I've always thought the one way to make the world a better place, would be to randomly throw people who say "I just want to make the world a better place" into the nearest active volcano. For maximum effect it would have to be completely random, but semi-frequently, until that trope is finally dead.
Man Sequoia hops from one wave to the next without any thought or worldview. The FTX fawning is still fresh and they are on to the next hypetrain. No lessons learnt, no shame at all.
I think the next gent LLMs will have enough reasoning capability to create tools that glue all these point solutions together. That's more likely to be the true revolution. Looking at this zoo of vendors, many have great point solutions but so much engineering time is spent connecting them together.
When the industry threw in the garbage the concept of an app server (remeber those) they also threw away the tight integration of code, deploy, database and queueing integration, built in monitoring and so on. Love them or hate them, but EJB app servers came with _all_ batteries included when that phrase actually meant something.
They provide no real data, so let's take a look ourselves at even one piece of real data. Go look at copilot's total number of downloads on the plugin marketplace - not dau, not mau, just people ever clicking the install button, hopefully filtered a little so that install on 5 machines by same person does not count as 5 (if not, the number looks even worse). Despite massive hype, full court press, etc, the number is small. The MAU is even smaller.
Then try to convince yourself it will really make a billion at 10-19 a seat by attracting 1000x-10000x the users it currently has, and retaining them monthly, anytime soon.
This isn't the only piece of data that suggests wild optimism at work.
The entire DevOps market, copilot included, is estimated at 10-12 billion now, and won't double until 2030. That includes the effects of AI. The estimates are actually down from a few years ago (it was estimated to reach 25B by 2025 before).
The sort of nonsense handwaving with no backing in data that goes into these articles is impressive.
Even rough back of the envelope says it's currently a pipe dream.
Maybe some future copilot will attract more, but based on current data and trajectory, when I hear this kind of article claim it will make billions with not a single real current piece of data supporting the claim it just comes off as delusional.
Or you know, they could try to provide any real data that supports their claims.
Really though, I think they just hope it's all wrong because they are too vested in the narrative.
I have tons of faith that ai will change development as we know it, and in fact making it happen is a significant part of my job. I just don't at all buy all the wild pronouncements of how soon everyone will make tons and tons of money doing it, nor how much they will make.
> Then try to convince yourself it will really make a billion at 10-19 a seat by attracting 1000x-10000x the users it currently has, and retaining them monthly, anytime soon.
According to the hype train people are losing jobs and getting replaced by copilot so there won't even be 100x the users left.
I've recently had to spend a lot of time using ChatGPT professionally, and holy hell have I found it to be entirely overblown. In my personal experience it sometimes can't even follow basic instructions.
I have an inkling that people really don't know how much of a black box the entire AI experience really is, and why that means that trust for mission-critical items is an impossible sell. I can't rely on it for anything meaningful and I'm no longer convinced I ever will.
For non-mission-critical things like code help, analysis, and hell even observability / cost management, maybe it can eventually be a valuable resource that serves as a partner to a human. But for managing hosting, networks, deployments, builds, terminal commands? Absolutely not, absolutely never.
https://news.ycombinator.com/item?id=33571734