rvachev.orgEN / RU / 🤖
← Back to essays
· Essay · 1 min

Issues with Pet Projects and AI Email Processor

I've realized that the main problem with my Pet projects is that they remain in the shadows.

I've realized that the main problem with my Pet projects is that even if they do “take off,” they remain in the shadows because I don’t talk about them at all, and as a result, they don’t go beyond MVP – this needs to change.

One of the projects I’m working on in my free time is an AI email processor. I’ve tried different solutions (from https://fyxer.com/ to https://superhuman.com/) and found that they are all very superficial. The attempt to simply overlay a prompt leads either to a complete lack of flexibility or to a drop in quality.

I decided to combine a deterministic approach (if/else) with the use of LLM.

Minimum task (which is already more or less solved): sorting emails into folders (notifications, emails related to kids, everything about AI, etc.)

Maximum task:

- to understand where actions are expected from me, but it’s not spam “reply if you’re interested”
- to automatically respond to standard requests (or at least create drafts)
- to launch some external processes
- to summarize all emails in the folder once a week (convenient for subscriptions)

Would you like such a thing in your email? What else should it be able to do besides what has already been described?

Issues with Pet Projects and AI Email Processor — illustration

Newsletter

Get new posts by email

A weekly digest of what I write about AI, software and products. Unsubscribe anytime.