I ran, an experiment in Straive. I asked my colleagues to mail me any questions or tasks that they had. I’d use my agents + my knowledge as context and reply based on that.

A few interesting things came out of it.

First: It often does a better job than me - it has more patience and speed. For example:

  • One colleague sent a detailed pitch. My agent spotted a tiny arithmetic averaging error.
  • Another asked if I knew an ontology SME. The agent found a perfect match in my contacts - someone I would not have thought of (and reached out to.)
  • Another asked why a slide (screenshot) was empty. The agent searched Google Drive, found the slide, analyzed it, and replied: “It’s white-on-white text. But don’t bother. The slide two pages earlier is better for your meeting!”

One colleague asked: “Instead of sending this to your email id, why can’t we ask this directly into Claude or so?” It replied:

The difference is not the model. It is the context and tools. This agent can search my email, calendar, meeting transcripts, Drive, notes, demos and code, then answer using my recent decisions. A plain Claude, ChatGPT or LLMFoundry session can do the same only if it is given or connected to those sources.

… and then listed 3 answers that relevant to him that only I could give!

Where it does worse is, perhaps 2–5% of the time, I read the answer and think, “WTF? You’ve missed the whole point.” Maybe it missed the politics / psychology / nuance. But since I disagree with myself and can’t even remember what I wrote some of this might be noise.

Second: I went to Inbox Zero! Once I got there, it’s actually a slightly lonely feeling (“Oh, no one wants to talk to me any more!”) so I’ve started reaching out to people, asking them to mail me so that I can have AI reply to them.

So, here’s an offer:

Until 31 Aug 2026, you may email [email protected]. My AI agent (with my knowledge, code and tools) will reply within 24 hours.

This is best for questions where:

  • AI won’t know some things (e.g. based on my research / experiments / classes)
  • I wouldn’t have replied to you (e.g. I’m busy, or the question takes too much effort)

Examples:

  • What’s the most relevant thing you’ve already written, built, taught, or seen for this problem?
  • Given everything you know about this situation, what would you do?
  • You said X a year ago. Do you still believe it? What has changed?
  • What have you learned about X from your recent experiments, classes, or client conversations?
  • I’m pitching X to Y. Based on what you’ve seen work, what would you change?
  • Who do you know who might be unusually good for X?
  • Here’s my deck / proposal / code / model. What am I missing? What would you do differently?
  • This is going to take me three hours to figure out. Can your agent figure it out instead?

The best requests are probably ones where asking plain ChatGPT would give you a reasonable generic answer, but something I know, have done, or have access to could substantially change that answer.

Long, detailed emails are fine. Please attach/forward the context (document, email, links, screenshots, …). Short, 1-line questions are fine, too.

I batch runs daily, so please expect a response within about 24 hours.