Post

I think calling design deliverables 'artifacts' is a misnomer. A 10 page document that nobody will ever read is not some artifact. The only reason it would be called that is because in 200 years someone will put it in a museum to ask why the previous generations were wasting so much time.

Those 10x more artifacts that none will read will get summarized by AI. Then this summary will be sent in an AI generated email with convoluted fluff because the more words you write the more important you are, to other people. Who will use AI to summarize that email. Rince. Repeat.

I feel like AI gave people the power to pretend they do more, while also wanting to do less, and we are just losing information along the way in a never ending game of "make it bigger" then "summarize it" along the way. For someone with a brain that seeks efficiency, this is honestly a nightmare.

When folks say they have 10x'd their productivity with AI, most of the time this is what is actually happening:

This made me think of Rust’s recent policy on LLM use, where the point out that long, detailed pull requests “no longer indicate effort” blog.rust-lang.org/inside-rust/... #rust #llm

When people say they are 10x more productive I just ask them if they have finished all the work they expected to do before 2030.

I ask them if their employees get to end their week at lunch time on Monday or if they get paid 10x. gotta be one of the two, right?

I feel very old school but I always start my stakeholder presentation with an executive summary that contains the conclusion and main action point. Then they can dig in for the details. Because I know most will not care about the details (it's fine they are mostly to explain how we arrived there)

Or maybe it's the academic background where you have to put the thesis and conclusion in the intro of your paper 😅

I made the flippant video above, but now here is data and telemetry on the exact issue of AI creating more content, while the content takes way more time to fix: bsky.app/profile/rtm2...

It's no surprise that we are seeing the result of this issue now: A lot of tokens being used with lots of cost, but not much to show for it. The problem, of course, isn't directly AI. It's management. AI has amplified the issues of bad management, and lays them bare much more quickly.

To think big documents means lots of work has happened, or to think many new features or many lines of code means lots of work has happened, is a problem of management.