Build for the model that will exist six months from now

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 1:10:28

   was different about cloud code that kind of won out and just caught like wildfire internally at the time coding looked very different if you thought about ai coding people thought about autocomplete. I think there was some very early agents, but it was kind of a secondary thing next to the autocomplete. And oftentimes it was used for Q&A. It wasn't really used for coding. And so when people thought about AI for coding, it was just a completely different product that you imagined. It was like tab to autocomplete. And I thought that's what it was. And I thought that was kind of the state of it. And then Ben, who was my manager at the time, he pushed me to think a little bit bigger. And he, I think, really internalized because, you know, he was there from the beginning of Anthropic. And, you know, he was like at other labs before. And so I think

2. 1:11:23

   he really understood the scaling laws kind of internally about how quickly the models are improving. And so he actually pushed me really hard to be like, don't build for the model of now um and so honestly for a long time quad code was not a great product and even when it was used internally i used it for maybe like 10 of my code or something like this you know use it sometimes but it really just can't do most things because the model is not capable enough and then at some point we released uh sonnet and opus 4 and i think this was maybe march of this year and the product just worked and we saw this in kind of the usage data and i saw this in my own coding I started to be able to use it for probably like half of my code.

3. 1:12:04

   And this was totally borne out because this was actually like literally six months after starting the project. This was the timeline. And, you know, at this point, most of quad code is written using quad code. I think it's like 80 or 90%. If you look at teams at Anthropic, there's some teams that, you know, like 90% of their code is written using quad code. This is not just our team. and if you look at the impact on productivity even though anthropic has tripled i think since the start of the year uh productivity per engineer measured in you know like poor cause we were talking about this before has grown like almost 70 percent per engineer because of quad code and so you know like as a product person i usually think a step ahead and i think being at a lab you have to actually think kind of differently you have to think a step ahead um not two steps