As someone who is working in this space, I found this article to be quite vague, and couldn't take too much out of it. It would be great to see specifics on one SLO, the associated error budget and what graphs Twitter uses to track this.
And while SLOs are a good start - I assume this will be SLOs across services on latency and availability - the next step is measuring key user journeys. "What % of users get stuck on a page, getting the same error message over and over?". "Where can we see a sharp increase from p95 to p99 on success or latency on a user journey, that we can click into?"
I have a feeling Twitter is still at the first level, as most companies I know are. Clearing this one, the next step is what is even more interesting.