BrowseSprint

BrowseSprint is an independent research project about slow Codex browser workflows. This listing covers what is available today: a free troubleshooting guide and a downloadable, blank CSV timing worksheet. It does not describe an available browser extension or an automated optimization service. The project also maintains a research waitlist for conversations and possible future tests.

The resource is for people who use Codex with browser tools and want to investigate a slow task before choosing a fix. A long elapsed time can include navigation, page reading, decision gaps, retries and timeouts. The guide encourages recording what can actually be observed, keeping unknown stage timings blank, and checking whether the requested outcome was completed correctly.

Start by defining a small task on a public page. Specify what counts as success, such as finding one fact and its supporting source. Keep the starting page, session conditions and expected result comparable between attempts. If a run fails, record the failure rather than removing it because it took longer than a successful run. Avoid putting private page contents, credentials or confidential work into a shared example.

The CSV worksheet provides columns for approach, run, public page, outcome, session state, total seconds, success, tool calls, retries and stage timing notes. It includes three baseline rows and three changed rows, ready for the user to fill in. These are blank recording slots, not measured results or evidence that an optimization works. Three runs are a practical starting comparison; they are not a statistical guarantee.

Change one factor at a time and compare both elapsed time and reliability. A shorter run that returns the wrong answer is not an improvement. Repeated page reads may suggest checking how much relevant information a tool returns. Slow clicks may suggest checking whether the page is ready and the target is actionable. Navigation failures deserve their own notes. Each observation should lead to a focused next test rather than a claim about every possible workflow.

The guide and worksheet are accessible on the public website without joining the waitlist. There is no payment or installation needed to read the guide or download the CSV. Users can keep their filled worksheet locally. Optional research signup is separate and is intended for project research and future tests.

BrowseSprint is at the research stage. There is no released extension, confirmed launch date or verified speed improvement. It is not affiliated with or endorsed by OpenAI. The useful outcome available now is a clearer measurement of a browser task and a better documented question about where the wait occurs.

What problem it solves

A slow browser task is hard to diagnose when elapsed time, correctness, repeated reads and failed retries are mixed together. This free guide and blank CSV help people record comparable public-page tasks, mark unknowns and decide what to measure next. It does not automatically speed up browsing.