← Back to context

Comment by fragmede

13 hours ago

You're assuming all women in your cohort start not pregnant. However, given a random sampling of women across the entire human race, if you have approximately 14,000 women, statistics says you'll have a baby in a month. That is to say, the chances of one of those woman being 8 months pregnant reaches close enough to 1, given about 14,000 randomly selected women.

Also, you can get a baby tonight if you steal one from the maternity ward.

The real question is, how do LLMs turn the mythical man month on its head. If we accept AI generated code, can an agentic AI swarm make software faster simply by parallelizing in a way that 9 women can't make a baby in 1 month because they're am AI, not human, and communicate in a different way.

The pitfall of AI coding is that previously every shiny tangent that was a distraction, is now a rabbit hole to be leaped into for an afternoon, if you feel like it. It's like that ancient Chinese curse, may you live in interesting times. Everybody can recreate an MVP of Twitter in a weekend now when previously that was just a claim a certain type of people made.

6 comments

fragmede

Reply
majormajor  12 hours ago

> You're assuming all women in your cohort start not pregnant. However, given a random sampling of women across the entire human race, if you have approximately 14,000 women, statistics says you'll have a baby in a month. That is to say, the chances of one of those woman being 8 months pregnant reaches close enough to 1, given about 14,000 randomly selected women.

There's a good point in here along the lines of "if you need X in a month, and someone else has something that's 90% of what you want X to be, can you buy it from them before starting any crazy internal death marches instead?"

> The real question is, how do LLMs turn the mythical man month on its head. If we accept AI generated code, can an agentic AI swarm make software faster simply by parallelizing in a way that 9 women can't make a baby in 1 month because they're am AI, not human, and communicate in a different way.

This is quite possibly only a one-time switch from a changed baseline, though. Give it a few years and "the fastest way an LLM tool can do it" will be what gets tossed out a an estimate, and stakeholders will still want you to do it in a tenth the time...