This is not a criticism he really mentions explicitly, but the issue I see with the valuation is that these products cannot be used in a production system by a responsible engineer. As in, I can't have LLMs autonomously plugged in as part of my product. The failure modes are not ever going to be predictable enough. Now, Microsoft will probably be able to charge a fortune at enterprise level, and managers will dream of a day that the LLMs can replace all those weirdo devs at their company, but that'll stay a dream.
All of the valuable uses are personal. It makes me personally feel more productive at work. It helps me personally understand some topic better. It gives me an idea in a personal project.
That's all really cool, but that is not what the valuation is about. The valuation is about a false science fiction and hype bubble about agentic this or that or AGI or whatever, and this is driving very questionable decisions for wasting possibly trillions of dollars and tons of energy.
The plus side is that there is some really cool personally useful tech here, and we will probably end up with very good open source implementations and cheaper used GPUs once the bubble bursts.
You also can't responsibly have a prolific apprentice (intern, first or second year, etc.) plugged directly to production.
On the other hand, today most money for SWEs goes to people who aren't "staff engineer" level. What if most money went to staff engineers who directed these interns and paired with new human apprentices to learn staff eng?
In the past few thousand years, apprentice/journeyman/mastery of trade was how trades worked, with an "up or out" model that kept the role pyramid the right shape for skill to shape the outcomes.
These days, far too many careers stay apprentice skill the entire career, mostly thanks to enterprises failing at engineering management as a skilled trade, so being unable to value and raise staff engineer caliber contributors. The enterprise learning machine is broken by false "efficiency".
LLMs change this. Staff engineer caliber SWEs are able to direct these indefatigable assistants, as if each staff engineer has a team of 10 who never need mental breaks to remain productive. There will of course be some number of junior devs who themselves have enough affinity for the role they will want to stick with the apprentice model and work to the staff engineer level. (And will always be solo or boutique teams of app/saas SWEs.)
As for the enterprise engineering management that couldn't tell the difference between a staff engineer and an apprentice, the LLM multiplies the difference to the point the outcomes are evident even to a non technical observer.
So one possible timeline for this is a raising of the median human skill level by attrition of those unskilled enough or unable to think critically enough to leverage the machine assistants as force multipliers or unable to survive directly mentored skill-up training and observation from the staff engineers.
You talk about personal value. Roughly, I agree with you completely, and am adding who I think those persons could be (or have to be given the current level of "thinking" by these tools) for the hype to deliver on value. (At a higher level of machine, closer to "AGI", this scenario changes.)
As is evident by downsizing a mediocre team and observing output go up and work more reliably, these forces could, if playing out this way, make dollars per human go up, productivity go up, quality go up, and enable a return to the millennia-proven model of apprenticeship for the trade.
All of the valuable uses are personal. It makes me personally feel more productive at work. It helps me personally understand some topic better. It gives me an idea in a personal project.
That's all really cool, but that is not what the valuation is about. The valuation is about a false science fiction and hype bubble about agentic this or that or AGI or whatever, and this is driving very questionable decisions for wasting possibly trillions of dollars and tons of energy.
The plus side is that there is some really cool personally useful tech here, and we will probably end up with very good open source implementations and cheaper used GPUs once the bubble bursts.