Very interesting post from the founder of Three.js
I hope we will not receive the same from our beloved @Deltakosh
In his feelings there are a LOT that apply to me as well
Frustration of seing (my) 1 month of work being done in 20 minutes
Loosing the pleasure of searching/coding/looking for the best way to put an idea in code
Feeling of becoming less and less valuable
Tired of seeing people on LinkedIn who didn’t know how to code three weeks ago but are suddenly fluid dynamics simulator developers () or such stuff
But also there is hope. Let’s don’t forget what’s nice :
I’ve never been that efficient into evolving my code (at work)
On personal projects, I’ve revived TONS of projects which were buried in their GitHub grave for years and are now, from a day to another, back to life with state of the art code update
(New) Ideation and fast POC creation has never been that fast and that exciting (at least for me)
He is NOT the founder of three.js lol
He is the founder of Three.js Roadmap which I have no idea what it is
I’m not quitting on the contrary I want to double down on Babylon with Lite. THis is the way to go: Provide the most efficient framework that humans and agents can bet on.
I think it’s a disruption how fast AI gets you from nothing to a decent PoC. But it’s still that, a PoC. It may be prettier than the demos we could do before without AI for the same effort, but from that demo to a real product there’s a lot of human involvement still required.
Of all these factors, I think the most significant is “Loosing the pleasure of searching/coding/looking for the best way to put an idea in code”
I do think, however, that this experience varies depending on age. I’m not exactly young anymore (54 years old) and I’ve been programming for over 40 years; I have been around the block a few times, so I’m not affected to the same extent as someone just starting out in the field. At the very least, this is offset by the fact that I can now tackle tasks and projects that I would never have considered before the advent of AI.
It’s actually really sad, because the guy is quite accomplished; Google says he even worked for Jeff Bezos and Elon Musk at certain points in his life. He creates interesting shaders (though they get like 2 fps in my browser, even on my super-powerful PC), but I can see a tricky way he makes people think he’s the founder of Three.js. For a long time, I also thought it was this guy, not Ricardo Cabello (mr.doob). On one hand, it’s my own fault for not paying attention to details, but on the other hand, it’s hard for him to pretend he didn’t want this wording to be read ambiguously. I’m not judging, it’s just kind of sad.
Honestly the "a month of work in 20 minutes" thing is real, I just don't believe the 20 minutes. It's 20 minutes to get something that looks done, and then two weeks finding out why it's quietly wrong. The code compiles, the tests pass, the demo looks great on LinkedIn, and the test was checking nothing the whole time.
Funny thing is the fluid dynamics guys from three weeks ago are going to be fine. They'll keep posting demos. The ones who get burned are the teams who try to ship that demo.
And mrdoob 3 hours in with Opus 5.5 is the most honest AI benchmark I've seen all year. Make a video, 3 hours, still going. That's the real number nobody puts in the launch post.
So no, AI won't kill us all. It just made us all into very fast reviewers of code we didn't write and don't fully trust. Same job, worse vibes, way more output.
is right though, frameworks that are boring and predictable win here, because the agent is just as confused by magic as we are.
This is so funny. I just been dismissed by a company that believes AI will do everyone’s job faster and better. I am so happy to see these moves. Now I can focus on potatoes! My long life dream!
Thank you AI I love you
TLDR; I’ve been working professionally with AI for a while now and it is very useful ( I choose not to do it in my personal/hobby because I love coding and I believe there is still room for useless curated human creativity).
My experience is that I have a Junior coding pal. Great to thinker with but I cannot trust his code. It is so savvy, confident and has always an answer for everything ( in a very verbose way, wonder why… cof cof… tokens driven monetization)… but it breaks more stuff than added value out of the box. On the other hand it can behave as a solid Senior Engineer providing right on code snippets and very thoughtful reviews of actual performance and O(n) optimizations. A dumb loop here a unused variable there. Very good at that indeed.
For for complex long and multidisciplinary issues… no way it is not even close to be able to code by itself with prompts for really unique proprietary code bases. LLM’s are data/stored information, try/fail inference engines, not sources of universal truths and definitively not capable of creating unique ingenious solutions without human interaction/validation. There is over 8 Billion humans alive atm. If you ever though… hmmm I am the first one to think about something… well think again… math/logic/odds will be against you on that ( obviously one needs to be the first one, but to be bold enough to be absolutist on it, is being quite cocky and bit over our heads).
Sorry for the ramble. I am waiting on my potatoes to grow, and it takes a bit.
Our past developer job is dead. It’s time to move to other positions: programming director, project manager, UX manager, product owner, designer, storyteller, QA …
) or such stuff
that!
