Use AI, but don’t become a “meat proxy” for its output

addyosmani · x · 2026-08-04

Don’t be a “meat proxy” for AI output

The post argues that using AI is fine, but blindly pasting its answers into Slack, PR comments, or group chats is not useful. The author says the extra step is to read, understand, and validate the model’s output before replying in your own words.

Key points:

AI responses are often verbose, dense, and can contain plausible nonsense.
Repeating them verbatim adds friction for the recipient, who could query the model directly.
In code review, this matters even more: if you copy a ticket into Claude Code and then relay its output without understanding it, the human reviewer is effectively doing the work with the model while you act as a proxy.
The value you add is the judgment and interpretation, not the raw model text.

Original post →

Use AI, but don’t become a “meat proxy” for its output

More from coding & agent