Hacker News new | past | comments | ask | show | jobs | submit login

I don't really get the comments from HN people. I could understand my neighbour (non-tech) saying "Ha ha, you software engineers will become jobless because of AI", but the HN crowd saying that? Here's my rationale:

- AI is a catalyzer. It will let us do more with less

- Most of the companies out there are greedy. They don't look for stability, but for growth

- If right now, it takes a company X resources to build product Y in N months, then with AI the company will have to decide between a) build the same product Y with X/10 resources and in N/10 months, or b) build 10 different products with X resources and in N months. Greedy companies will not choose option 'a' because that options leads to stability; they will choose option 'b' which is the one that leads to (hyper) growth

Just imagine your usual startup in 2024 invigorated with AI:

- Founder 1: I want to build product X. Cool, with this AI thingy I can hire only 1 engineer, 1 product manager and maybe 1 designer to build what it would have cost 20 engineers, 2 managers and 3 designers in 2020.

- Founder 2: Cool. So we just hire 3 people and launch X in 1 month, right?

- Founder 1: Well, we could do that. Or, we could hire 20 engineers, 2 managers and 3 designers to build not just product X, but also product Y and product Z. Right? I mean, why should we limit ourselves with one product only? What if it fails? At least if we launch 3 products, we'll have more chances of "winning"

- Founder 2: brilliant!

So, AI will just boost customer demands and greedy companies (the majority of companies). There will be plenty of jobs for all of us. If you don't believe this, not long ago, Twitter had around 7000 employees. Yeez, could you believe that? And Twitter had technology of the 21st century!

My bet is that AI will enable tech companies to look more like Google: small-medium companies will be able to do A LOT more (e.g., we'll see small companies launching complete G-suite-like software in less than 6 months, we'll see YouTube-like competitors like we see nowadays TODO lists)... but that won't be achieved with 10 or 20 engineers only.




I have two comments, both subject to change. One is that I think most startups hire too many people anyway, you don't need 20 people to build a TODO app. Depends on the scope sure, but I see it all the time. Too many chefs, too much code.

Second though, no one has demonstrated an AI that can write programs with complex requirements yet. Even light requirements it doesn't approach holistically or with the goal of a coherent software ecosystem. It might give a junior the tools to muddle through code snippets from ChatGPT and copilot but you would still need just as many of them. And your senior engineers derive even less value from what language models provide, since they are not struggling to write programs, coding is never the hard part of building an application. There is limited value there currently. Maybe help with an algorithm or two but it's not about to start making changes to your eCommerce platform for you.

I think we need AI specific frameworks for that to occur, ground up AI codebases, like a DSL for the somewhat disjointed and abstract approach that language models have toward code. Current frameworks are aimed at providing guidance and strict scaffoling to humans to reason about business logic in but the grey matter is actually busy solving the very specific business problems, and those problems are spread over a fourth dimension, ever evolving with time.

The other issue is that language models rely on prior art, for human language that is pretty abundant. But I suspect the problem solving skills of resolving business logic into code are not well represented online, since no one is writing blogs about their every thought at BigCo Commerce and Analytics.


Nobody's invented an orchestrator yet. There will be a kAIbernetes invented by someone, which will analyze structures. It seems to me that not many people are really trying hard to project and extrapolate in a realistic way. This will develop much like other techs have developed, but express itself (literally, expression is what these fuckers do) very differently in its various stages.

Ten years got us from seeing dog faces in everything to verbal diarrhea that makes sense most of the time and award-winning Instragram photographers that only exist as a StableDiffusion delusion. You and I might be gone in twenty years, but the world will still be here in a thousand. We just don't know, but I do know one thing - AI is still at the "Steam Thresher" stage of things.

We programmers will definitely be the last to go, I think, but we'll go too, because fundamentally, we became programmers to make machines do stuff we don't want to do.


kAIbernetes will become what jQuery became: someone will need to create something better at some point. It's all the same again. AI won't affect shit. Did we lose our jobs because of compilers?


Follow your example: someone did create something better at some point. Hence, someone will create something better at some point.

Automating people away without taking care of them has been given a pass because so far, the people being automated away were not People Who Matter.

Let's watch.


Yeah. If it works for more than just generating code snippets with no actual understanding, then we get to level up.


I've been thinking about it in terms of the transition from pre-1950s human computers to electronic computers. How much should we morn the loss of rooms of people doing calculations by hand? Making it easier for people to solve problems seems like an obvious win even if puts the "just give me the requirements in a ticket and let me code" programmers out of their jobs. Even outside of big software projects, imagine the business benefits from allowing an office worker to describe a problem and get an excel formula to try.




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: