David Schur

Check out this session that Prashanth Nanjundappa and I had with TechGig which included topics such as Agentic RAG and how it's one of the best usecases to get started with AI, but also why it can get hard to move to production - and also about how we are leveraging maturity models to meet teams where they are to ensure everyone is fully enabled with AI.

12

EDB Day Session — Part 3: PGD 6.x — Five Nines Isn’t a Number. It’s a Promise. In the same EDB Day session (Dec 8, 2025), the conversation shifted in a way I’ve learned to respect. We stopped celebrating what’s new… and started protecting what must never break. Because in enterprise systems, availability isn’t a KPI. It’s payroll. It’s payments. It’s trust. It’s the quiet agreement we make with the business: “You will not feel our complexity.” That’s where PGD 6.x came into focus—not as “distributed architecture,” but as a platform built for a world where always-on is not ambition… it’s expectation. Five-nines-class availability. Scalability that doesn’t demand heroics. And most importantly: operations that don’t turn into a permanent war room. One capability from the discussion stayed with me because it sounds simple, but it carries years of engineering pain inside it: regional workload transfer to the closest database—ensuring no conflicts occur. That’s the real magic: not “more nodes.” A system that stays coherent when the world is not. And it ties directly back to the heartbeat of this series: Modernization doesn’t slow down because teams lack features. It slows down when change starts to feel unsafe. PGD 6.x is about restoring safety at scale: - availability that holds under pressure - scale that doesn’t create data fights - predictability during patching, upgrades, and planned change - and resilience that feels calm, not theatrical Part 1 was about operational trust. Part 2 was about reachable releases. Part 3 is about keeping both intact—when the platform has to be everywhere and up. Next: Aurora migration thinking—and what changes when you move from AWS-managed constraints to an architecture you can control end-to-end. #EDB #PostgreSQL #PGD #HighAvailability #Scalability #DatabaseReliability #SRE #Modernization

31

Gemini 3 is impressive, but another model beating another benchmark isn’t the story that matters. Most of the conversation right now is still about the demos. The leaderboard wins. The moments that make people say “wow.” And while that’s interesting, it shows how early we still are in actually understanding what this technology enables. We’re reacting to the hype instead of doing the hard thinking. So far, most GenAI work has focused on making our existing processes a little more efficient. Faster documentation, cleaner code, smarter search, improved internal workflows. All good steps, but they force a transformational technology into a world that wasn’t designed for it. We adapt the tool to our current systems instead of adapting our systems to really embrace the tool. There’s a ceiling to this approach. If we keep treating AI like an add-on, that’s exactly how it will behave. Things get interesting when teams start asking a different question: “What would this look like if we designed it for AI from the start?” That’s when the surface area opens up. New workflows. New interaction patterns. New reasoning loops. New boundaries between human and machine work. It stops being about efficiency and becomes about possibility. (Hint: it’s not just about putting a rainbow border and a sparkle emoji around a chat box.) An AI-powered product is very different from a product with an AI feature bolted on. One requires rethinking the entire flow. The other just decorates the old one. This is when the engineering work actually starts. Not at the benchmark stage. Not in the glossy demo. But at the moment when people stop celebrating the model and start asking how to make it reliable, observable, safe, and boring enough that you trust it in production. Benchmarks are fun. The competitive advantage comes from what you redesign once the hype settles. Where do you see the biggest opportunity to redesign something from the ground up with AI, rather than bolting it onto the old flow?

10

3 Comments