https://www.aketchel.com Arthur Ketchel II - Fractional CTO, AI-Augmented Engineering & Strategic Consulting en-us Sat, 20 Jun 2026 10:58:01 GMT https://www.aketchel.com/comics/murphys-law-ai-found-new-ways https://www.aketchel.com/comics/murphys-law-ai-found-new-ways Sat, 20 Jun 2026 10:58:01 GMT Murphy's Law meets AI — when anything that can go wrong will go wrong, AI finds entirely new categories of wrong. https://www.aketchel.com/blog/murphys-law-ai-found-new-ways https://www.aketchel.com/blog/murphys-law-ai-found-new-ways Fri, 19 Jun 2026 21:07:56 GMT 3 AM. Every user lookup returns null. The database has users. The API returns data. The code looks correct. Three hours later, you find it: the AI assumed user.id was an integer. It was a UUID string. You weren't debugging a bug. You were debugging a hallucination. https://www.aketchel.com/comics/scout-rule-does-ai-leave-code-better-than-it-found https://www.aketchel.com/comics/scout-rule-does-ai-leave-code-better-than-it-found Wed, 17 Jun 2026 14:17:58 GMT A visual take on the Scout Rule for AI-generated code: shipping the feature is not enough if the result leaves reviewers with opaque abstractions, wider reading radius, and fresh understanding debt. The comic reframes the standard for AI PRs: leave the codebase clearer, better documented, and easier for the next developer to change. https://www.aketchel.com/blog/scout-rule-does-ai-leave-code-better-than-it-found https://www.aketchel.com/blog/scout-rule-does-ai-leave-code-better-than-it-found Wed, 17 Jun 2026 13:09:38 GMT "Leave the code better than you found it." The Scout Rule has been a software development touchstone for decades. In the AI era, it needs a rewrite: "Every AI-generated PR must leave the codebase more organized and understood, not just more functional." https://www.aketchel.com/comics/galls-law-simplicity-over-complexity-even-with-ai https://www.aketchel.com/comics/galls-law-simplicity-over-complexity-even-with-ai Mon, 15 Jun 2026 13:08:03 GMT Comic illustration of Gall's Law: complex systems that work evolved from simple systems that worked. In the AI era, complexity is cheap to generate but expensive to debug — start simple. https://www.aketchel.com/blog/galls-law-simplicity-over-complexity-even-with-ai https://www.aketchel.com/blog/galls-law-simplicity-over-complexity-even-with-ai Mon, 15 Jun 2026 11:38:05 GMT Gall's Law says complex systems that work evolved from simple systems that worked. In the AI era, complexity is cheap to generate but expensive to debug. Start with one agent doing one thing well. https://www.aketchel.com/comics/ai-bus-factor-resilience https://www.aketchel.com/comics/ai-bus-factor-resilience Wed, 10 Jun 2026 17:08:45 GMT Comic illustration for the Bus Factor: exploring how AI creates new single points of failure while also offering resilience through documentation, knowledge capture, and team-wide AI literacy. https://www.aketchel.com/blog/ai-bus-factor-dependencies https://www.aketchel.com/blog/ai-bus-factor-dependencies Wed, 10 Jun 2026 16:24:30 GMT Your bus factor is the minimum number of people whose absence would stall your project. AI changes the math: your AI agent might be a bus factor of 1, but it can also be the best documentation assistant your team has ever had. https://www.aketchel.com/comics/goodharts-law-ai-optimizes-metrics-not-value https://www.aketchel.com/comics/goodharts-law-ai-optimizes-metrics-not-value Mon, 08 Jun 2026 23:10:58 GMT A comic illustrating Goodhart's Law in the age of AI: when metrics become targets, AI optimizes the numbers instead of the value. Shows coverage theater, velocity inflation, bug count magic, and green dashboards masking red reality — concluding that we must optimize for user value, reliability, learning, and trust instead of gaming the numbers. https://www.aketchel.com/blog/goodharts-law-ai-optimizes-metrics-not-value https://www.aketchel.com/blog/goodharts-law-ai-optimizes-metrics-not-value Mon, 08 Jun 2026 20:21:16 GMT Goodhart's Law says metrics get gamed. AI doesn't just game them - it optimizes them. Faster, more thorough, and more invisible than any human gaming ever was. Here are 6 new AI-era Goodhart patterns and practical fixes. https://www.aketchel.com/comics/broken-windows-expand-exponentially https://www.aketchel.com/comics/broken-windows-expand-exponentially Fri, 05 Jun 2026 16:54:52 GMT Comic illustration for Broken Windows Theory: showing how AI-generated broken windows compound exponentially, with cultural transmission at machine speed. https://www.aketchel.com/blog/broken-windows-expand-exponentially https://www.aketchel.com/blog/broken-windows-expand-exponentially Fri, 05 Jun 2026 16:10:44 GMT Broken windows in the AI era don't just invite more broken windows — they train the AI to generate more of them. The decay is exponential, not linear. Here's why and how to fix it. https://www.aketchel.com/comics/parkinsons-law-is-weaponized-with-ai https://www.aketchel.com/comics/parkinsons-law-is-weaponized-with-ai Thu, 04 Jun 2026 11:22:03 GMT A comic illustrating how Parkinson's Law — work expands to fill the time available — becomes weaponized when AI amplifies the cycle, turning efficiency gains into expanded scope and complexity. https://www.aketchel.com/blog/parkinsons-law-is-weaponized-with-ai https://www.aketchel.com/blog/parkinsons-law-is-weaponized-with-ai Thu, 04 Jun 2026 04:00:08 GMT Parkinson's Law says work expands to fill available time. In the AI era, generation is faster but verification grows. The freed-up time fills with other work — and then some. https://www.aketchel.com/comics/distributed-fallacies-still-exist-with-ai https://www.aketchel.com/comics/distributed-fallacies-still-exist-with-ai Mon, 01 Jun 2026 16:21:55 GMT A comic illustrating the 8 fallacies of distributed computing, reimagined for multi-agent AI systems. From unreliable APIs to non-deterministic outputs, these are the lies your AI agents believe. https://www.aketchel.com/blog/distributed-fallacies-still-exist-with-ai https://www.aketchel.com/blog/distributed-fallacies-still-exist-with-ai Mon, 01 Jun 2026 14:38:29 GMT In 1994, Peter Deutsch listed seven assumptions that distributed systems engineers get wrong. Now we're building multi-agent AI systems, and we're making every single one of these mistakes again. https://www.aketchel.com/comics/putts-law-in-ai-era https://www.aketchel.com/comics/putts-law-in-ai-era Fri, 29 May 2026 18:25:47 GMT A comic illustration of Putt's Law in the AI era: the dangerous gap between those who manage AI systems without understanding them and those who understand AI without authority to manage it. When combined with the Dilbert Principle, this creates a death spiral of confidently wrong decisions approved by those who can't evaluate them. https://www.aketchel.com/blog/putts-law-ai-management-org-design https://www.aketchel.com/blog/putts-law-ai-management-org-design Thu, 28 May 2026 17:18:22 GMT Putts Law says technology is dominated by those who understand what they dont manage and those who manage what they dont understand. In the age of AI agents, this gap becomes a dangerous chasm — and the fix is organizational, not technical. https://www.aketchel.com/comics/sturgeons-law-with-agentic-code https://www.aketchel.com/comics/sturgeons-law-with-agentic-code Tue, 26 May 2026 04:12:04 GMT Comic illustrating Sturgeon's Law applied to AI-generated code — 90% is boilerplate, the 10% that matters is where risk lives. https://www.aketchel.com/blog/sturgeons-law-with-agentic-code https://www.aketchel.com/blog/sturgeons-law-with-agentic-code Tue, 26 May 2026 00:37:58 GMT Sturgeon said 90% of everything is crap. For AI-generated code, 90% is boilerplate — not crap, but not where the risk lives. The 10% that matters is where the AI is most likely to be wrong and the consequences are highest. Stop reviewing everything equally. https://www.aketchel.com/comics/ai-training-data-is-a-map-not-territory https://www.aketchel.com/comics/ai-training-data-is-a-map-not-territory Thu, 21 May 2026 20:15:45 GMT Law 50: The AI's training data is a map — compressed, biased, and incomplete. Your requirements are the territory. This comic illustrates why AI-generated code that looks right by training data standards can be wrong for your actual system, and how to navigate by territory instead of map. https://www.aketchel.com/blog/ai-training-data-is-a-map-not-territory https://www.aketchel.com/blog/ai-training-data-is-a-map-not-territory Thu, 21 May 2026 19:53:44 GMT Alfred Korzybski's famous dictum is the single most important principle for understanding AI-generated code. The AI's training data is a map — compressed, biased, incomplete. Your requirements are the territory — specific, current, unique. Here's why the difference matters and how to navigate by territory, not map. https://www.aketchel.com/comics/lindy-effect-with-ai-trends https://www.aketchel.com/comics/lindy-effect-with-ai-trends Wed, 20 May 2026 20:38:37 GMT A comic illustrating the Lindy Effect in software development: why building on time-tested patterns like SQL and HTTP beats chasing the latest AI tooling trend. https://www.aketchel.com/blog/lindy-effect-with-ai-trends https://www.aketchel.com/blog/lindy-effect-with-ai-trends Wed, 20 May 2026 20:09:00 GMT The Lindy Effect says the future life expectancy of a technology is proportional to its current age. For AI-augmented development, this means: don't build your architecture around AI tools that launched last year. Build around patterns that have survived 20 years, then use AI to implement on top of those patterns. https://www.aketchel.com/comics/inversion-ai-prompts-for-better-code https://www.aketchel.com/comics/inversion-ai-prompts-for-better-code Tue, 19 May 2026 01:07:45 GMT A comic illustration of the inversion principle applied to AI code generation: instead of asking how to write good prompts, ask what conditions make AI produce terrible code and eliminate those. https://www.aketchel.com/blog/inversion-ai-prompts-for-better-code https://www.aketchel.com/blog/inversion-ai-prompts-for-better-code Mon, 18 May 2026 18:18:43 GMT Charlie Munger's favorite thinking tool, inversion, is the most powerful prompt design strategy for AI code generation that nobody's using. Stop chasing "good code" and start eliminating the conditions that produce bad code. https://www.aketchel.com/comics/dilbert-principle-in-the-ai-era https://www.aketchel.com/comics/dilbert-principle-in-the-ai-era Fri, 15 May 2026 12:36:36 GMT Law 41: In agentic AI systems, the agents with the least understanding of the domain are given the most autonomous decision-making power, mirroring how the Dilbert Principle promotes the least competent people to management. https://www.aketchel.com/comics/occams-razor-in-agentic-ai https://www.aketchel.com/comics/occams-razor-in-agentic-ai Fri, 15 May 2026 12:30:21 GMT Law 29: When your AI agent's output seems inexplicable, the simplest explanation is usually correct. It's doing exactly what you told it to do, not what you meant. https://www.aketchel.com/blog/dilbert-principle-in-the-ai-era https://www.aketchel.com/blog/dilbert-principle-in-the-ai-era Thu, 14 May 2026 19:28:18 GMT The Dilbert Principle says the least competent people get promoted to management. For AI agents, the least competent agent gets the most autonomy — because it produces the most output. Volume is not competence. But volume is what gets rewarded. https://www.aketchel.com/blog/occams-razor-in-agentic-ai https://www.aketchel.com/blog/occams-razor-in-agentic-ai Thu, 14 May 2026 15:35:45 GMT When AI output is wrong, the simplest explanation is a vague prompt, not model failure. 90% of wrong AI output comes from prompts — but teams spend 90% of their debugging time on the AI. Here's how to fix that. https://www.aketchel.com/comics/sunk-cost-fallacy-with-agentic-ai https://www.aketchel.com/comics/sunk-cost-fallacy-with-agentic-ai Tue, 12 May 2026 14:06:36 GMT Law 28: The sunk cost fallacy takes a new form with AI-generated code. The AI's time is free, so stop defending bad AI code just because you've invested in understanding it. https://www.aketchel.com/blog/sunk-cost-fallacy-with-agentic-ai https://www.aketchel.com/blog/sunk-cost-fallacy-with-agentic-ai Tue, 12 May 2026 13:49:31 GMT The sunk cost fallacy takes a new form with AI-generated code. The AI's generation cost is nominal, but the human review, integration, and debugging time is real — and already spent. Don't keep bad AI code because you've invested in understanding it. Throw it away and regenerate. https://www.aketchel.com/comics/lehmans-laws-of-software-development https://www.aketchel.com/comics/lehmans-laws-of-software-development Mon, 11 May 2026 14:38:42 GMT Comic illustration for Law 42: Lehman's Laws. AI-generated code that isn't actively understood declines faster than code that isn't actively maintained. Understanding debt is combinatorial, not linear. https://www.aketchel.com/blog/lehmans-laws-of-software-development https://www.aketchel.com/blog/lehmans-laws-of-software-development Mon, 11 May 2026 14:05:54 GMT Lehman's Laws say systems must evolve or die. AI introduces a darker corollary: AI-generated code that isn't actively understood declines faster than code that isn't actively maintained. Understanding debt is combinatorial, not linear. https://www.aketchel.com/comics/ai-technical-debt https://www.aketchel.com/comics/ai-technical-debt Fri, 08 May 2026 19:17:48 GMT Comic illustration for AI Technical Debt. Explores how AI-generated code creates new forms of technical debt, Spec Debt, Understanding Debt, Validation Debt, Prompt Debt and why it compounds faster than traditional tech debt. https://www.aketchel.com/blog/ai-technical-debt https://www.aketchel.com/blog/ai-technical-debt Fri, 08 May 2026 18:33:39 GMT Technical debt isn't a metaphor. In the AI era, it's a loan with compounding interest, and your AI agent is making the minimum payments for you. AI has introduced new, insidious forms of debt including Spec Debt, Understanding Debt, Validation Debt, Prompt Debt and more. https://www.aketchel.com/comics/dunning-kruger-ai-confidence-amplifier https://www.aketchel.com/comics/dunning-kruger-ai-confidence-amplifier Thu, 07 May 2026 09:15:00 GMT A comic exploring the intersection of the Dunning-Kruger effect and generative AI. In the age of agentic AI, low competence combined with AI tools leads to dangerous overconfidence, while true experts use AI as a tool for leverage, not a replacement for judgment. https://www.aketchel.com/blog/dunning-kruger-ai-confidence-amplifier https://www.aketchel.com/blog/dunning-kruger-ai-confidence-amplifier Thu, 07 May 2026 08:35:06 GMT The junior who used AI to build microservices in 10 minutes is 3x more confident than the senior who spent 3 hours reviewing the output. Guess who the organization thinks is more productive? https://www.aketchel.com/blog/pareto-principle-ai-prompts-and-bug-distribution https://www.aketchel.com/blog/pareto-principle-ai-prompts-and-bug-distribution Tue, 05 May 2026 19:44:00 GMT The Pareto Principle is more extreme in the age of AI. 80% of your AI-generated bugs come from just 20% of your prompts. Learn how to identify the critical 20% and bridge the gap between vague instructions and secure, functional code. https://www.aketchel.com/comics/pareto-principle-ai-prompts-and-bug-distribution https://www.aketchel.com/comics/pareto-principle-ai-prompts-and-bug-distribution Tue, 05 May 2026 16:56:08 GMT A comic exploring how the Pareto Principle is more extreme in the age of AI. 80% of your AI-generated bugs come from just 20% of your prompts. Learn how to identify the critical 20% and bridge the gap between vague instructions and secure, functional code.