Process 80 documents with an agent
The poster asked Dot to process 80 documents; it finished about 60 and stopped, and refused to commit or merge a PR. ·
Done withDots
- What they did
- The poster used Dot on coding work tied to their own GitHub repo and also tried having it work in its own VM. They told it to process 80 documents, create and merge PRs, and use the browser, but Dot repeatedly paused for approvals, refused to commit or merge, and stopped after 60 documents when its Sentry safety layer flagged the batch as too risky. The poster says they kept resetting authorizations, memory, custom rules, and even tried gh/git CLI access.
- How it went
- It processed about 60 of the 80 documents before stopping. The reported failure mode was not bad output so much as repeated refusals and interruptions: browser access, file reads, git/gh actions, and PR merging were denied or stalled.
- Worth knowing
- The main catch was constant babysitting: the poster says they had to restate approvals repeatedly because Sentry denials stuck in context and derailed later steps.
Try it yourself with Dots
Process these [number] documents in [folder or location]: for each one, [extract fields, summarize, or classify]. Keep a checklist of every document with its status, and keep going until all are done; if you stop, tell me exactly which are left and why. Put the results in [output file], and show me any changes before committing or merging anything.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Oct 20, 2026
Related use cases
Plan a trip and book flight, hotel and shuttle with an always-on agent
A user had Dots plan a trip and remind them to book the shuttle; the agent filled booking forms but kept asking for confirmation and wouldn't enter 2FA codes, so it only got partway.
Turn customer feedback into tested pull requests
In OpenAI's own launch example, a Dot watches feedback for recurring requests, scopes small fixes, builds and tests them, and returns PRs with demo videos.
Turn GitHub issues into draft pull requests with long-running agents
The author used Dots for read-heavy long-horizon tasks like reading GitHub issues and making PRs, which were typically poor but gave a starting point.
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Write end-of-sprint changelogs from completed tickets
A PM has Claude Cowork read completed Linear tickets, pick what matters, draft the update and suggest email or in-app format, replacing about an hour per sprint.
Collect monthly invoices and email them to an accountant
The poster built a bot that logs into each site monthly, grabs the invoice, and emails it to their accountant.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.