cate

"That Slack discussion that aligned the team on an architectural pattern? If it isn't discoverable to the agent, it's illegible in the same way it would be unknown to a new hire joining three months later."

Some of these multi-agent workflow writeups are basically reinventing engineering management from first principles - except with agents instead of humans.

https://openai.com/index/harness-engineering/

Harness engineering: leveraging Codex in an agent-first world

By Ryan Lopopolo, Member of the Technical Staff

Mar 14 at 12:18pmWeb
Show threadGreg WilsonMar 14
@cate Silicon Valley has reinvented the bus at least three times, so why not basic management practices?
Show threadmhoyeMar 14
@gvwilson @cate For people who never take public transit or any history courses or management training, all of this must seem like they're doing the most groundbreaking work in history.
Show threadcateMar 14
@gvwilson urgh
Show threadDanielMar 14
@cate one of the saddest things for me to see in this time of AI being shoved into everything is how much a lot of people are willing to write down and explain to an agent that they never did or would've for a human
Show threadDonald BallMar 14
@mrtazz @cate If this is the development that gets management to value and reward documentation…
Show threadChristinaMar 14
@cate I’ve just nodding along as so many “AI makes this totally different” conversations eventually get to “we should do what we always knew we should do”. It’s a little depressing how much we’re willing to help the bots but didn’t have time for all that work when it was only helping humans.