@nate_5126% Is the number that got me. Out of every engineering team that's brought AI into the fold, only that sliver has built the infrastructure to run an AI-Native SDLC at real scale. The other 94%? They've got adoption and not much else to show for it yet. That tension is basically the whole reason the Atlassian State of AI SDLC Digital Summit exists, and I was in the room for it last week. Mike Cannon-Brookes (CEO + Co-Founder at Atlassian) and Guillermo Rauch (CEO of Vercel) spent their session circling that gap. I rewound their back-and-forth more than once, not because it was confusing but because the way they framed the problem kept landing harder on the second listen. The through-line for the day: the edge isn't in whether you've adopted AI, it's in whether you've wired your context (projects, owners, dependencies, decisions) tightly enough to the agents actually doing the work. Tamar Yehoshua (Atlassian) and Elena Verna (Lovable) went deep on judgement and speed, and Tamar's line stuck with me: "When you can develop so fast, you have to pick what you're developing and make sure you're doing the right thing." Justin Reock (DX) and Uma Namasivayam (Dropbox) pushed past the vanity numbers, asking what actually signals progress once you look beyond PR volume and token spend. Matt Canham and Ming Wu brought the scale receipts: 1M agents across 6K engineers at Atlassian. Their closer: "The agent is interchangeable. The system of record and the context underneath it are not." If you're trying to get from AI adoption to real delivery, there's a lot worth digging into. Every session is on demand now → https://fandf.co/4y0yjde
View original post
