Comment by storus
6 days ago
Soon all coding will look like L3 support - debugging something you've never seen before, and under pressure. AI is really taking away the fun parts from everything and leaving just the drudgery in place.
6 days ago
Soon all coding will look like L3 support - debugging something you've never seen before, and under pressure. AI is really taking away the fun parts from everything and leaving just the drudgery in place.
“What do you mean you want to think about our architecture? Just get the LLM to do it, and we’ll get it to fix it if anything goes wrong”
“No we’re not allocating any time to thinking about the design, just get the LLM to do it”
I’m so excited for the bleak future.
The key is to figure out how to move up the ladder of abstraction. You don't want to be a "coder" in a world where AI can code, but you do want to be a "person who makes software" in a world where making software just got easier.
Most people who chose the profession don’t want that, though. They like the coding, and dislike managing.
How does "person who makes software" imply "managing"?
I understand that "coding" is the fun part for lots of people, especially younger people. This is me as well, so I'm definitely sympathetic to it, and feel quite a bit of sadness about this.
Lots of people also enjoy woodworking and machining by hand, but that's not how most furniture or machines are made.
If I were independently wealthy, I might well spend some of my time making artisan software, but as a professional and entrepreneur, I'm going to try to use efficient tools for the job.
4 replies →
People said the same about VB style coding then low-code and now AI.
They have been wrong every time and will continue to be wrong.
This feels different; I asked DeepSeek R1 to give me an autoregressive image generation code in pytorch and it did a marvelous job. Similar for making a pytorch model for a talking lip-synced face; those two would take me weeks to do, AI did it in a few minutes.
Autoregressive LLMs still have some major issues like over-dependency on the first few generated tokens and the problems with commutative reasoning due to one-sided masked attention but those issues are slowly getting fixed.
People used to tell me all the amazing things no-code and low-code was able to do as well.
And at the end of the day they went nowhere. Because (a) they will never be perfect for every use and (b) they abstract you from understanding the problem and solution. So often it will be easier to just write the code from scratch.
3 replies →