Arjun's hiring panel scores candidates in their heads, and the bar moves with the mood of the room. He asks Claude for a scorecard. It takes 13 turns — because the rules that make a scorecard fair (weights, must-have gates, a threshold that a high score cannot override) all live in his head. Then he writes them down, folds them into one prompt, and rebuilds it in a fresh chat in one message.
This is what she ended up with. Every rule below it — the slabs, the cap, the order of operations — had to be dragged out of her head one turn at a time. Change the inputs and watch them fire.
Click through it. Notice what each turn actually is: not Claude failing, but a rule arriving late. Every correction is a piece of context that was missing from turn 1.
At the end of the messy chat she asked Claude one question: “list every correction I gave you in this chat.” Claude has the whole conversation in front of it, so the log costs almost nothing to produce — and this log is the context she was carrying in her head all along.
ScaleEvery criterion is scored 1–5. Never out of 10.
WeightsCoding 40% · System design 25% · Communication 20% · Culture 15%.
GateCoding is a must-have. Under 3 on coding = NO HIRE, whatever the weighted total says. The gate beats the arithmetic.
Gate displayWhen a gate fails, say NO HIRE loudly and name the gate that failed. The score stops being the headline.
ThresholdWeighted score is out of 5. 3.5 or above is a hire — inclusive.
Tie-breakOn a tie, the higher coding score wins.
BlanksA blank is not a zero. Any unscored criterion = incomplete card, and no verdict at all.
JustificationA score of 1 or 2 requires a written note. Nobody kills a candidate without a reason.
TransparencyShow each criterion's contribution (score × weight), not just the total.
BehaviourRecalculate live, and never wipe the notes when a score changes.
OutputOne screen, no login, prints cleanly for the panel file.
The log, folded into a single message: role, spec, the rules, the order they fire in, the edge cases, the output format. This file is the real deliverable of HO4 — not the tool. The tool is just the proof that the context is complete.
Build a single-file HTML interview scorecard for hiring engineers. No login, no backend, one screen, prints cleanly. CRITERIA — each scored 1 to 5 (never out of 10), each with a weight: Coding 40% System design 25% Communication 20% Culture 15% SCORING RULES — follow these exactly 1. The weighted score is out of 5. Show each criterion's CONTRIBUTION (score x weight), not just the final number. 2. HIRE at 3.5 or above. The threshold is INCLUSIVE — 3.5 exactly is a hire. 3. CODING IS A MUST-HAVE GATE. If coding scores under 3, the verdict is NO HIRE no matter what the weighted total says. The gate BEATS the arithmetic — do not show a healthy total as the headline and bury the gate. Say NO HIRE loudly and name the gate that failed. 4. A blank is NOT a zero. If any criterion is unscored, the card is INCOMPLETE: give no verdict at all, and say what is missing. 5. Any score of 1 or 2 requires a written justification note. No note, no verdict. 6. On a tie, the higher CODING score wins. Say so. BEHAVIOUR - Recalculate live as scores change, and never wipe the notes when a score changes. - Show the verdict as one unmissable line: HIRE / NO HIRE / INCOMPLETE.
She opened a new chat, pasted that one prompt, and got the same tool — working, first try. Same features, none of the archaeology.
This is the whole lesson. Claude was never the bottleneck. The context was — it just arrived one turn at a time instead of all at once. Say it up front and the fourteen turns collapse into one.