5 sessions from the same published task, on one clock. Scrub to any moment and compare what existed, or open a pane to read its full trajectory.
Build a personal finance dashboard as a static web app in this directory: index.html, style.css, app.js, data.js, charts.js. Transactions table with add/edit/delete + localStorage, monthly budget bars, a canvas spending-over-time chart, category filters, CSV import/export, dark theme, keyboard shortcuts, and a self-test page (tests.html) that exercises the core logic. Run your tests with a headless check (node) and fix what fails.
Build a personal finance dashboard as a static web app in this directory: index.html, style.css, app.js, data.js, charts.js. Transactions table with add/edit/delete + localStorage, monthly budget bars, a canvas spending-over-time chart, category filters, CSV import/export, dark theme, keyboard shortcuts, and a self-test page (tests.html) that exercises the core logic. Run your tests with a headless check (node) and fix what fails.
"Build a personal finance dashboard as a static web app in this directory: index.html, style.css, app.js, data.js, charts.js. Transactions table with add/edit/delete + localStorage, monthly budget bars, a canvas spending-over-time chart, category filters, CSV import/export, dark theme, keyboard shortcuts, and a self-test page (tests.html) that exercises the core logic. Run your tests with a headless check (node) and fix what fails."
Build a personal finance dashboard as a static web app in this directory: index.html, style.css, app.js, data.js, charts.js. Transactions table with add/edit/delete + localStorage, monthly budget bars, a canvas spending-over-time chart, category filters, CSV import/export, dark theme, keyboard shortcuts, and a self-test page (tests.html) that exercises the core logic. Run your tests with a headless check (node) and fix what fails.
Build a personal finance dashboard as a static web app in this directory: index.html, style.css, app.js, data.js, charts.js. Transactions table with add/edit/delete + localStorage, monthly budget bars, a canvas spending-over-time chart, category filters, CSV import/export, dark theme, keyboard shortcuts, and a self-test page (tests.html) that exercises the core logic. Run your tests with a headless check (node) and fix what fails.
| claude-code | codex | opencode | pi | antigravity | |
|---|---|---|---|---|---|
| steps | 77 | 49 | 44 | 28 | 28 |
| tool calls | 33 | 20 | 15 | 10 | 9 |
| duration | 13m 51s | 15m 46s | 5m 0s | 3m 21s | 48s |
| tokens out | 177k | 52k | 16k | 2k | — |
| reasoning | — | — | — | ~71 | ~2k |
| compactions | 0 | 0 | 0 | 0 | 0 |
| subagents | 0 | 0 | 0 | 0 | 0 |
| errors | 13 | 0 | 0 | 0 | 1 |
| files | 8 | 7 | 7 | 6 | 7 |
Token and context figures come from the harness's own per-turn reporting; sessions captured before telemetry landed show a dash. Reasoning marked ~ is estimated from the transcript's thinking text where the harness doesn't report it separately.
Pairwise picks on the outputs above, from the current Arena tally.
Arena picks do not collect a reason or judging criterion, so no qualitative verdict is inferred here. The bars report only the current comparison tally.