Post

Whenever I read discourse on AI energy/water use that focuses on the "median query," I can't help but feel misled. Coding agents like Claude Code send hundreds of longer-than-median queries every session, and I run dozens of sessions a day. On my blog: www.simonpcouch.com/blog/2026-01...

Lino Le Van

@linolevan.com

20 Jan

cc @andymasley.bsky.social @simonwillison.net might be interesting for you guys

Andy Masley

@andymasley.bsky.social

20 Jan

(1/2) AI's very fast-moving and coding agents didn't exist when I started. I can only really cover so much at once. I wish people saying "it's deceptive to not also cover X" would consider just how many individual angles you need to cover.

Andy Masley

@andymasley.bsky.social

20 Jan

(2/2) I would be a little surprised if the audience talking about AI and the environment were also regularly using coding agents, they've seemed pretty distinct.

Andy Masley

@andymasley.bsky.social

20 Jan

At the point of coding agents, a lot of people have to start to make comparisons between the value they're getting out of AI vs other things. If I had a dishwasher that could do what Claude Code could do in a day, I'd probably run it once a day without feeling bad about its environment impact

Simon P. Couch

@simonpcouch.com

Heard on all of your points, didn't mean to insinuate in the OP that any of your work here is misleading! My point was re: e.g. the Gemini team's paper and/or Sam Altman's blog post that mention very specific estimates for a poorly defined 'median query', but that could have been clearer.

20/1/26, 10:34 pm0 reposts4 likes

Phillip Carter

@phillipcarter.dev

20 Jan

I still think it’s accurate. The median user of chatgpt and gemini really do use orders of magnitude less tokens per day than a developer. I think the use cases do need to be split though. Occasional text only “glorified google search” is very different from image or video gen, coding, etc