byzantinegene

china?

not if china gets into the picture

exactly, the people who copy-paste AI responses are typically the ones who don't do due diligence in verifying the responses.

I would be curious to see if the current AI valuations continue to balloon if everyone eventually shifts to local models. Hardware demand would still be high, but circular financing might no longer continue,

why not? if he's hired to do the task he's supposedly an expert in, and that is now done by AI, what's stopping the company from replacing him with any other person at a lower cost that operates the same AI?

this seems to happen far more than i would like

this pretty much sums up what i feel about AI currently. It made my life significantly easier for most tasks I already breeze through, yet tasks I used to struggle with are still the equally difficult

vast majority of average users don't use llms for coding, and for those purposes, local llms with low param count are a far cry from SOTA models.

i'm not sure the tradeoff in code readability is worth it as of now.

i would argue we don't need anything near Opus to be productive. Sonnet is plenty productive enough

what are those limitations we're talking about? seems most of those the original limitations that people complained about were resolved through workarounds like tools and skills which are more software-engineering than llm advancement.

it is a fruitless endeavour to try to appeal to a crowd that does not and will never understand the fundamentals of how llms work.

you have 35 years of experience and have already built up the learning capability and general framework to acquire new knowledge. you know how to use agentic coding as a tool to supplement your work. the juniors who start today don't have that, they overrely on agentic coding and do not know what they don't know

we're already doing that, it's called distillation and how models like deepseek are trained.

factor in depreciation and energy costs, and a subscription might just be cheaper.

your spec is a guideline, not something the LLM has to adhere to. it is definitely not guaranteed to work without error

you don't need to build systems from scratch to acquire problem-solving skills. even routine maintenance problems require to dig into documentation, look at github issues, and do root-cause analysis. These skills are eliminated from reliance on AI and there is no fallback if one never acquired them in the first place.

seems like what you're suggesting to token efficiency is to simply use less of it?

in theory yes, but all the money is going into AI or AI-adjacent startups that no one would actually build a product that solves problems if it doesn't incorporate AI in it.

the near term is not an issue, because most ai code is still reviewed by experienced engineers with experience. the problem comes in the future where junior engineers who never acquired enough experience to handle engineering problems

IMO this is the only way model providers can survive in the long run, bank on their users overreliance on them resulting in diminishing capabilities. This gives them leverage to increase prices without any pushback

you're right but for now it doesn't matter if both competitors are running on infinite vc money, we as consumers benefit from it. it only matters if they cause negative externalities in the meantime

how have you coded before the era of llms?

it's not worth $5000 if people are not willing to pay that amount for it

good article!

the problem now is how much can they hike the costs without people just going back to coding by hand.

too bad search is gradually made obsolete by ai

even more reason to cheer for the rest of us

from experience, AI is bad at TDD. they can infer tests based on written code, but are bad at writing generalised test unless a clear requirement is given, so you the engineer is doing most of the work anyway.

seems like a security risk