notes and training from a small software studio in denmark. building solutions for problems since 2014.https://scalable.dk/enhttps://scalable.dk/craft/ts-types-in-anger/https://scalable.dk/craft/ts-types-in-anger/branded types, template literals, conditional types. make illegal states unrepresentable, not just inconvenient.Mon, 25 May 2026 12:00:00 GMThttps://scalable.dk/posts/shipping-an-oauth-protected-mcp-server/https://scalable.dk/posts/shipping-an-oauth-protected-mcp-server/adding model context protocol to indebtio. notes from the security review that gated the launch: scope, dcr threats, refresh-token replay, jwt key separation.Sun, 24 May 2026 12:00:00 GMThttps://scalable.dk/posts/enablement-isnt-evangelism/https://scalable.dk/posts/enablement-isnt-evangelism/i wanted to lead the ai enablement work in my area, and got it. the tools are landed, the early adoption is real; the next chapter is widening the circle. here's the shape this job has to take.Sat, 23 May 2026 12:00:00 GMThttps://scalable.dk/posts/spec-kit-in-practice/https://scalable.dk/posts/spec-kit-in-practice/github's spec-kit makes you write a spec before the agent writes code. a few evenings in, here's what changes about the agent loop — and what doesn't change at all.Mon, 18 May 2026 12:00:00 GMThttps://scalable.dk/posts/meaningful-reporting/https://scalable.dk/posts/meaningful-reporting/performance metrics in one tool, incidents in another, vulnerabilities in a third, change requests in a fourth. a meaningful weekly report is a deterministic synthesis from source systems — humans curate, scripts collate.Mon, 11 May 2026 12:00:00 GMThttps://scalable.dk/posts/template-literal-types-in-practice/https://scalable.dk/posts/template-literal-types-in-practice/template literal types went from cute trick to load-bearing the day i used them to derive route params from a path. one pattern, one cost, a few wins.Mon, 04 May 2026 12:00:00 GMThttps://scalable.dk/posts/agentic-chat-over-synthetic-bank-data/https://scalable.dk/posts/agentic-chat-over-synthetic-bank-data/first useful workflow on the ollama+n8n homelab: a synthetic bank api, an n8n agent that calls it, a dashboard with a chat bubble. what worked, what didn't.Wed, 29 Apr 2026 12:00:00 GMThttps://scalable.dk/ai/claude-agent-sdk/https://scalable.dk/ai/claude-agent-sdk/build your own coding agent in typescript. tool use, sub-agents, hooks, permissions.Mon, 27 Apr 2026 12:00:00 GMThttps://scalable.dk/posts/ollama-and-n8n-on-proxmox/https://scalable.dk/posts/ollama-and-n8n-on-proxmox/setting up local ai on a 3-machine proxmox homelab: ollama, n8n, the ram gotcha that bit me, the cpu-only model shortlist, and what the first prompts actually returned.Sun, 26 Apr 2026 12:00:00 GMThttps://scalable.dk/posts/the-translators-job/https://scalable.dk/posts/the-translators-job/modernization loses to features when the why doesn't travel. metaphors are how leaders make technical reasoning portable into rooms they're not in.Sun, 26 Apr 2026 12:00:00 GMThttps://scalable.dk/ai/agentcore/https://scalable.dk/ai/agentcore/aws bedrock agentcore in the browser — no account requiredFri, 24 Apr 2026 12:00:00 GMThttps://scalable.dk/scale/team-topologies/https://scalable.dk/scale/team-topologies/the four team types, interaction modes, and why conway's law wins anyway.Fri, 24 Apr 2026 12:00:00 GMThttps://scalable.dk/ai/agentic-patterns/https://scalable.dk/ai/agentic-patterns/skills, extensions, personas, handoffs — the building blocks of multi-turn agent systems.Fri, 24 Apr 2026 12:00:00 GMThttps://scalable.dk/ai/agentcore/what-is-agentcore/https://scalable.dk/ai/agentcore/what-is-agentcore/the seven primitives, the boring parts aws decided to own, and the shape of a minimal agent.Fri, 24 Apr 2026 11:59:00 GMThttps://scalable.dk/ai/agentic-patterns/what-are-agentic-patterns/https://scalable.dk/ai/agentic-patterns/what-are-agentic-patterns/the shape of the problem. why patterns, not frameworks. the four primitives you'll see in every later lesson.Fri, 24 Apr 2026 11:59:00 GMThttps://scalable.dk/ai/agentcore/runtime/https://scalable.dk/ai/agentcore/runtime/define an entrypoint, handle a payload, emit events. the agent app object, up close.Fri, 24 Apr 2026 11:58:00 GMThttps://scalable.dk/ai/agentic-patterns/skills/https://scalable.dk/ai/agentic-patterns/skills/a tool call plus the prompting, validation, and return-shape glue around it. reusable across agents.Fri, 24 Apr 2026 11:58:00 GMThttps://scalable.dk/ai/agentcore/memory/https://scalable.dk/ai/agentcore/memory/session memory vs long-term memory. events, strategies, retrieval across sessions.Fri, 24 Apr 2026 11:57:00 GMThttps://scalable.dk/ai/agentic-patterns/extensions/https://scalable.dk/ai/agentic-patterns/extensions/memory injection, retrieval, safety, telemetry. the pattern equivalent of middleware.Fri, 24 Apr 2026 11:57:00 GMThttps://scalable.dk/ai/agentcore/gateway/https://scalable.dk/ai/agentcore/gateway/take an openapi spec (or a lambda) and make it callable from your agent as an mcp tool.Fri, 24 Apr 2026 11:56:00 GMThttps://scalable.dk/ai/agentic-patterns/personas/https://scalable.dk/ai/agentic-patterns/personas/a bundle of prompt, voice, allowed skills, and guardrails. swap it and the same loop behaves differently.Fri, 24 Apr 2026 11:56:00 GMThttps://scalable.dk/ai/agentcore/identity/https://scalable.dk/ai/agentcore/identity/workload identity, oauth to third parties, iam without writing policy by hand.Fri, 24 Apr 2026 11:55:00 GMThttps://scalable.dk/ai/agentic-patterns/handoffs/https://scalable.dk/ai/agentic-patterns/handoffs/how one agent passes control to another with just enough context. most 'the agent got confused' bugs live here.Fri, 24 Apr 2026 11:55:00 GMThttps://scalable.dk/ai/agentcore/browser-and-code-interpreter/https://scalable.dk/ai/agentcore/browser-and-code-interpreter/two managed tools you get for free — when to reach for each.Fri, 24 Apr 2026 11:54:00 GMThttps://scalable.dk/ai/agentic-patterns/loops/https://scalable.dk/ai/agentic-patterns/loops/the control flow inside a single agent turn. stopping criteria, budget, and failure modes.Fri, 24 Apr 2026 11:54:00 GMThttps://scalable.dk/ai/agentcore/observability/https://scalable.dk/ai/agentcore/observability/spans, tool-call boundaries, cost attribution. the parts cloudwatch doesn't show you.Fri, 24 Apr 2026 11:53:00 GMThttps://scalable.dk/ai/agentic-patterns/evals/https://scalable.dk/ai/agentic-patterns/evals/offline scorecards, trace-based checks, and the gap between demo and production.Fri, 24 Apr 2026 11:53:00 GMThttps://scalable.dk/ai/agentcore/shipping-to-aws/https://scalable.dk/ai/agentcore/shipping-to-aws/what changes when you leave the sandbox. cli, iam, costs, and the smallest deploy that works.Fri, 24 Apr 2026 11:52:00 GMThttps://scalable.dk/posts/the-production-readiness-review/https://scalable.dk/posts/the-production-readiness-review/a one-hour meeting with a structured checklist that catches the last 20% of failures before they're customers' problem.Mon, 20 Apr 2026 12:00:00 GMThttps://scalable.dk/posts/the-editor-as-a-force-multiplier/https://scalable.dk/posts/the-editor-as-a-force-multiplier/the editor is the tool you'll touch ten thousand hours. small wins in it pay back forever.Mon, 13 Apr 2026 12:00:00 GMThttps://scalable.dk/posts/eval-driven-prompt-iteration/https://scalable.dk/posts/eval-driven-prompt-iteration/treat prompt engineering like test-driven development — write the failing case before you change the prompt.Mon, 06 Apr 2026 12:00:00 GMThttps://scalable.dk/posts/illegal-states-unrepresentable/https://scalable.dk/posts/illegal-states-unrepresentable/a small refactor: replace the optional-fields-and-flags pattern with a discriminated union. invariants enforced by the compiler.Mon, 30 Mar 2026 12:00:00 GMThttps://scalable.dk/posts/the-senior-ic-who-isnt-a-lead/https://scalable.dk/posts/the-senior-ic-who-isnt-a-lead/the senior individual contributor who isn't on a management track is a real role with real value, and most companies bury it.Mon, 16 Mar 2026 12:00:00 GMThttps://scalable.dk/posts/tool-use-error-handling-patterns/https://scalable.dk/posts/tool-use-error-handling-patterns/throwing inside a tool handler is the wrong abstraction for an agent loop. here's a result shape that works.Mon, 09 Mar 2026 12:00:00 GMThttps://scalable.dk/posts/git-bisect-the-bug-finder/https://scalable.dk/posts/git-bisect-the-bug-finder/an underused git command that finds the commit that introduced a bug, in roughly log(n) tries.Mon, 02 Mar 2026 12:00:00 GMThttps://scalable.dk/posts/structured-outputs-claude-vs-openai/https://scalable.dk/posts/structured-outputs-claude-vs-openai/two providers, two approaches to constrained output generation, and the small-but-real differences when you actually rely on them.Mon, 23 Feb 2026 12:00:00 GMThttps://scalable.dk/posts/prompt-caching-real-numbers/https://scalable.dk/posts/prompt-caching-real-numbers/anthropic's prompt caching cuts the per-call cost of long-context calls dramatically. here's what it actually saved on a workload i measured.Mon, 09 Feb 2026 12:00:00 GMThttps://scalable.dk/posts/pre-mortems-vs-post-mortems/https://scalable.dk/posts/pre-mortems-vs-post-mortems/imagine the launch failed. you have an hour. write the post-mortem now, before the launch.Mon, 02 Feb 2026 12:00:00 GMThttps://scalable.dk/posts/typescript-types-as-documentation/https://scalable.dk/posts/typescript-types-as-documentation/comments rot. types compile. branded types let you encode intent into the type system so the next reader can't miss it.Mon, 26 Jan 2026 12:00:00 GMThttps://scalable.dk/posts/claude-code-hooks-and-agent-design/https://scalable.dk/posts/claude-code-hooks-and-agent-design/claude code ships hooks instead of opinions. that's a design choice, and it has implications for anyone building their own agent harness.Mon, 19 Jan 2026 12:00:00 GMThttps://scalable.dk/posts/training-as-a-way-to-learn/https://scalable.dk/posts/training-as-a-way-to-learn/an experiment for 2026: one publish a week, on top of everything else. the bet is that making training material forces me to explain, and that's how i learn it.Mon, 05 Jan 2026 12:00:00 GMThttps://scalable.dk/posts/welcome-in/https://scalable.dk/posts/welcome-in/a first post about what this site is, and a few things i'm planning to write about.Sun, 28 Dec 2025 12:00:00 GMT