9 chunks · 9 episodes
With LLMs now the ability to ask interesting questions is the limiting factor for quality of output. Everyone can generate interesting answers to a given question. So who can come up with the most interesting questions?
...e all the balls it throws you. It's capable of throwing you infinite balls. The limiting factor for juggling is you, which guarantees you'll be at your limit. If you're using agents, your attention will, guaranteed, be saturated.
...u. If you want to maximize the AI system's impact for you, you have to not be a limiting factor.
...er of their insight to an extraordinary degree. In a world of agent swarms, the limiting factor is your judgment and intelligence.
... made software, you made economic value, almost automatically. Software was the limiting factor. The bottleneck. Now LLMs make it no longer the limiting factor.[bm] A lot of software you could make won't add value. No one wants a TEMU for apps.
...e is "if you put a question in a box it can come up with the right answer." The limiting factor for LLMs is increasingly not the intelligence, it's the nescience. Claude is adamant that that's actually a word!
Often the technology isn't the limiting factor in a pipeline. The limiting factor often isn't what we think it is. For example, making a movie faster is not about making more images. It's about ge...
... can't tell the difference in intelligence anymore. Intelligence used to be the limiting factor, now it's not.[cu] There's a ceiling for intelligence in a given application. Peter Wang: "You wouldn't put a Xeon processor on your lawn mower". How...
... harder to steer… but also gives some insulation from disastrous decisions. The limiting factor of corporations is not the intelligence of any one individual, it's the challenge of central coordination with the fractally complex facts on the gro...