HN Explorer

Software is rapidly being commoditized

5 points by squatin64 ↗ HN
software is rapidly being commoditized and it seems that within a year people will be able to build out their own applications and likely eventually even complex bespoke things like 'make an AI doctor' or other tools

for a long time software has been the lifeblood of the tech startup ecosystem, but it seems like it would be unsustainable to build a business on if anybody can spin up a team of agents and copy something with some effort in rapid time

i never really thought of developers as the 'middlemen' between computers and people but i guess thats true and therefore they are ripe for automation

as a developer it feels really demoralizing to realize that your skillset is obsolete and that even if i moved to the product or ai research side of things that im not sure the whole space isn't washed out

in the past more software has not brought down the value of software, just accelerated its advancements but i think we're seeing a shift here and i also don't think there are any limits on the kinds of software that the models can't create, in fact, i think they can create unbelievable software that humans would never be able to make by hand in the long tail

let me know what you guys think about this and how you're planning to pivot if you are planning that

i personally want to continue to be pushing innovation forward and found something that brings value to the world, learning whatever i have to along the way whether thats hardware or physics or everything about a domain/business area

4 comments

[ 5.1 ms ] story [ 28.3 ms ] thread
squatin64
ive also realized lately that the value for software creation/agents comes when we take humans out of the loop entirely and don't do code reviews or even that much work on architecture, just presentation of product/technical ideas & tastemaking based on that, thats the real unlock

i hope many swes can transition to something like that but then again i think commoditization makes it such that there isn't as much opportunity economically to do something like that

GianFabien
Similar angst was commonly expressed back in the days when high level languages were replacing assembly language and also when OOP was intruding upon FORTRAN, COBOL spaces.

With each technological advance we are able to work at a higher level of abstraction. As you already suggest, broadening your knowledge to domain specific areas is a good idea.

>people will be able to build their own applications Not so fast. Most people aren't willing to spend hours learning prompt engineering to get the latest tools to produce a correct application program that covers all the edge-cases, security and privacy requirements, meet performance targets, etc.

Application systems builders for specific domains will still be in demand. You will need both domain knowledge/expertise as well as prompt engineering (or whatever it is going to be called) skills to have the tools produce exactly what is required.

chadcmulligan
I worked with these guys that kept sending me AI explanations for SQL queries, like it was somehow helping. The explanations were fairly sound, though didn't seem to help them achieve the results they were after, and any Sql guy could have figured it out with a few minutes thought. They just had this chasm between what the AI told them and what it actually meant, words without understanding are just words. I am becoming concerned this will be the future, people saying words AI tells them and they have no idea what it means, I'm starting to see it already to a small extent, I'm hoping my sample is not representative.
benoau
I think it's amazing working with Cursor. Sometimes it's dumb as shit but I can keep it focused and on track and working so fast I feel like the bottleneck is my 100+ WPM typing speed. I am looking forward to making software like never before. I am using AI for art, for video, for 3d, for writing tests, for proofreading and revising, and so much more. I feel like "full-stack" is a much broader set of technologies this year.

But I am also not considering this as a "career" anymore, at best there will just be continuous attrition from now on and maybe some of the time I can have a job. I don't see a path to retirement as a programmer. I don't know how much of my software will be written for or on behalf of other people moving forward.

But I am also optimistic, the main thing AI does is give me the output of an entire team even if I have to find my own avenues to publish and monetize it.