[dakiol]

Same can be said about Claude, Codex, etc. These tools are amazing (technically speaking) but they don't play in our favor (most of us are regular, replaceable employees). Only the usual suspects benefit from AI (executive layer, investors, etc)

Still amazes me how engineers on HN are in awe of AI and LLMs knowing that 90% of us will be affected (we won't be able to bring money to the table) once the higher ups start to normalize even more the usage of AI to reduce headcount. Not everything is about the technical details people, grow up

[fragmede]

It's an iterated prisoner's dilemma with all the other developers in the world, and some are vocally choosing to defect. The only rational strategy then is to also defect.

[dakiol]

Right. It seems then that all these "elite" engineers on HN aren't as smart as we thought (and yeah, I include myself in that bag).

It's deeply sad to see how our most beloved work (those side projects we pour ourselves into purely for the joy of it) will, at the end, be the very reason most of us lose our jobs (not all of us, but the majority). Openai/antrhopic/etc and others simply took all of that and turned it to their advantage. It's capitalism, sure, but it's heartbreaking... I wouldnt mind be out of job for another reason, but not for that one pls

[apsurd]

All is not lost though is it? We can invest our efforts into local models and frontier competitors.

I'm not blind, I have Claude pro (not max) and Cursor subscription. But I'm really hesitant to go balls to the wall on the most powerful models because it isn't sustainable; I don't want it to be. So how much can I get from the older models, the smaller, cheaper ones that will hopefully inevitably be commoditized. I think the harness improvements are making headway. I continue to think Cursor Composer 2 is more than adequate.

Then again if one believes it's a race to the singularity, then that's another story. I don't.

[fragmede]

Why not?

[apsurd]

The most concise answer as of now is because AI has no "will".

LLMs are objectively smarter than any one person so in some definition we've already created super-intelligence. The problem is they just sit there. They have all the answers already, if you think about it. Whenever we ask it something it gives us the answer, it's amazing, we can even say it can synthesize new information. We can agree with all the claims.

But what does it do with that super-intelligence? Nothing. It can't. it doesn't have will. Or interest. Curiosity? Biological imperative. Who knows.

So we create loops and introspection and set them free. Does giving AI a goal make the AI conscious? That's easily silly if you ask me.

(I'm trying really hard not to make this philosophy. I really like the philosophy aspect, but this is my 30 second answer to the question)

[fragmede]

The singularity won't happen because sticking a cron job in front of an LLM and telling it to do something (make money) is "silly"?

I am no philosopher but https://poc.bcachefs.org/ seems conscious.

[securicat]

As if Claude and Instagram are remotely similar products. But again, these products make it incredibly easy to cancel. If work requires that you use it, make the next job you get not require it or just use it on the job.

[AvAn12]

Both try to maximize engagement. Both (soon to be) ad supported. Both driven by algorithms that show the user what they want to see.

[dakiol]

I see engineers addicted to Claude the same way non-tech people (friends of mine) are addicted to instagram. At the end it's all the same: making multibillion dollar companies richer every day

[joe_mamba]

>These tools are amazing (technically speaking) but they don't play in our favor (most of us are regular, replaceable employees).

I'm a mid programmer at best, like compared to top guys in the industry, who built stuff like OpenClaw or those prodigy 16 year-old coders who became millionaires, and yet I don't fear the LLM assisted coding future. I'm at peace knowing that I will adapt to the LLM programming world using my knowledge in my favor, or adapt to a world where I will no longer be a SW engineer, but something else.

Also I find it ironic and poetic how some SW devs here want us to rise up and fight LLMs and the companies making them for disrupting this profession, when the SW dev profession was so well paid precisely because the SW products they wrote, disrupted other peoples' professions, moving the savings from labor costs into the pocket of employers, who used SW to optimize processes and repetitive labor and not have to hire as many people, yet they never saw an issue with other people losing their jobs. "Learn to code" eh?

Oh how the turntables.

[LPisGood]

I haven’t looked at OpenClaw but I get the impression anyone could build it. It doesn’t do anything technically impressive, does it?

[joe_mamba]

>anyone could build it

Then why hasn't anyone else done it before?

With hindsight, it's always easy to say anyone could have done it too, but there's more to product success than just coding and shipping an app out the door.

The first iPhone was built using COTS(commercial off the shelf) parts that Nokia, Ericsson and Motorola also had access to, and SW tools they also had access to, yet Apple won and buried the other companies because their end-product was way more popular with the customer base. I'm sure engineers from Nokia, Ericsson and Motorola also said "we could have done exactly the same thing with the right leadership" when they saw that.

I also say "I could have done that" when I see how the maker of Flappy Bird became a multi millionaire, or how any other top 100 AppStore slop app has 100+ million downloads.

Coding skills are dime a dozen these days. A lot of people can do 95% of these things now. The differentiator between failure and success, comes with the 5% rest: network effects, market know-how, promotion, timing, outreach, UI, UX, luck, etc.

[LPisGood]

I agree it was a good idea and there’s more to product success, but you were specifically talking about coding skill level.

There are some things I could easily say I (and many others) could not build even in retrospect. Solidworks, for example is beyond a lot of people’s skill level and very difficult to build.

Flappy bird and open claw, not so much.

[gavmor]

Many people have! Nanoclaw, LocalGPT, Moltis, Thoth, Q-Claw... the list goes on.

[Dylan16807]

Well your previous comment sure made it sound like you were talking about level of coding skill.