Since I previously shared how we are scaling AI usage so quickly without costs rising at the same rate, there’s been a lot of interest in the engineering details.

So I asked the team to open up the hood and share what we’ve learned.

One thing we’ve always been very good at