Software jobs are being automated away, but not by AI. If you think about it, any time-saving library, service, or tool is going to reduce headcount at an efficient organization.

For example, I currently have a company that I technically manage. It has one employee who does maintenance on the infrastructure (AWS Elastic Beanstalk), database (AWS RDS), and code for five different products. The only other people who touch the code are security auditors.

That just wasn't possible 10 years ago. Managing servers alone for those products (~50 EC2 instances, ~10 load balancers, 10 database instances) would've been a full-time job.

Yep, this came up at a previous company (with 700 people at the time) that had existed before the cloud did. As AWS became more and more prevalent in 2010/11 they had to decide whether they wanted to be in the business of running physical servers and racks in data centres or whether they wanted to move to cloud hosting providers.

It was a non-trivial exercise, and to the best of my knowledge they still have some physical servers, but slowly but surely almost everything moved to Amazon.

Those physical server operations employees slowly left for other jobs.

They company kept growing and employing tech people though. Just tech people doing other things.

Apparently AI gives companies an excuse to avoid dealing with false positives.

If your human support team made a mistake, there was an appeal process. If your AI makes a mistake, sorry, bad robot, deal with it.

Some companies would be happy to pay that AI to develop their b$ software, because once in a position to abuse you customers, why not abuse them with less costs.

AI won’t be replacing software engineers anytime soon but something to keep in mind:

While people talk up Googles ai (at least they once did) the truth is while they may put out some good research now and again most of the stuff they deploy is crap by design - they can’t afford to spend a dollar per day per customer to run a complex algorithm on your data, they can only spend fractions of a penny.

This severely limits the quality of the ai deployed by large consumer oriented megacorps.

On the other hand you look at some of the ai being deployed in the b2b space or anywhere there is a real exchange of money and you’ll see some pretty expensive and interesting ai research and deployments.