Where does AI go after chat?
I chatted with Charlie Guo of Artificial Ignorance about the ongoing shift from AI as a chatbot to something that can actually do work.
Here’s some stuff we cover:
Why the biggest leap may be the “harness” around models, not just the models themselves.
How Codex, GPT Image 2, and agentic workflows are changing software and knowledge work.
Why agents still use human-facing interfaces today, and what agent-native tools might look like.
How much do you still need to review your agent’s work, and when might you be able to trust it to do things independently?
Ambient AI assistants, multimodal interaction, open-source models, and where practical use cases are heading next.
A lot of this stuff is speculative, but it was fun to pick Charlie’s brain and share some of my own thoughts on where we’re heading.
We discussed doing a more hands-on session with Codex for practical tasks, so if you have any specific things you’d like to see covered, reach out to me at whytryai@substack.com or send me a message directly here on Substack:
