Comment by shadowgovt | Hacker News Reader

Comment by shadowgovt | Hacker News Reader

Parent Full thread

I'm a big fan of "the bazaar," but it requires a lot more cognitive load and a lot more legwork to patch together how to do things.

Living in the Google ecosystem means a lot of your services are integrated, and that's actually pretty great. Docs knows about email. Email knows about photos. My mobile device knows about all this stuff.

Those ecosystems are not just the software units, they're the integration. And I've done enough decades of running my own bespoke GNU/Linux-derived distros to know that the integration is an extremely nontrivial lift.

There's a reason people keep checking into these software hotels instead of building their own shacks in the woods, even if they don't own the walls. I don't think users will be particularly happy with those who force them to live in the shacks, and the idea that they now own the peeling paint and malfunctioning floorboards will be cold comfort.

That's all 100% true, but at the same time, wading through and figuring out how to dodge all the issues that come with being in "the cathedral" requires increasingly more and more cognitive load as well. If both are full of landmines, you might as well go with the bazaar.

I used to be hardcore "use Google services for everything," and while I still use a lot of them, I'm gradually replacing them with more disparate and often self-hosted options. I supposed I lose out on a lot of that integration, but personally I've found that I kinda just don't need it. The integration offered by living in an "ecosystem" has always been a kind of vague promise, defined more by what it potentially enables rather than what it actually delivers. At the end of the day, how much integration do I really need between my docs and my emails?

Even business use-cases for this stuff still require a lot of "going to the bazar," whether it's in the form of employees writing bespoke code, paying for another company like Zapier to actually do the heavy lifting for you, etc. This of course is the raison d'etre of LLMs in the workplace -- offering a comfortable interface, natural language, for making use of all this data and finally real-izing the mostly wasted potential for integration we've been living with for the last decade or so. It's "no code" on steroids, but IMO like most "no code" solutions, it too will fall short, promise too little juice for too much squeeze, and "the bazaar" will continue to live on.

Reply on news.ycombinator.com