Top 4 Themes inthe Discussion
| Theme | Summary & Key Quote |
|---|---|
| 1. Token cost & inefficiency of vision‑based agents | Many users point out that “computer use” is dramatically more expensive than calling an API. “The vision agent took almost 20 minutes… the API approach… 0.5 s – 2.8 s … a 17‑minute (!!) total time for the vision agent vs. 0.5s‑2.8s for the API approach.” — palashawas |
| 2. Preference for structured APIs / MCPs over raw UI navigation | Several commenters stress that “structured APIs (or MCP) beat vision agents in speed and cost.” “Structured APIs… are 40× cheaper, more deterministic, and pay the server bills.” — jacktu |
| 3. Real‑world constraints & skepticism about AI handling sensitive tasks | Users highlight practical barriers: security, privacy, and the need for reliable APIs. “I don’t see AI ‘agents’ being trusted with taxes, background checks, or creating an LLC – those require human oversight.” — overgard |
| 4. Future OS/API redesign to expose functionality to agents | There’s a recurring call to rethink operating‑system design so that every app’s functionality is “exposed via an API while remaining human‑friendly.” “In an agentic world, the OS needs to be completely rethought… every single app functionality should be exposable via an API.” — aurareturn |
The themes above capture the most common points raised, each backed by a direct quotation from the participants.