There were A16z articles I remember reading and being better off for. This… I struggle to… separate the managerial crack from what insights remains of a field of engineering.
It isn’t helped that at random sections it feels like I’m reading Claude speak. Every instance of which the article is better off by its excision.
The combination of statements that sound like voodoo, followed by the articulation of something obvious (“you need to have a way to encode what good is.”), is disheartening to say the least.
At one point I wondered if this article was actually being subversive; attempting to bring back rationality by stating the need for Loops to be constrained by economic limits
> “ The missing piece is boring yet necessary: cost per iteration, progress per dollar, a curve someone can see while the loop is still running.”
Where engineering used to be, a religion seems to be standing. Opinion and positions instead of measurement and clarity.
i’m not trying to dismiss this article entirely as i did get value around from discussion around verifying
but this coming from a VC firm invested in token producing companies and quoting the “genius” behind openclaw makes me think this is just an psyop of over engineering to get ppl to blindly spend tokens on the LLM slot machine
Put another token in the slop machine!
Not really; looping is a requirement when there's an imperfect generator: LLMs. I read the article some days ago and included the citation in my own.
For a longer explanation why, here's mine:
https://memseek.ai/blog/the-loop-has-always-been-there/
ps: not affiliated with a16 or similar
I failed to think of any polite words to express my feelings about this article.
It's nonsense written by a chinstroker, quoting the guy whose ilk causes Github to crash every damn week.