https://altitudes.cloud/en/insights/ Writing from the team that also builds the platforms. One practical write-up per month. en-GB altitudes.cloud Fri, 29 May 2026 15:53:07 GMT https://altitudes.cloud/en/insights/dora-evidence-without-evidence-theatre/ https://altitudes.cloud/en/insights/dora-evidence-without-evidence-theatre/ Thu, 15 Jan 2026 00:00:00 GMT Most DORA evidence packs we inherit are PDF graveyards. The auditor wants four things. The team can give them four things. Everything in between is theatre. This is the minimum that works. noreply@altitudes.cloud (Maikel Vlasman) REGULATORY https://altitudes.cloud/en/insights/finops-savings-not-where-you-think/ https://altitudes.cloud/en/insights/finops-savings-not-where-you-think/ Thu, 08 Jan 2026 00:00:00 GMT Most mid-market cloud bills carry 40 percent waste. Most of it is in three line items. None of them are EC2. The savings story most consultancies tell is the wrong story. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/agentic-pilots-that-scale/ https://altitudes.cloud/en/insights/agentic-pilots-that-scale/ Thu, 18 Dec 2025 00:00:00 GMT The pilot worked. The demo got applause. Six months later, the rollout has stalled, the dashboard is dead, and no one wants to be the one who turns it off. Three failure modes we keep seeing in EU mid-market. noreply@altitudes.cloud (Sebastiaan van Parijs) AI ENABLEMENT https://altitudes.cloud/en/insights/finops-tagging-debt/ https://altitudes.cloud/en/insights/finops-tagging-debt/ Thu, 21 May 2026 00:00:00 GMT Every untagged resource today costs three times to fix: once to find it, once to attribute it, once to justify the bill it produced. Teams that tag retroactively spend around three hours per resource. Teams that enforce at creation spend three seconds. The gap is not discipline — it is architecture. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/kubernetes-cost-visibility/ https://altitudes.cloud/en/insights/kubernetes-cost-visibility/ Thu, 07 May 2026 00:00:00 GMT Your cluster runs at 68 percent CPU utilization. Capacity planning says that is healthy. The FinOps review says you are spending €180,000 a month on compute. Neither number tells you which team owns which slice. That gap is an engineering problem, and it stays until you build the attribution layer. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/commitment-discounts-math/ https://altitudes.cloud/en/insights/commitment-discounts-math/ Thu, 23 Apr 2026 00:00:00 GMT Reserved Instances get the headline at re:Invent. Savings Plans are harder to explain on a slide. Spot looks like the obvious choice for anything that can tolerate interruption. On the estates we audit, the wrong commitment strategy costs an average of 11 percent of the compute bill. The math is not hard, but it starts with a different question. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/landing-zone-mistakes/ https://altitudes.cloud/en/insights/landing-zone-mistakes/ Thu, 12 Mar 2026 00:00:00 GMT The most expensive landing zone project is not the one you never started. It is the one you started wrong and retrofitted at 80 engineers. Retrofitting a flat-network, single-account structure onto a running estate costs an average of 14 weeks of platform team capacity and leaves a security debt that takes another quarter to clear. Starting with a correct structure costs four weeks. noreply@altitudes.cloud (Maikel Vlasman) PLATFORM ENGINEERING https://altitudes.cloud/en/insights/finops-cost-accountability-org-design/ https://altitudes.cloud/en/insights/finops-cost-accountability-org-design/ Thu, 26 Mar 2026 00:00:00 GMT The average engineering team we work with runs 3 to 5 cost visibility tools before they call us. The average unallocated cloud budget on those same estates is 28 percent. The tools are not the problem. The org is. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/platform-engineering-vs-devops/ https://altitudes.cloud/en/insights/platform-engineering-vs-devops/ Thu, 09 Apr 2026 00:00:00 GMT The question we hear most from engineering leaders thinking about team structure is whether they should be doing DevOps or platform engineering. The framing is wrong. Platform engineering is what DevOps looks like when it works at scale. Most teams asking the question are already somewhere on the path. noreply@altitudes.cloud (Sebastiaan van Parijs) PLATFORM ENGINEERING https://altitudes.cloud/en/insights/nis2-dora-platform-engineering/ https://altitudes.cloud/en/insights/nis2-dora-platform-engineering/ Thu, 26 Feb 2026 00:00:00 GMT The legal analysis of NIS2 and DORA is thorough. The GRC consulting analysis is thorough. The platform engineering analysis is not. Most teams know they need to comply. Very few know what that means for the services they build and operate. This is the platform engineering read. noreply@altitudes.cloud (Maikel Vlasman) REGULATORY https://altitudes.cloud/en/insights/finops-maturity-assessment/ https://altitudes.cloud/en/insights/finops-maturity-assessment/ Thu, 22 Jan 2026 00:00:00 GMT The FinOps Foundation's crawl/walk/run model is useful as a map. It is less useful when every team self-reports as walk. Here is a one-day assessment that produces an honest score, a prioritised improvement list, and a baseline you can measure against. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/kubernetes-showback-vs-chargeback/ https://altitudes.cloud/en/insights/kubernetes-showback-vs-chargeback/ Thu, 05 Feb 2026 00:00:00 GMT The ambition is chargeback: teams own their compute costs, invoiced internally. Most organisations implement showback instead — you can see your costs, no one is invoiced — and call it chargeback. The sequence is usually right. The naming is the problem. noreply@altitudes.cloud (Danny Zak) FINOPS https://altitudes.cloud/en/insights/internal-developer-platform-build-vs-buy/ https://altitudes.cloud/en/insights/internal-developer-platform-build-vs-buy/ Thu, 14 May 2026 00:00:00 GMT Most internal developer platform discussions are framed as build versus buy. There is a third path most teams end up on without deciding to: assembling from open-source primitives. Each path has different maintenance costs, different capability ceilings, and different adoption dynamics. The decision turns on one question most teams skip. noreply@altitudes.cloud (Maikel Vlasman) PLATFORM ENGINEERING https://altitudes.cloud/en/insights/llm-production-readiness-eu/ https://altitudes.cloud/en/insights/llm-production-readiness-eu/ Thu, 28 May 2026 00:00:00 GMT The US-focused production checklists for LLMs cover latency, evals, fallbacks, and cost. For EU enterprise teams, three more dimensions apply: data residency under GDPR at inference time, AI Act obligation classification for your specific use case, and cost governance for GPU workloads that behave nothing like the compute you have been managing. The extra dimensions are not bureaucratic overhead. They are the parts that break first. noreply@altitudes.cloud (Sebastiaan van Parijs) AI ENABLEMENT