Day-of Operations
AI Coding Cage Match · 4 Teams · 4 Tools · 85-minute run
Setup
Build
Demo
Score
Pre-Event Runway
T − 7 d
Subscriptions & authAll tool seats paid; accounts authenticated per machine. Credit balances noted.
T − 7 d
Repo readyFork from post-cutoff codebase; lock
tests/; install dependencies; baseline tests green.
T − 48 h
Tool assignments sentEmail participants their assigned tool. No self-selection on the day — clusters on the "best" tool.
T − 48 h
Scoring rubric publishedCorrectness 40% · Completeness 25% · Quality 15% · Edge cases 10% · Speed 10%.
T − 24 h
Credit balance re-checkCopilot switched to 1,500 credits/mo (June 2026) — agentic loops drain fast.
T − 24 h
VLANs & hotspotsPer-team SSID confirmed. Hotspots charged and sealed in envelopes. Judges calibrated on sample output.
Day-of AM
No Claude Code rehearsals5-hour usage reset window can exhaust session allocation before the event clock starts.
Day-of AM
Notifications clearedOS alerts, Slack, email — silenced on all presenting machines. Font ≥18 pt in IDE + terminal.
ClockPhaseAction
T − 60
Setup
Accounts verified, repo cloned, all baseline tests passing on every machine
T − 30
Setup
Teams seated by cluster; Slido join code on screen; per-team WiFi SSID confirmed from each device
T − 10
Setup
Facilitator reads task spec aloud; questions answered; countdown timer visible to audience
0 : 00
Build
BUILD STARTS — all 4 teams simultaneously
Drivers at keyboard · Navigators on requirements · Observers logging prompts & dead-ends
+ 15 : 00
Build
Facilitator check-in: "any tool crashed?" — surface blockers early, don't wait for stoppage
+ 30 : 00
Build
BUILD STOPS —
git commit or screenshot to lock stateNo further edits · Observers submit prompt logs to judge panel
+ 32 : 00
Demo
Team 1: 3-min walkthrough → demo operator runs test suite on projector → Slido quick vote
+ 45 : 00
Demo
Teams 2–4 in sequence (~13 min each · walk + test run + audience vote). Display Slido join code before each poll.
+ 57 : 00
Score
Judge scores finalised; Slido ranked-choice overall-winner poll opens
+ 65 : 00
Debrief
Results announced + structured debrief: privacy posture, cost-per-token, latency, spec-alignment failures
Room & Display Setup
🖥 PRESENTER DISPLAY — Scoring dashboard + countdown timer
Team A
Claude Code
💻💻💻
Team B
Cursor
💻💻💻
Team C
Copilot
💻💻💻
Team D
Devin Desktop
💻💻💻
📽 AUDIENCE DISPLAY — Mirrors active team IDE during demos
Each cluster: 4-outlet surge strip · HDMI + USB-C adaptors + 2 spare sets · dedicated VLAN
Font ≥18 pt IDE/terminal; bump to 22 pt for rooms deeper than 8 m
Black-on-white outperforms white-on-dark below 3,500 lm projector output
Font ≥18 pt IDE/terminal; bump to 22 pt for rooms deeper than 8 m
Black-on-white outperforms white-on-dark below 3,500 lm projector output
Network Infrastructure
5 Mbps
per device
1.5×
devices per head
≤30
users per radio
210 Mbps
minimum (20 pax)
30 devices × 5 Mbps × 1.4 headroom = 210 Mbps
- VLAN per team — isolates each agentic loop from the others
- Wi-Fi 6 / 6E preferred for dense concurrent connections
- Bring managed switch if venue can’t provision VLANs
- One charged hotspot per team in a sealed envelope — brief upfront
Per-Tool Pre-Event Setup
$17–20 / mo (Pro)
npm i -g @anthropic-ai/claude-codeAuth once per machine. VS Code / JetBrains extension or terminal CLI.
Heavy agentic sessions hit the 5-hour usage reset — no rehearsals morning-of
$20 / mo (Pro)
Cursor Pro required for Composer 2 multi-file editing. Supports Claude, GPT-5.5, Gemini — decide model before the event.
Free-tier trials expire unpredictably — use paid seats only
$10 / mo Pro · $19 / seat Business
Extension for VS Code, JetBrains, Neovim, Xcode. Confirm participant IDE before the day.
June 2026: usage-based credits (1,500/mo Pro) — verify balance 24 h before; agentic loops drain fast
$40 / user / mo (Teams)
Rebranded from Windsurf on June 2, 2026. Existing Windsurf accounts carry over — download new installer from cognition.ai.
Docs still say “Windsurf” in many places — brief participants on the rebrand
$20 / mo (Pro, 1,000 credits)
AWS-backed. CLI on Windows 11+; macOS/Linux via desktop. Stage a
SPECS.md in the repo the week before to cut cold-start friction.
Spec-driven workflow needs SPECS.md pre-staged — create it the week before, not the morning of
Pre-Flight Checklist — 12 items
All tool subscriptions paid; credit balances confirmed 24 h before (2)
Repo cloned and dependencies installed on every participant machine
All baseline tests passing before clock starts (locked state)
VLAN or dedicated SSID per team configured and tested (15)
HDMI + USB-C adaptors for each team; 2 spare sets for presenter display (4)
Slido/Mentimeter event created; join code tested from non-presenter device
Printed requirements doc (1 per team) as offline fallback (1)
Mobile hotspot per team, charged and ready in sealed envelopes
Power strips (4-outlet, surge-protected) at each team cluster (1)
All OS/app notifications cleared on presenting machines (4)
Judges calibrated on sample output; scoring rubric published to all participants