1.
2.
3.
4.
5.

If Claude’s response didn’t give you enough to fill in any cell, ask a follow-up:

“For option 3, what specific components would we use, and which exist in our stack?”

Cross out the options whose audience isn’t your audience.

You’ll usually be left with 2-3.

Step 4 — Write the recommendation (15 min)

Section titled “Step 4 — Write the recommendation (15 min)”

Write a 1-page recommendation that includes:

  1. Recommended option. Which one, and one sentence on why.
  2. Why not the others. Two sentences each on why the rivals were rejected.
  3. What the recommended option requires. Components new vs. reused. Estimated effort.
  4. What this doesn’t include. Be explicit about scope cuts (“This does not include direction 5’s customizable widgets”).
  5. Verification plan. What checks will determine “done”?

Time-box this to 15 minutes.

If you’re working in a pair, swap recommendations. Each person plays the skeptic for 2 minutes.

The skeptic asks:

  • “Why didn’t you pick option X?”
  • “What if our user base changes — does this still work?”
  • “What components are you inventing that we already have?”
  • “What happens on mobile / at low data / on slow networks?”

If you can defend the recommendation under 2 minutes of skeptic questions, ship it.


  • Differentiation. Are your 5 options meaningfully different, or are 3 of them slight variations? (10 = clearly different; 0 = palette swaps)
  • Audience clarity. Did you cross out options whose audience isn’t yours? Could you name the audience? (10 = explicit; 0 = vague)
  • Reuse. Does the recommended option specify components that exist in your stack? (10 = yes, with paths; 0 = invented from scratch)
  • Scope cuts. Did you write down what this doesn’t include? (10 = yes; 0 = no)
  • Defendability. Could you survive 2 minutes of skepticism without backpedaling? (10 = yes; 0 = collapsed under questioning)

35/50 = you can ship. 45/50 = you can lead this workshop.


  • Five options that are all slight variations. Force differentiation. Make them disagree on the user model, not just the layout.
  • Picking option 1 because it’s the cheapest. Cheapest isn’t always right. The right test is: “what does this product need to be in 12 months?” If option 1 plateaus and option 2 scales, pick 2.
  • Skipping scope cuts. This is where most projects fail. Write down what you’re NOT building, and don’t quietly slip it back in.
  • Not naming the audience. “Power users” isn’t an audience. “Field technicians on phones in trucks with intermittent signal” is.