Post

Tim Fernholz
@timfernholz.com

Here's a test of Bluesky's crowdsourcing power: Later this week I'll be interviewing the manager of OpenAI's Dots product on stage at TechCrunch Disrupt. What questions should I ask?

6 reposts14 likes

Ken '|<3ns41' 👻 - Pro-Ukraine 🌻
@k3n5s.bsky.social

What steps are being taken to protect against intentional or unintentional malicious use? Given the challenges in sandboxing models, what steps are being taken to create isolation and ensuring privacy is maintained with the dots?

Drew Breunig
@dbreunig.bsky.social

I've used several general agents like Dots, for over 6 months. Hermes, etc. The issue all of them run into is they store state and notes in Skills, memory, and other natural language files. Assumptions by the model corrupt behavior and these texts rot, requiring maintenance.

Drew Breunig
@dbreunig.bsky.social

Especially when they're used across different contexts (work, clients, personal life). How do they think Dots (or Bot or Muse) will perform for generalist users after 6 months of heavy usage? A year? What are the challenges they expect to hit and how are they thinking about them?

werewolf ver mitzvah 🗿
@ver.ooo

i don't know how to phrase the question in a way that will ellicit an earnest answer but i am curious what they perceive as the hurdles to broad adoption of persistent agents: is this bottlenecked on their product velocity? standards/integration? customer appetite?

Tim Fernholz
@timfernholz.com

Certainly has a lot of similarities with others in the field. I'm interested in this question too and will find a way to ask it, but suspect the answer will be "our model is the difference"