<?xml version="1.0" encoding="utf-8"?><rss version="2.0"><channel><title>J. Michael Constans</title><link>https://constans.dev</link><description>Research-grounded systems architecture. Deterministic, audit-ready AI infrastructure for the enterprise.</description><item><title>Token Econ</title><link>https://constans.dev/writing/token-econ/</link><description><![CDATA[aka Model pricing]]></description><category>llm</category><category>economics</category><category>snark</category><guid>https://constans.dev/writing/token-econ/</guid><pubDate>Tue, 02 Jun 2026 23:33:30 -0700</pubDate></item><item><title>My current architecture methodology</title><link>https://constans.dev/writing/my-current-architecture-methodology/</link><description><![CDATA[<h2>Architecture? Er...</h2>
<p>Architecture is one of those words that means different things to
different people, and carries different emotional valence according to
the a person's history with people that called themselves architects,
or with groups called `architecture team'.  All of the definitions I've
heard carry truth, but the only one that struck me as universally
applicable was "architecture is what everybody contributing to the
project needs to know."</p>
]]></description><category>architecture</category><category>ai</category><category>agents</category><category>philosophy</category><guid>https://constans.dev/writing/my-current-architecture-methodology/</guid><pubDate>Fri, 29 May 2026 14:58:00 -0800</pubDate></item><item><title>Determinism under load</title><link>https://constans.dev/writing/determinism-under-load/</link><description><![CDATA[A working definition of audit-grade inference, and what it costs.]]></description><category>determinism</category><category>audit</category><category>inference</category><guid>https://constans.dev/writing/determinism-under-load/</guid><pubDate>Tue, 26 May 2026 09:00:00 -0500</pubDate></item><item><title>The Unreasonable Effectiveness of Category Theory in Preventing LLM Drift</title><link>https://constans.dev/writing/the-unreasonable-effectiveness-of-category-theory-in-preventing-llm-drift/</link><description><![CDATA[Can Category Theory tame the &ldquo;AI slop&rdquo; that coding agents produce? We pressure-test the thesis that a CT-derived skeleton lets an LLM iterate toward production code while catching intent drift at checkpoints.]]></description><category>ai</category><category>agents</category><category>philosophy</category><guid>https://constans.dev/writing/the-unreasonable-effectiveness-of-category-theory-in-preventing-llm-drift/</guid><pubDate>Mon, 25 May 2026 12:00:00 -0500</pubDate></item><item><title>That feeling when Google I/O catches up to you...</title><link>https://constans.dev/writing/google-io-and-t-shape/</link><description><![CDATA[Spooky coincidence, or just something in the air?]]></description><category>software</category><category>architecture</category><category>philosophy</category><guid>https://constans.dev/writing/google-io-and-t-shape/</guid><pubDate>Wed, 20 May 2026 20:57:00 -0700</pubDate></item><item><title>Cops or plumbers? The culture of your security team matters</title><link>https://constans.dev/writing/are-your-it-security-people-cops-or-plumbers/</link><description><![CDATA[<p>Now that LLMs are flipping the table on so much of what we knew about
IT best practices, and we are struggling to feel at all confident that
we're not exposing ourselves to new unknown risks when we add
inference engines to our pipelines, my mind came to think about the
many models adopted by organizations to deal with these issues. I've
found it to be a telling indicator of organizational culture.  Does
your security team see themselves as cops or plumbers?</p>
]]></description><category>industrial-organization</category><guid>https://constans.dev/writing/are-your-it-security-people-cops-or-plumbers/</guid><pubDate>Wed, 29 Apr 2026 23:24:00 -0800</pubDate></item><item><title>Morphology becomes ontology</title><link>https://constans.dev/writing/morphology-becomes-ontology/</link><description><![CDATA[To make LLMs useful we wrap them in a 1990s software firm: actor, critic, pair-programming. The morphology we impose ends up dictating the ontology we get — intelligence as social architecture, not raw horsepower.]]></description><category>ai</category><category>agents</category><category>philosophy</category><guid>https://constans.dev/writing/morphology-becomes-ontology/</guid><pubDate>Sat, 04 Apr 2026 12:00:00 -0500</pubDate></item><item><title>Categorical learners and governance for insurance contracts</title><link>https://constans.dev/writing/categorical-insurance/</link><description><![CDATA[We develop a category-theoretic treatment of insurance contracts in which policies, endorsements, and claim events are objects and morphisms in a structured category. Claim resolution is recast as composition of morphisms, yielding a deterministic and auditable semantics for contract execution. Worked examples are drawn from the `categorical-insurance` reference implementation, which exposes the constructions described here as a small embedded DSL. A category-theoretic model of insurance contracts; deterministic, auditable claim resolution via composable morphisms.]]></description><category>applied-category-theory</category><category>large-language-models</category><category>bayesian-optimization</category><category>pareto-optimization</category><category>monoidal-categories</category><category>haskell</category><category>category-theory</category><category>insurance</category><category>governance</category><guid>https://constans.dev/writing/categorical-insurance/</guid><pubDate>Sat, 04 Apr 2026 09:00:00 -0500</pubDate></item><item><title>A Category Theoretic Framework for Multi-Agent Workflows</title><link>https://constans.dev/writing/pi-agent-space/</link><description><![CDATA[Explore the multidimensional solution space for optimized agent construction by treating it as a combinatorial Bayesian Optimization problem and defining the categorical semantics of agent workflows.]]></description><category>applied-category-theory</category><category>large-language-models</category><category>bayesian-optimization</category><category>pareto-optimization</category><category>monoidal-categories</category><category>haskell</category><guid>https://constans.dev/writing/pi-agent-space/</guid><pubDate>Sun, 15 Mar 2026 09:00:00 -0500</pubDate></item><item><title>Decisions in a Box</title><link>https://constans.dev/writing/decisions-in-a-box/</link><description><![CDATA[As software engineers, what are we even doing anymore?]]></description><category>software</category><category>philosophy</category><guid>https://constans.dev/writing/decisions-in-a-box/</guid><pubDate>Sat, 28 Feb 2026 14:36:00 -0700</pubDate></item><item><title>What bets are you making with your attention?</title><link>https://constans.dev/writing/what-bets-are-you-making-with-your-attention/</link><description><![CDATA[Spend your time on refining and articulating your aesthetics]]></description><category>software</category><category>philosophy</category><guid>https://constans.dev/writing/what-bets-are-you-making-with-your-attention/</guid><pubDate>Sat, 28 Feb 2026 14:36:00 -0700</pubDate></item></channel></rss>
