Hacker News Reader

Comment by mark_l_watson | Hacker News Reader

Nice blog, keep writing.

I manage a machine learning team and I also think that at least partially automated data curation and modeling will reduce the number of people required in my field. It might take 5 or 10 years, but I think it will happen.

I think you are spot on that IT and devops will take a hit. I look more at Heroku’s model that AWS and GCP as the future. That said AWS and GCP will keep getting more ‘Heroku like’.

View on HN

Heroku and its parent company, Salesforce. Salesforce databases are code and no code manipulated and should already carry most of the business and customer data and offer several different off the shelf products for plugging other data sources in. For most companies, it could cover the majority of what their legacy IT department does, and for many companies in the Bay Area it handles everything, except their product and what’s in JIRA. For others, it is woven into the product.

It’s a complete blind spot for most engineering minded people because they never realized how flexible the platform was, and with Bret Taylor running the show now, it’s miles away from just being a clunky Sales CRM.

Couple of examples of recent developments:

https://developer.salesforce.com/blogs/2018/12/introducing-l...

https://lightningdesignsystem.com

https://developer.salesforce.com/platform/dx

My experience with Salesforce (mostly with Marketing Cloud) has been that the developer experience is truly awful. Mediocre documentation, no debugging tools to speak of and systems that never quite do everything you want. Maybe other parts of the ecosystem are better but I'd be very reluctant to take another job where I had to work with it.

Yeah, Marketing Cloud is a collection of acquired products that is not really on the actual Salesforce platform. It’s very much its own thing, kinda like Heroku is its own thing.

Actual Salesforce is just expensive, complicated, cumbersome, opaque, legacy-encumbered, and programmed in a unique language and UI framework. Marketing Cloud is all that, plus what you said.

I was a SQL/Tableau analyst and a SF Admin before I turn to software engineering. SF is truly a fantastic piece of shit to do any proper development for. I am so glad i've stepped away from it. It has been built with the idea to sell to managers that will not be handling the implementations, and with the idea that non-programming staff can configure it. The front end is decent and most people can make it do... something. But custom integrations are just a pain. Salesforce impose lots of fair usage policies on upserts for example.

I was at a Money Transfer company and we couldn't move our Transaction Monitoring staff over to SF without hitting limits

I had some cursory experience with SAP, which is another platform that can do anything. From that I got the impression that while these things are extremely flexible, you're often better off just hiring normal software engineers to write normal software for your problem instead of throwing millions at $PLATFORM expert consultants that know the arcane platform.

"I manage a machine learning team and I also think that at least partially automated data curation and modeling will reduce the number of people required in my field."

I mean, if it doesn't, what the hell are you even doing?

The whole point of technology and modern capitalism is to increase automation, increase the amount produced by the same number of workers, and increase the overall amount of wealth in the world and improve overall living conditions for everyone (setting aside very important questions of distribution). I just find it odd people in the computer technology industry find this shocking or especially worrying.

"I look more at Heroku’s model that AWS and GCP as the future."

Google's App Engine was much closer to the Heroku approach, and the AWS approach won. So I will be pretty surprised if the Heroku approach wins out.

Reply on news.ycombinator.com