Slalom Build

“Aren’t we Agile?” It sounds like a simple question, but it often leads to the wrong conclusions. I’ve seen it used to justify: - Changing scope mid-sprint - Skipping structure - Reshaping events on the fly But Agile isn’t about removing structure. It depends on it. Without clear boundaries, teams don’t become more flexible—they become unpredictable. I wrote a short piece on where this misconception shows up most and how it impacts delivery. https://lnkd.in/g-jWSXkT #agile #Scrummaster #Leader #Scrum

Excited to announce that I'll be speaking at AWS Summit Sydney 2026! 🎤 📅 13–14 May | 📍 ICC Sydney I'll be presenting session DEV401 – Build Intelligent Memory Systems for AI Agents, where I'll dive into how we can design smarter, context-aware memory architectures that make AI agents truly useful in the real world. If you're attending the summit, I'd love to connect and chat about all things AI, agents, and engineering at scale. See you there! #AWSSummit #AWSSummitSydney #AIAgents #GenerativeAI #AWS #CloudComputing #PublicisSapient

Working on the AWS Agent Registry this week, and one of the most useful things I've done isn't even part of the newly released feature — it's my Claude Code /statusline. Four fields: • Model — which model I'm using • ctx — % of context window used • sess — total tokens this session (input + output) • 7d — weekly subscription usage % + reset date When you're using Coding Agents, context and token usage stop being abstract. ctx tells you when the working memory is getting full. sess shows you the real cost of a loop. 7d stops you running out of runway on a Friday. Observability for your own usage is underrated. The model can't tell you when to /clear — but a good status line can.

7 connections. That’s what some of the most talented young people I work with have on LinkedIn. And we sit here talking about talent shortage. This month it’s 20 years since I graduated from DTU - Technical University of Denmark . If I’ve learned one thing since then, it’s this: talent is overrated without access. At HealthBuddy , we’re working with 22 young talents right now. They are sharp, they understand users, they understand technology, and honestly they understand the future better than most of us. But they don’t have the network. And without that, doors stay closed. So here’s a challenge. If you are hiring, building, investing, or just sitting on a strong network, prove that “we need more talent” is not just talk. Go connect with them. Open one door. Give one opportunity. It takes you 30 seconds and it can change their trajectory. Tagging a few of them here: Frederik Houman-Jensen , Tobias Kold , Magnus Velling Rasmussen, Helena Do , Freja Juul Skafte , Maria Staalhagen Korsgaard , Maria Fischer Poulsen , Maya Elsborg , Jonas Lindgaard Drögemüller , Louise Cathrine Iversen , Mathilde Stanborough , Christoffer Holst , Simon Finderup Sørensen , Malene Howe Andersen , Ulrikke Dein , Maj Hjortskov Jensen , Mette Bertelsen IT-Universitetet i København Let’s see what your network is actually worth.

I've been running experiments using AI to build a software product end-to-end. Not to become a developer. There are enough of those. But to stay honest with myself about what the work actually involves. I come from a UX and Software Craftsmanship background, and I've been fortunate to work alongside engineers who made the depth and discipline of their work visible. But there's a difference between understanding something intellectually and feeling it firsthand. What struck me most wasn't how much AI enables. The barrier to building is lower than it's ever been, and that's real. What I didn't expect was how much the cognitive load would remain, even on a small, self-contained project with no team, no deadlines, and no pressure. Claude Code and nWave do extraordinary things: they compress what would have taken days into hours, sometimes minutes. When they work well, what once took days feels like a different era entirely. But that only holds if you stay in the loop. Letting AI build features blindly is not how this should work. The feedback loop has to be there: you build incrementally, test constantly, and stay close enough to the output to catch things before they compound. The moment you step back, the work gets away from you. And this is the easy version: no competing priorities, no unclear requirements, no stakeholder changes mid-implementation. A significant number of managers leading software teams have never worked directly in the discipline. Not as a developer, not as a UX designer, not as a QA engineer. They've observed the work, reviewed it, and been accountable for it, but they've never been inside it. That gap matters more than most organisations acknowledge. That's where unrealistic deadlines come from. Not bad intent, but a lack of proximity to the work itself: the scoping conversations that need to happen before a line of code is written; the edge cases that only emerge once you're deep in the work; the cost of switching context; the compounding effect of small decisions made under pressure. If more managers had that visibility, their instincts would shift: slower to commit to arbitrary timelines, more focused on the questions that actually matter. Are we solving the right problem? Have we defined it clearly enough? Is this the moment to build it at all? Good prioritisation is hard. It requires understanding what delivery costs. Without that, it's easy to confuse being busy with making progress. AI doesn't remove the need for deep thinking. In some ways it increases it. The judgement required doesn't disappear. It just moves. If you manage software product teams and haven't spent time close to the work, I'd encourage you to try building something small using AI. Not to prove anything. Not to become technical. Just to develop a more grounded sense of what your teams are navigating daily. It's a few hours well spent. #ProductManagement #ProductCraft #AgenticAI #SoftwareDevelopment #Innovation #nWave #ContinuousLearning