This sounds possible but how can we know? It could just as well correlate. Effort of all kinds correlates with success even when it's not obvious or not... linear.
>This sounds possible but how can we know?
If usage correlated with ROI, we would never stop hearing about it from the big AI firms.
The fact is that a bunch of marks and rubes got suckered into buying snake oil.
Sure, you could say all digital transformations are nebulous snake oil and don't deliver but I just don't think this is the case.
That's a pretty low bar if you ask me.
I'm sure someone is going to reply to and say "actually it's not clear and the tech debt is about to explode everything real soon now" or something to that effect, but it's hard for me to explain this theoretically when it is so obvious in practice in my day to day.
ROI is about how much its costs vs the value of the “help.” That’s where the present crisis is. Is the ROI there to pay for the trillions in commitments that have been bet on that ROI? That’s a clear no at this point hence the growing panic.
That's only a potential crisis for those who have made concrete investments.
On the use side, the 'cost' of AI spans more than two orders of magnitude. Looking at recent models (<6mo) with reasonable performance (intelligence index >= 45) on OpenRouter, the output cost ranges from $50/MTok (Fable) to $0.153/MTok (DeepSeek Flash 0731).
From the perspective of a user of LLM/agent assistance, there's very likely a range where the benefits outweigh the costs.
If the ROI for the model developers isn't there, then that just impairs the future trajectory of the field. Current models are just bits that aren't going anywhere, and as long as they can be served (in inference) above their marginal cost they will continue to be so-delivered.
All other things being equal, increasing the speed of some part of the development process will increase the overall pace of development. However, By Amdahl's law that increase will be sublinear, and that is why we should take "pull requests" as an imperfect metric.
We also don't get to pick the form that 'better technology products' take. While we'd probably like to keep cost(/effort) and complexity constant and increase robustness and performance, the market equilibrium might be 'worse is better' and reward whiz-bang features and lower effort.
Also, large companies are still experimenting on how to integrate LLMs into their workflows. Due to the fast cadence of releases, people forget that LLMs became robust (regardless of the capability level/parameter count/data size) enough to use semi-reliably in company-specific ways only 1 year ago. And bigger the org, the slower the process. I don't expect it to settle and get productive used across the majority of very large companies for another year atleast.
High prices are mostly a result of DC capacity. As more and more DCs get built out, prices will drop. At a unit level the inference business is extremely sound regardless.
Right, but we didn't invest a trillion dollars of capital into any of those things in the span of a couple years, thus forcing them to capture value and show a return on such a massive investment.
AI has created a generation of 0.125x engineers.
If you make an effort to use the new tools effectively, the gains are wild. Don't fall into a grumpy luddite trap, the train is leaving the station and you'll struggle to catch up if you don't learn and grow.
Not using AI for software development in 2027 will be like only knowing how to program via punchcard in 2010.
It's new. It's different. It's hard. It's your job. Learn how to use it effectively, embrace the new abundance mindset.
The key question is how much value has been delivered on the other end for a given cost. Well used, tokens would be cheap at 10x the price right now.
If that's true things are truly hopeless for pre-school/school children who are not in position to jump on the train just yet.
This. Except it’s not that hard to get proficient. Either learn it or get left behind same as it ever was in software tech.
Correct, in fact you can tell most AI detractors aren't thinking about this from a _computer science_ point of view. You'll have many people claiming "why prompt an LLM when you can write it and commit it yourself in 45 minutes?" True, you could. But that's a purely _serial_ workload. LLMs are effectively data-independent and they can.. work in _parallel_. Meaning instead of working on a single task, you can juggle multiple simultaneously. So even if you're as fast or faster on a one-on-one basis, you _cannot physically_ be faster than 10 separate sessions running in parallel, working on wholly separate tasks. And that's actually where most 10x gains come from. See Amdahl's/Gustafson's laws.
Also, somewhat amusingly, the “founder” consistently being at the top of the use curve might just have something to do with everyone else also using it, but that more implies people are using it because the chief at the top wants them to use it… not because it’s actually useful. A pattern that’s typical of bad AI deployments.