https://chapsoft.com/cairns Daily trail markers from building production AI. Short technical notes, lessons learned, and observations. By Russ Kohn. en-us info@chapsoft.com (Russ Kohn) info@chapsoft.com (Russ Kohn) Wed, 08 Apr 2026 10:51:00 -0700 https://chapsoft.com/cairns A model's competence in one domain doesn't transfer to the next — but it all too often feels like it does. Never assume the model knows what you haven't explicitly had it remember, or find. Context in the window is not attention in focus. A model's competence in one domain doesn't transfer to the next — but it all too often feels like it does. Fight the temptation toward complacency. Never assume the model knows what you haven't explicitly had it remember, or find.

]]> Wed, 08 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#context-in-the-window-is-not-attention-in-focus cairn https://chapsoft.com/cairns We surveyed 47 public analyses of the Claude Code leak. Every one asked 'What's inside?' We asked a different question. We surveyed 47 public analyses of the Claude Code leak. Every single one asked the same question: "What's inside?" We asked a different question: "What does Claude Code tell us about building better tools for agents?"

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#what-47-analyses-of-the-claude-code-leak-missed cairn https://chapsoft.com/cairns A regex fires every time, costs zero tokens, has zero hallucination risk. Use deterministic methods for what can be determined deterministically. The pundits laughed at Anthropic's frustration regex — a simple pattern matcher inside the most sophisticated AI system in the world. But a regex fires every time, costs zero tokens, has zero hallucination risk, and is trivially debuggable. The principle: use deterministic methods for what can be determined deterministically. Save inference for what requires judgment.

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#dont-infer-what-you-can-detect cairn https://chapsoft.com/cairns MCP server tool responses land inside someone else's context window. Cap responses. Include truncated flag and total_count. If you're building an MCP server, your tool responses land inside someone else's context window. Unbounded results waste tokens the agent needs for reasoning. Cap responses. Include a truncated: true flag and a total_count. Budget your output like a conference talk — say what matters, leave room for questions.

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#response-budgets cairn https://chapsoft.com/cairns The cost of an unused field is zero. The cost of retrofitting a field into a running system is a migration. We added an agency field to log events — default "none," purely informational. Today it's metadata nobody reads. When autonomous agents arrive, it's the audit trail that tells you which actions were human-directed and which were agent-initiated. The cost of an unused field is zero. The cost of retrofitting a field into a running system is a migration.

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#metadata-now-enforcement-later cairn https://chapsoft.com/cairns Self-correction catches tactical errors. Verification catches strategic errors. Better execution demands better oversight, not less. Next-gen AI models can identify and correct their own errors recursively. Self-correction catches tactical errors — typos, logic bugs, off-by-one mistakes. Verification catches strategic errors — wrong requirements, missing edge cases, building the wrong thing well. Better execution demands better oversight, not less.

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#self-correction-is-not-verification cairn https://chapsoft.com/cairns Writing prompts for a model that doesn't exist yet. You're writing letters to a future colleague and hoping they'll be kind. There's a strange feeling in writing prompts for a model that doesn't exist yet. You're designing a conversation with a future colleague — someone smarter than the current version, with capabilities you can only estimate. The plan: use the next model to improve the tools that serve it. It's an introspection loop — the system examining itself through a more capable lens. You're writing letters to a future colleague and hoping they'll be kind.

]]> Sat, 04 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#the-introspection-loop cairn https://chapsoft.com/cairns When setting up an agentic orchestration workflow to overcome the limitations of vibe coding, don't vibe code the orchestration layer. When setting up an agentic orchestration workflow to overcome the limitations of vibe coding, don't vibe code the orchestration layer.

]]> Fri, 03 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#dont-vibe-code-the-orchestration-layer cairn https://chapsoft.com/cairns Define a Person entity with @id on your homepage, reference it by @id on every other page. AI knowledge graphs use this to deduplicate. When you define a Person entity with @id: "https://example.com/#person" on your homepage, every other page can reference it by @id instead of re-declaring the full entity. AI knowledge graphs use this to deduplicate — one entity, many pages. The consistency matters more than the completeness of any single block.

]]> Fri, 03 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#schema-org-id-references-across-pages cairn https://chapsoft.com/cairns Naming a class .geo-quotable instead of .highlight-box tells developers what the element is FOR and makes the design system self-documenting. Naming a class .geo-quotable instead of .highlight-box does two things: it tells future developers what the element is FOR (AI citation extraction), and it makes the design system self-documenting. The class name is the spec.

]]> Fri, 03 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#css-class-names-as-documentation cairn https://chapsoft.com/cairns Source Serif 4 as a variable font is one 417KB WOFF2 file covering weights 200-900. The non-variable alternative would be 6+ files over 1MB. Source Serif 4 as a variable font is one 417KB WOFF2 file that covers weights 200-900. The non-variable alternative would be 6+ files totaling over 1MB. For Core Web Vitals (LCP < 2.5s), fewer HTTP requests with font-display: swap wins every time.

]]> Fri, 03 Apr 2026 00:00:00 -0700 info@chapsoft.com (Russ Kohn) Russ Kohn https://chapsoft.com/cairns#variable-fonts-one-file-every-weight cairn