A digital third place for thinking out loud about AI, data, and the infrastructure underneath.https://hey-mhk.com/https://hey-mhk.com/blog/what-an-ontology-explorer-taught-me/https://hey-mhk.com/blog/what-an-ontology-explorer-taught-me/I set out to build a tool that explains ontologies. I ended up building one that's careful about how it explains them, and that was the harder, better project.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/the-95-dollar-education/https://hey-mhk.com/blog/the-95-dollar-education/Pre-writing 6,000 field notes with an LLM cost more than the plan said and taught more than expected, including which model quietly narrates its own prompt.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/org-chart-vs-kinship/https://hey-mhk.com/blog/org-chart-vs-kinship/Every ontology declares how it's organized. Its concepts have an actual shape that's often different. You can measure the gap, and CIM's is wide.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/teaching-a-graph-to-rank-itself/https://hey-mhk.com/blog/teaching-a-graph-to-rank-itself/A thousand-node ontology is a hairball until you ask it the right questions. Three graph algorithms and one design lesson turn it into a map.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/triplestore-app-with-no-server/https://hey-mhk.com/blog/triplestore-app-with-no-server/On paper Emic is a database application with four moving parts. In production it's a folder of static files. Here's how the backend got compiled away.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/why-the-most-impenetrable-ontology/https://hey-mhk.com/blog/why-the-most-impenetrable-ontology/If a comprehension tool can make DMTF's Common Information Model legible, everything easier comes for free. So that's where I started.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/going-emic/https://hey-mhk.com/blog/going-emic/The hardest part of cross-industry research is learning a new field fast. There are two ways in, and one of them is built from the domain's own words.Sun, 14 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/what-the-engine-room-taught-me/https://hey-mhk.com/blog/what-the-engine-room-taught-me/These systems are more comprehensible than marketing suggests, and more limited than hype implies. Both are true simultaneously.Thu, 11 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/ai-readiness-is-a-governance-question/https://hey-mhk.com/blog/ai-readiness-is-a-governance-question/Every AI conversation eventually becomes a data conversation. Every data conversation eventually becomes a governance conversation. The sequence matters.Wed, 10 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/working-with-the-machine/https://hey-mhk.com/blog/working-with-the-machine/The most valuable prompting skill isn't writing better prompts. It's knowing when prompting isn't the answer.Tue, 09 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/hallucination-is-a-feature-not-a-bug/https://hey-mhk.com/blog/hallucination-is-a-feature-not-a-bug/Vendors frame hallucination as a bug they're fixing. It's actually inherent to how these systems work, which changes the governance question entirely.Mon, 08 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/ui-as-the-ultimate-guardrail/https://hey-mhk.com/blog/ui-as-the-ultimate-guardrail/AI systems speak with uniform confidence whether they're right or wrong. Interface design determines whether that difference becomes visible.Sun, 07 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/why-vectors-arent-enough/https://hey-mhk.com/blog/why-vectors-arent-enough/Vector search finds what's similar. Expertise often requires connecting things that aren't similar at all.Sat, 06 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/the-tacit-knowledge-bottleneck/https://hey-mhk.com/blog/the-tacit-knowledge-bottleneck/The most valuable knowledge in any organization is often the hardest to capture, because the people who have it can't fully explain it.Fri, 05 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/the-multiplier-effect/https://hey-mhk.com/blog/the-multiplier-effect/Proprietary data isn't the AI advantage most organizations think it is. What that data connects to is.Thu, 04 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/context-windows-and-memory/https://hey-mhk.com/blog/context-windows-and-memory/The most common misconception I encounter: people assume these models remember previous conversations. They don't.Wed, 03 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/the-efficiency-illusion/https://hey-mhk.com/blog/the-efficiency-illusion/That benchmark showing the smaller model matches GPT-4? Probably accurate. Also hiding tradeoffs that only surface in production.Tue, 02 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/training-economics/https://hey-mhk.com/blog/training-economics/"Let's just train a model on our data" sounds straightforward. The actual cost structure tells a different story.Mon, 01 Jun 2026 00:00:00 GMThttps://hey-mhk.com/blog/the-attention-mechanism-explained/https://hey-mhk.com/blog/the-attention-mechanism-explained/"Attention" is a brilliant name for what these models do, and also misleading in ways that matter.Sun, 31 May 2026 00:00:00 GMThttps://hey-mhk.com/blog/ive-heard-this-engine-before/https://hey-mhk.com/blog/ive-heard-this-engine-before/I worked on attention mechanisms before ChatGPT made them famous. That perspective has shaped how I separate AI signal from noise.Sat, 30 May 2026 00:00:00 GMT