Use the machine.

The moral panic is too blunt to be useful.

There is lazy use of AI. There is also liberating use. Pretending these are the same act is intellectually unserious.

We already accept tools that alter expression: spellcheck, translators, editors, ghostwriters, accessibility software, dictation. AI sits on the same continuum, different in degree from those earlier tools and occasionally in kind. The ethical question is what role did AI play, and who is accountable for the result?

I'd rather be open about my answer than have it inferred.

Articulation, editing, research. None of those corrupt authorship.

Used as an articulation aid, converting fluid associative thought into linear prose, AI is translation. Used as an editor cleaning up rough drafts, it does the work editors have always done. Used as a research instrument with verified sources, it earns its place. In all three cases the originator remains human, and the human remains accountable for meaning.

AI becomes a problem when it manufactures the appearance of having thought without the thinking happening. When someone outsources judgment and accepts the output. When AI-generated claims get presented as firsthand expertise. When the user cannot explain or defend what was published under their name.

That is the hinge.

As a thinking partner. Originator, verifier, accountable party — all me.

What that looks like in practice:

• To draft and refine prose I've already thought through.
• To argue with my own positions before publishing them, finding the weak point I haven't yet considered.
• To compress dense reasoning into something a reader can actually finish.
• To translate between registers: internal memo into public position, operator vernacular into analyst prose.
• To pull together public information faster than I could alone, with every factual claim verified against an original source before it lands in client work.

Every analytic position on this site is one I can defend without the machine. Every framework is one I built and stand behind. Every recommendation is mine to be wrong about.

Sources have names. Trust has limits.

When this work touches third-party data — funding rounds, vendor metrics, market intelligence, technographics, talent flows — most of it flows in through MCP servers from a small set of sources I can name.

Two of them are people I know professionally and have followed for years. Ross Haleliuk publishes Venture in Security and writes the most honest analysis of the cybersecurity business I've found in long form. Mike Privette publishes Return on Security, which carries the cleanest cyber funding and M&A data I've encountered. I trust both of them as people who do real work — meaning their judgment, their methodology, their editorial standards. That trust does not extend to guaranteeing any particular number. No data set is perfect, and they would say so themselves.

Other inputs come from established commercial analytics providers, the kind every analyst desk in this space already uses.

The discipline is the same regardless of source: cross-reference, triangulate, never conclude from a single number. If a claim in client work depends on a data point, that point gets at least two independent sources before it survives. If two sources disagree and I can't reconcile them, that disagreement becomes part of the answer.

There are uses I won't go to.

• I won't cite sources the AI invented. Hallucinated citations are the single fastest way to lose trust, and every source goes through a check.
• I won't manufacture social proof, fake personalization, synthetic case studies, or fabricated client quotes.
• I won't deliver a client engagement where AI is doing the analytic work and I'm laundering it. Clients pay for judgment.
• I won't present an AI-generated position as one I arrived at independently.

If you ever wonder whether a specific claim came from me or from a tool, ask. I'll tell you.

For some of us, AI is not a cheat code. It is a bridge.

Some people can speak fluently in meetings but cannot write cleanly. Some can think in systems but struggle with linear prose. Some can see the whole architecture of an argument and still not get the first sentence out.

For those people, and I am one of them, AI is not a cheat code. It is a bridge.

The line is unearned authority.

Using AI to articulate your own mind is legitimate. Using AI to improve communication is legitimate. Using AI to explore, challenge, and structure thinking is legitimate. Using AI to fabricate competence, evidence, originality, or expertise is not.

The future belongs to neither the AI purists nor the AI maximalists. The purists mistake tools for corruption. The maximalists mistake output for judgment.

The harder position, and the one I try to hold: