The question of what AI (or AI Agents) can actually do has been long discussed on the internet. Honestly, the answer is really yet to come, however this week, it’s been THE hot topic on r/AI_Agents.

    1. Top answer is: custom agents, this particular redditor built them off of smolagents, the Hugging Face framework

    2. The only other useful answer is someone talking about using LLMs to expand their data and create synthetic data

    1. This is a dumb title, and the post got reported, but it doesn’t look like it’s breaking the rules so I approved it.

    2. Someone is using an AI Agent “manager” that decides what needs to be done and then “hires” other specialized agents to go do those things.

    3. Someone lists that in more complex agents they need to be able to:

      1. plan multi step workflows

      2. handle failures

      3. track cost and latency

      4. adapt when goals shift

    1. Probably the most practical question on here

    2. From someone doing a PhD, they have 1500+ PDFs to search through

    3. Top response? You don’t need an AI, you need an indexing/retrieval system, basically a good RAG system

    1. really weird that this is the top voted comment yet also the one “no one wants to use”

    1. this is great! this is a real world application

    1. this is 10 projects by the Azure team that are basic agents

    2. also I posted this 🙂 (I’m completely unaffiliated with Microsoft)

Community Message Board

Special Announcements

88% of the people who voted have said YES they want a hackathon so we’ll be doing an AI Agents hackathon soon!

Reminder that Seattle Startup Summit is happening at the end of this month on 3/28. It’s basically a “who’s who” of AI dev tools.

Reply

Avatar

or to participate