Hacker News new | ask | show | jobs
by simonw 347 days ago
Because I continue to collect definitions of "agent", here's what Andy Jassy said in that memo: https://www.aboutamazon.com/news/company-news/amazon-ceo-and...

> Think of agents as software systems that use AI to perform tasks on behalf of users or other systems. Agents let you tell them what you want (often in natural language), and do things like scour the web (and various data sources) and summarize results, engage in deep research, write code, find anomalies, highlight interesting insights, translate language and code into other variants, and automate a lot of tasks that consume our time. There will be billions of these agents, across every company and in every imaginable field. There will also be agents that routinely do things for you outside of work, from shopping to travel to daily chores and tasks. Many of these agents have yet to be built, but make no mistake, they’re coming, and coming fast.
2 comments
imiric 347 days ago
> Many of these agents have yet to be built, but make no mistake, they’re coming, and coming fast.

This is the same wishful thinking that AI companies are heavily marketing.

Nobody will want to use an "agent" that makes mistakes 60% of the time. Until the industry figures out a way to fix the problems that have plagued this technology since the beginning―which won't be solved by more compute, better data, or engineering hacks―this agentic future they've been promising is a pipe dream.

link
edanm 347 days ago
I appreciate you keeping up the fight for correct terminology! I think at this point though there is a "standard" definition of agent - an AI that can actually utilize external real-world tools to do things on behalf of users. That fits coding agents, web-using agents, etc.

Do you think there's still confusion around it like there was a year ago?

link
simonw 347 days ago
Definitely. The one you are using there is pretty much the accepted software engineering definition now, but if you talk to non-engineers you'll still hear all sorts of variants about things like travel agents or UI automation or "autonomy" without explaining what that means.

Anthropic use the tools-in-a-loop one quite consistently now, but OpenAI still sometimes say things like "AI agents are AI systems that can do work for you independently. You give them a task and they go off and do it." - https://simonwillison.net/2025/Jan/23/introducing-operator/

link