The Agent in the Terminal — and the Human Behind It
Claude Code, Cursor, and Codex are reshaping what it means to be a software engineer. The strategic and human question every technology leader needs to answer.
All Posts
23 articles
First upstream contribution from the AI-Farm: verifying vLLM PR #35568 on DGX Spark GB10
A short field note from the AI-Farm: we left a verification comment on vLLM PR #35568 — and found the patch had been quietly fixing our GB10 dispatch for 38 days.
Agents Everywhere: Part 10 - Why Current Approaches Break (From AI Features to AI Infrastructure)
AI is shifting from feature to infrastructure — and that changes everything. When agents become commodities, the value moves to orchestration, control, and system design. The question is no longer what an agent can do, but how the system operates.
Agents Everywhere: Part 9 - Why Current Approaches Break (The Missing Layer)
We have powerful models, flexible tools, and capable agents. And yet systems still struggle at scale. The problem isn't intelligence — it's the absence of a layer that governs execution, manages interactions, and ensures reliability across the whole system.
Agents Everywhere: Part 8 - Why Current Approaches Break (Why Workflows Are Not Orchestration)
Many teams believe they are building orchestration. Most are building workflows. The difference seems subtle — but as systems scale, it becomes the gap between a system that holds together and one that doesn't.
Agents Everywhere: Part 7 - Why Current Approaches Break (The Hidden Cost of Multi-Agent Systems)
More agents should mean more capability. That's the intuition. But in practice, each new agent adds communication paths, dependencies, and coordination challenges. Complexity grows faster than capability — and most teams don't see it coming.
Agents Everywhere: Part 6 - Why Current Approaches Break (And Why Most Systems Never Reach Production)
Most agent systems work in demos. Very few survive contact with production. We explore the gap between what we can build and what we can reliably operate — and why that gap is becoming the defining challenge of the next phase of AI.
You're Already Using AI. The EU Just Started Watching.
The EU AI Act is already in force — and for Nordic companies rushing to adopt AI tools, the compliance clock is running whether you've noticed or not. If you're using AI in your business, you're a 'deployer' with real obligations. Here's what that actually means.
Agents Everywhere: The Complete 5-Part Series on Agentic AI
From the building blocks of agentic systems to the future of AI ecosystems, this 5-part series traces the full arc of where agentic AI is today and where it's heading. Built for practitioners and leaders navigating the shift from AI tools to AI systems.
Agents Everywhere: Part 5 - Open Ecosystems vs Controlled Intelligence — What Happens Next?
Two very different futures for AI agents are taking shape. One is open, flexible, and dynamic. The other is structured, reliable, and controlled. Understanding the trade-offs — and why both will likely coexist — matters for anyone designing AI systems today.
Agents Everywhere: Part 4 - The Rise of Agent Orchestration
The interesting problems in AI are no longer inside the agent — they're around it. Orchestration is emerging as the critical layer that separates experimental demos from production-ready systems. Here's what it means, why it matters now, and what it looks like when done deliberately.
Agents Everywhere: Part 3 - Why Agentic Systems Break (And What That Tells Us)
AI agents don't fail the way we expected. They fail quietly — through interpretation drift, tool misalignment, and coordination breakdowns. Understanding these failure patterns reveals what the next generation of reliable AI systems actually needs to be built on.
Agents Everywhere: Part 2 - Three Deployment Architectures Shaping Modern AI
Most organisations deploying AI agents today fall into one of three patterns. Only one of them scales reliably. Here's a clear breakdown of single-agent systems, multi-agent systems, and orchestrated architectures — what each means in practice, and what to consider before you build.
Agents Everywhere: Part 1 - Seven Powerful Insights Into the Rise of Agentic Systems
Everyone is talking about AI agents. But what does 'agentic' actually mean — and why does it matter now? We explore 7 insights into what's really driving the rise of autonomous AI systems, what most deployments actually look like under the hood, and why orchestration is becoming the defining challenge of the next era.
Vibe Coding Has a Ceiling. You Just Hit It.
You have three browser tabs open — Claude, Cursor, Codex. You're copying context between them, forwarding outputs, catching errors, and re-explaining the same architecture for the third time today. You are not an AI-powered engineer. You are a message queue with a salary. Time to step up to agentic engineering.
Your Data Has Left the Building — What European Companies Risk When They Use US AI Tools
72.6% of Swedish companies use cloud services. 85% of European cloud runs on US infrastructure. Microsoft testified under oath it cannot guarantee EU data sovereignty. The CLOUD Act, Schrems III, NIS2, the EU AI Act, and ISO 42001 are converging — and most Nordic organisations are not ready.
You're Already Using AI. The EU Just Started Watching.
The EU AI Act is already in force — and for Nordic companies rushing to adopt AI tools, the compliance clock is running whether you've noticed or not. If you're using AI in your business, you're a 'deployer' with real obligations. Here's what that actually means.
What Does an Interim IT Manager Actually Do?
When a permanent IT Manager leaves or a company hits a growth inflection point, an interim steps in. But what do they actually deliver? More than firefighting — a good interim brings structure, vendor management, budget discipline and a clear handover plan that leaves the organisation stronger than they found it.
Getting More from Jira: Five Patterns We See in Mature Atlassian Implementations
Most Jira projects start strong and drift into chaos. After implementing Atlassian tools across dozens of organisations, we keep seeing the same five patterns that separate teams who thrive in Jira from those who fight it every day. Here's what mature implementations actually look like.
Multi-Material 3D Printing: What Changes When You Can Use 15 Filaments at Once
Multi-material FDM printing with systems like the Bambu Labs AMS unlocks capabilities that were unimaginable five years ago. Here's what changes for prototyping, production and design — and what your first multi-material project should look like.
AI Agents in the Enterprise: From Proof of Concept to Production
A controlled demo is not a production system. Most enterprise AI proofs of concept die between pilot and scale — not because the AI failed, but because the surrounding infrastructure wasn't ready. Here's the gap we keep seeing and how teams are actually bridging it.
GDPR, NIS2 and ISO 27001: How to Prioritise When Everything Feels Urgent
Three frameworks, one timeline, finite resources. Nordic organisations face overlapping compliance pressure from GDPR, NIS2 and ISO 27001 simultaneously. This is how we help clients triage what actually needs to move first — and what can wait without creating real risk.
How to Win a Public Sector IT Tender in Sweden
Swedish public procurement follows strict LOU rules — and most IT vendors lose on process, not capability. This is the pattern we see in winning bids: how experienced teams structure their responses, price competitively and manage the evaluation phase to maximise their score.