The tech itself works, unlike blockchain, but the narratives are way overblown. I know a few small companies that have legitimately let people go without even having a proper replacement.
The leadership then hyped up Devin or whatever, only to find the whole workplace in chaos after a few months. All this was done to attract VC money, and now they're hiring for the same position again.
What about blockchain tech doesn't work? It's a relatively more tame technical problem than any GenAI.
Usability. The use case of GenAI is pretty clear at this point. With blockchain, the big question was, "Okay, we have this cool tech, but what do we do with it now?" If you have to think about what problem a tool solves, it's not a good tool at all.
To some extent, usability.
But since the target audience was always technical (although gullible) people, the incentive to make it useful beyond coin purchases never manifested.
But moreso: scalability.
The blockchains that are worth anything don’t scale to the level of total worldwide transactions per second.
I’ve heard “But Lightning!” — sure, so before I can send any money, I have to find a fellow to relay my transactions against.
I mean, Our Lord and Saviour The Almighty Blockchain (and its offshoot technologies, like DAOs) worked, for some value of worked; it just wasn't particularly useful. I'd put LLMs largely in the same bucket for now.
> I know a few small companies that have legitimately let people go without even having a proper replacement.
In terms of collateral damage, blockchain nonsense is probably still winning, at least for now, though its victims were often at least somewhat complicit in their own destruction.
I wouldn't be as harsh on LLMs as I was on blockchain. Blockchain didn't solve a single problem I had or understood. In contrast, LLMs already solve a few boring and annoying problems for me, which I find quite valuable.