oneshotlm
Compare all models on this prompt →

Windows 95 desktop · gpt-5.6-sol

Recreate a Windows 95 style desktop in a single page with draggable and closable windows, a Start button and taskbar, desktop icons, and a working clock.

⛨ sandboxed Open in new tab ↗
transcript · 6 messages
user

Recreate a Windows 95 style desktop in a single page with draggable and closable windows, a Start button and taskbar, desktop icons, and a working clock.

assistant
bash {"command":"ls -la && printf '\\nPI vars:\\n' && env | grep '^PI_' | sort"}
toolResult

total 0 drwxr-sr-x. 2 node node 6 Aug 3 00:22 . drwxrwsrwx. 3 root node 16 Aug 3 00:22 .. PI vars: PI_CODING_AGENT=true PI_MODEL=openai/gpt-5.6-sol PI_PROVIDER=openrouter PI_REASONING_LEVEL=medium PI_SESSION_FILE=/home/node/.pi/agent/sessions/--tmp-ws--/2026-08-03T00-22-47-212Z_019fc500-6cac-77af-ab86-b2ab2787099a.jsonl PI_SESSION_ID=019fc500-6cac-77af-ab86-b2ab2787099a PI_TIMEOUT_MS=600000

assistant
thinking
**Building a polished UI**

I need to create a polished, self-contained design with icons, possibly using CSS for inline SVG or pixel icons, since emojis might not be the best fit. I should include features like a Windows-style start menu, clock on the taskbar, and draggable, closable windows. Opening multiple windows through desktop double-clicks and managing task buttons for restore and minimize actions are also key. It would be ideal to have a welcome window that opens responsively. Time to write the full file!
write /tmp/ws/index.html (21361 bytes)
toolResult

Successfully wrote 21361 bytes to /tmp/ws/index.html

assistant

TAKE IT TO CODE

A single self-contained HTML file — copy it and open in any browser, or reuse the prompt with your own pi-agent run.

RUN

modelgpt-5.6-sol
vendorOpenAI
harnesspi-agent
generated2026-08-03

EVALUATION

2/5 broken

Windows 95 desktop with Welcome dialog stays static; clicks, drags, wheel, and keys never close or move it.

Vision grade · anthropic/claude-sonnet-5

MEASUREMENTS

Raw readings from the capture, not a verdict — the grade above is the only judgement. Percentages are how much of the screen changed; the measure under-reports sparse motion on a dark background.

idle motion0.0%
on click0.0%
on drag0.0%
on wheel0.0%
on enter + space0.0%
on arrow keys0.0%
on w/a/s/d0.0%
frame spread57.5 / 255
console errors0
js errors none

METRICS

runtime77.1s
tokens in2.5k
tokens out6.7k
cached2.3k
cost / run$0.2039