Docs, slides and sheets: 3 tasks #116

Test how well a UseAgent agent makes the files people use at work (documents, slides and spreadsheets) on three tasks, and report what happened.

Tasks

1. A slide deck. Ask for a 6 to 10 slide deck on a topic you know, then ask for two rounds of changes.
2. A document. A one to two page brief, plan or how-to, with headings, and a table where it fits.
3. A spreadsheet that works. A budget, tracker or calculator with real formulas (totals, averages, a lookup), not just typed-in numbers.

Free to try

Sign up free at app.useagent.org. New accounts get free model credits (GLM or Claude Opus, up to a limit) and free cloud sandbox time, enough to finish this task without paying or adding a key.

1. Sign up with your email at the link above. The link in the email works for 24 hours; check spam if it's missing. No API key, paid plan or card needed.
2. Type the task in plain words and press send. The agent gets its own computer with a browser, terminal and desktop, and you watch it work.
3. When it's done, copy the link from the address bar (app.useagent.org/session/...). That link goes in your report.

Stuck? Comment here with a screenshot and we'll help.

How to do the runs

• A task is 10 runs. Your report has one line per run.
• Start each run yourself in the app, one at a time, and watch it work. Never more than 2 runs going at once. No scripts, tools or other AIs sending prompts for you.
• Every run needs at least one follow-up that reacts to that run's actual result: ask for a specific change and check it landed, or point out a specific problem. One run without this rejects the whole report.
• Record the whole session in one continuous, real-time screen recording that shows you starting each run and sending each follow-up. No cuts, no speed-ups.

Your report

Post one comment on this issue, one line per run:

Task: <number>

1. <thread link> - follow-up: <what you asked> - did it work: yes / no - <one sentence on what happened>
2. ...
...
10. ...

Recording: <link to the one continuous recording of your whole session>

• Each line describes its own thread. We read every thread before paying. Lines that don't match their threads, or repeat each other, reject the whole report.
• Reports that look batch-run or scripted aren't paid.
• Runs that didn't work still count, as long as you followed up on them.
• Before paying, we may ask about your own threads (for example, a price or a heading on the page the agent built). Answer from your threads.
• For the level of detail we expect, see the accepted report on #112.

Claim it

Comment "Taking task N". You hold the task for 3 days, then it opens up again.

Payment

• $5 per task, to the first person whose report checks out.
• One report per person per task. The first report that passes the checks gets that task.
• To get paid, email hi@useagent.org from the same email you signed up to the app with, and include a link to your report.

Keep private details out of here

This page is public. Don't post your real email, phone number, address, payment details, API keys, or screenshots that show any of them. Send anything private, including how you'd like to be paid, to hi@useagent.org with a link to your comment here.

Rules

Any of these means no payment, and we hide the comment without replying:

• Made-up runs or links, runs copied from someone else, or the same prompt ten times with tiny changes.
• More than one account per person, or a report for a task someone else is holding.
• Runs done or simulated by another AI tool. Do the runs yourself.
• Spam or harm: no mass form submissions, no scraping behind logins you don't own, nothing illegal.