HEAT STRATEGY DAY 2026

The AI Jury has deliberated

AI Challenge · 4 groups · 4 skills · 60 minutes · one prize.
Scored against the published rubric — no favourites, no mercy.

AI SessionStrategy Day 2026

20 August 2026Königstein

JuryClaude, briefed & blindfolded to charisma

Hand-drawn robot judge with four nervous group submissions

How the jury scored

Five dimensions, anchored 1–10 — the plan was scored, not the pitch

25%

Usefulness for heaties

Real, frequent, painful job at HEAT?

25%

Functional soundness

Complete steps, inputs, rules, example — would it work?

20%

Reusability

Can any heatie call it by name and get the same result?

15%

Impact: time saved

Credible quantification — frequency × minutes × people?

15%

Clarity & feasibility

Half-page spec followed, buildable quickly?

Jury rules: only the written hand-in counts · every score cites the hand-in · a simple skill that clearly works beats a grand skill that can't be specified · tie-breaker: Usefulness, then Reusability.

The contenders

Four skills entered the courtroom. All four shipped.

Spidy the timesheet detective · the Pre-Bid Machine · the De-Hype 3000 · the Skill Compass. In order of appearance, Your Honour.

Group 1 · Steffi, Abbas, Hoshang, Rohit, Cesare

/expert-timesheet-check — "Spidy" counts every expert-day

Spidy the timesheet detective spider checking hanging timesheets

Usefulness25%

9.5

Functional soundness25%

9.5

Reusability20%

9.0

Impact: time saved15%

9.0

Clarity & feasibility15%

9.0

Weighted total9.3

"Eight complete steps, a working prototype with demo data, 6–8 PM-hours/month quantified — and they checked first that nothing duplicates /weekly. Textbook."

Group 2

/tender-prebid-agreement — from ToR to signature-ready

The Pre-Bid Machine turning ToR, partners and budget into a signed contract

Usefulness25%

8.5

Functional soundness25%

9.0

Reusability20%

8.5

Impact: time saved15%

7.0

Clarity & feasibility15%

8.0

Weighted total8.3

"Live and tested with three demo outputs, never invents a VAT number, always keeps a human before signature. The half-pager itself was thin — the package saved it."

Group 3

/language-dehype — wash the hype, keep the facts

The De-Hype 3000 washing machine with detect, rewrite and verify gnomes

Usefulness25%

8.0

Functional soundness25%

8.5

Reusability20%

8.0

Impact: time saved15%

6.0

Clarity & feasibility15%

9.0

Weighted total8.0

"The most precise design in the room: a fixed checklist, a three-agent chain, honest limitations. But not one number on time saved — the spec asked."

Group 4

/skill-compass — lost in 165 skills? Follow the needle

A heatie explorer with a glowing compass in a jungle of skill signposts

Usefulness25%

8.5

Functional soundness25%

8.0

Reusability20%

9.5

Impact: time saved15%

6.5

Clarity & feasibility15%

8.0

Weighted total8.2

"The most universal idea of the day — every heatie benefits, and the hash-cached card design is clever. No prototype yet, and the half-page grew into two."

The scoreboard

All four cleared 8.0 — but one ran a full point ahead

Skill	Useful 25%	Sound 25%	Reuse 20%	Impact 15%	Clarity 15%	Total /10
G1 · /expert-timesheet-check "Spidy"	9.5	9.5	9.0	9.0	9.0	9.3
G2 · /tender-prebid-agreement	8.5	9.0	8.5	7.0	8.0	8.3
G4 · /skill-compass	8.5	8.0	9.5	6.5	8.0	8.2
G3 · /language-dehype	8.0	8.5	8.0	6.0	9.0	8.0

Scores 1–10 per dimension against the published anchors; weighted totals to one decimal. Full one-line justifications per cell in the written jury report (jury-report.md).

The verdict

Group 1 wins — Spidy takes the prize

A real monthly pain, eight airtight steps, a working prototype built inside the session, and the only hand-in that quantified everything.

9.3weighted total

6–8 hPM-hours saved / month

8arms. zero mercy.

Winners podium: Spidy first, contract second, compass third, washing machine fourth

Credit where due

Every group did one thing better than everyone else

Group 1 · Spidy
Best engineering disciplineChecked for duplicates before building, shipped a prototype with mock data, and bonus-planned a second skill (/project-pulse) extending /weekly instead of cloning it.
Group 2 · Pre-Bid
Best safety rails on a legal documentNever invents a legal name, VAT number or percentage — missing data becomes a question, and a human always signs off. Already live in the KB.
Group 3 · De-Hype
Best-crafted hand-inA precise checklist, a three-agent chain with a skip threshold, disclosed limitations — and the only submission that actually fit on half a page.
Group 4 · Skill Compass
Most universal ideaThe one skill that makes all 165 other skills usable by the non-technical majority — with a smart hash-cached card design that answers instantly.

Final salute

Four skills planned in 60 minutes — two already live in the KB.
That's the whole point of skills.

You write the steps once. Every heatie benefits forever.
Congratulations, Group 1 — and thank you, everyone who shipped.

The AI Jury has deliberated

Five dimensions, anchored 1–10 — the plan was scored, not the pitch

Usefulness for heaties

Functional soundness

Reusability

Impact: time saved

Clarity & feasibility

Four skills entered the courtroom. All four shipped.

/expert-timesheet-check — "Spidy" counts every expert-day

/tender-prebid-agreement — from ToR to signature-ready

/language-dehype — wash the hype, keep the facts

/skill-compass — lost in 165 skills? Follow the needle

All four cleared 8.0 — but one ran a full point ahead

Group 1 wins — Spidy takes the prize

Every group did one thing better than everyone else

Best engineering discipline

Best safety rails on a legal document

Best-crafted hand-in

Most universal idea

Four skills planned in 60 minutes — two already live in the KB.That's the whole point of skills.

Four skills planned in 60 minutes — two already live in the KB.
That's the whole point of skills.