Ask HN: What are your AI assistants doing?
At them moment it has some tasks, mostly periodic/cron:
- Scan TenderNet (public procurement website of my government) API for new procurements. Match to my profile. Send categorized matches over Telegram (daily).
- Monitor an inbox (got one at AgentMail [0], a Ycombinator company, no affiliation) for emails with job descriptions. Match to my profile. Send categorized matches over Telegram (hourly).
- Create and monitor another inbox for invoices, make received emails into a PDFs, forward to my bookkeeping software (which can process invoices, but not if they are in the body of an email) (hourly).
- I also gave it a folder that I expose to the web via Caddy, so I can quickly whip up demos. I did one somewhat useful thing with that so far.
- Occasionally I ask it to research something or some company.
So the pattern is: I make small automations, 1 of them does not even need an LLM, only the profile matching does. I do not integrate it into systems with important information. It essentially only processes public data.
It's fun, certainly, the LLM assumes the ST Voyager computer persona, answers with Vulcan salutes, I call it nr. 1 from time tot time. The model underneath is easily switched (Claude Pro is limiting, now experimenting with a DeepSeek model, hope to one day do it all local on excess solar.)
So my question is, what do you use agents like this for? Is it strictly for fun or also for profit? I automate the tedious and try to find little things to do, but I could do these things by hand if I really wanted. Let's share some stories!
[0] https://www.agentmail.to/