Snowflake lays off documentation staff after they train AI replacement | Hacker News Reader

Snowflake lays off documentation staff after they train AI replacement

threadreaderapp.com

25 pointscedporter19 comments

Sounds like a company digging its own grave by forcing its employees to dig their own graves

It sounds like a company that will profit more by saving money and giving their customers a worse product, and they don't have to worry about the consequences of that because every other company is doing it too.

Yes, but what if some companies don't do it and instead make better products. Do we think that's impossible? Because with so many companies focused on making worse products more cheaply, it seems like it should be easier than ever to win the game by making good products...

>Do we think that's impossible?

Yes. Even if it weren't, it doesn't even seem to matter anymore. Making better products doesn't seem to lead to more money than churning out shit and financializing it, so why would anyone bother?

Consumers have voted over and over and over and they are very clear: The vast majority will choose cheap vs good

Commodity products win in the long term and “better” products that are more expensive will go out of business

> Consumers have voted over and over and over and they are very clear: The vast majority will choose cheap vs good

Snowflake customers have definitely not made this choice, as Snowflake is good but very, very expensive. They're basically the Oracle of cloud.

Isn’t Oracle the oracle of cloud?

And this still fits, if snowflake is feeling pressure from lower cost entrants or demand from investors for more profit then it would track

I don’t necessarily disagree that it’s a risk but people assume that companies optimize for product stickiness but in fact they don’t, most companies optimize for investor relationships

Yeah the real holy grail is a product that's both cheaper and better. If we're still climbing the ladder of abstraction, over time it does become possible to build products both cheaper and better.

But who cares about the long term when you could make a killing this year or even just this quarter and walk away with a fortune.

The brutal part: they made the senior writers spend their final 6 weeks "knowledge transferring" to the AI system

I’ve said this for decades: The future is transfer learning from humans to machines until the point where bootstrapping new behaviors doesn’t need a human

Interesting perspective, I honestly had not thought about it this way. I work on problems around knowledge transfer but always from the angle of people leaving or transitioning roles where the goal is to preserve knowledge so it does not get lost. Framing it as people effectively training the system that might replace them feels pretty brutal.

It’s literally how information transfer happens between humans and machines forever - how do you think factory automation works? The deming system was literally timing every human task and then replacing them mechanically one at a time

See one, do one, teach one has been standard in bootstrapping behavior learning in advanced mammals since the early 20th century

Why would it not be applied to non human systems that are capable of replicating it

This is so interesting. My work is having us do a AI hackathon where we have to attempt to do our daily job only using AI tooling. No writing any code yourself, no manual input besides to the agents. I guess I should prepare myself for the inevitable conclusion of this.

Move up the technical-product abstraction stack if you want to survive

A young Kim Catrall had a Vulcan commentary on this: https://www.elastic.co/customers

How is that link related to the post?

I'm sure this will work out well and in no way impact the quality of their product.

>Documentation quality hasn't dropped because the AI learned from the best

Yet

More disenfranchised employees to be soldiers in the Butlerian Jihad.

Reply on news.ycombinator.com