Aaron Walter (@awlevin) / X

Aaron Walter

779 posts

Aaron Walter

@awlevin

building @vellum_ai, prev: CTO @thinkfiveable (1k to 1M MAU)

free stuff 👉

Joined July 2010

Aaron Walter
@awlevin
Aug 6
thrilled to share all we've learned while building our own agent memory system! h/t to @pragunseth for this awesome piece
Vellum
@vellum_ai
Aug 6
Read this Vellum memory system breakdown to learn how we got your Assistant to remember you!
Article
How Vellum Remembers
Models alone have the memory of a goldfish. We all have the experience of reminding ChatGPT or Claude the info it missed when using it. We built a memory system to fix that. For AI assistants to work...
Aaron Walter
@awlevin
Aug 5
a list of 10 areas where AI still has room to improve to meet my bar for AGI. yes there are specialized models that solve these. but, to me, AGI is when we have all of them in a unified model that is economically viable for the masses. here's my list: 0. text - AI can't write
Aaron Walter
@awlevin
Jul 29
been using vellum voice mode a lot lately. everything you see in the video is realtime. imo we nailed it in a few areas: > proper-waiting on pauses while speaking > invisible handoffs during complex tasks > ambient usage UX > interruptions & latency all while avoiding
anita · vellum.ai 👾🦾
@anitakirkovska
Jul 29
Work will look like this sooner than you think. You can now speak to your @vellum_ai agent to handle complex tasks like plan your trip, push a PR, or optimize your Meta ads. Watch Alex get things done with just his voice:
00:00
Aaron Walter
@awlevin
Jul 28
do other assistant harnesses let you set permission modes per each channel your assistant is in? or is it just vellum?
Aaron Walter
@awlevin
Jul 25
It’s so frustrating to hear the widespread sentiment that AI is “ruining critical thinking.” But I get it. A lot of what we used to spend cycles thinking about is now solved. But the critical thinking still exists… *it’s just displaced* We now spend more cycles thinking about