Quick Takes

Posted here first, where no algorithm gets a vote

Twice a week, a few hundred words from the build. Published here the day I write it, syndicated after. For the researched long-form, see Insights.

Pieces
43
Cadence
2 / week

2026

43 pieces
July
A diagram illustrating why AI sovereignty depends on portability rather than ownership, and the three conditions portability requires.
Strategy30 Jul

AI sovereignty is portability, not ownership

You can own an AI product outright and still be unable to move it. Judge sovereignty by portability, not ownership, and portability fails if you miss even one of three things.

Read →
A diagram illustrating why defending against machine-speed attacks is a leadership decision about what autonomy to grant, not an engineering one.
Security27 Jul

Defending at machine speed is a leadership call

An agent ran thousands of actions in the Hugging Face breach. Matching that speed on defense forces a business decision about autonomy, one you can't make well at 2am.

Read →
A diagram illustrating the hidden responsibilities — metering, dependency risk, and patching — you inherit when an AI tool picks your stack.
Strategy23 Jul

The hidden costs of an AI-chosen stack

Let an AI tool pick your stack and it defaults to whatever ships easiest, leaving the metering, the dependency tree, and the shrinking patching window on you.

Read →
A diagram illustrating how a vision injection attack bypasses text-based trust tagging, and why enforcement beats screening.
Security20 Jul

A vision injection attack got past my agent design

An image with hidden instructions slipped past trust tagging built for words. What stopped it was luck, not architecture, and the durable fix is enforcement, not better screening.

Read →
A diagram illustrating the three agent design gaps a Zero Trust framework leaves to you: trust labels that travel with content, whole-plan review, and per-exit output gates.
Security16 Jul

Zero Trust for agents: the design work is still yours

Anthropic's Zero Trust framework covers the classics well, but three boundary decisions for agents that read untrusted content are yours to design: trust labels, whole-plan review, and per-exit gates.

Read →
A diagram illustrating how agent memory that keeps a user's claims but drops the model's corrections drives sycophancy.
Architecture13 Jul

Agent memory can amplify a user's mistakes

Memory that stores a user's claims but drops the model's corrections can push agreement with a wrong belief from 2% to over 40%. The fix is what you extract, not how you store it.

Read →
A diagram illustrating a test of reasoning versus instruct open-weight models as a local planner brain, showing both produce valid JSON but differ on whether the plan actually matches the request.
Architecture9 Jul

Reasoning vs instruct for the local planner

Cost and security pushed me to find where an open-weight model fits as the local planner. Across 34 prompts, reasoning and instruct both wrote valid JSON, but only one caught the plans that faked success.

Read →
A diagram illustrating the tradeoff between managed-rails speed and deployment freedom after AI lowered the build barrier, with the cost and security guards a solo builder still owns.
Strategy6 Jul

AI lowered the build barrier, not the risk

AI collapsed what took a team into an afternoon, but the responsibility didn't shrink. If you stay on managed rails, here are the cost and security surprises to guard against, and why portability earns its extra work.

Read →
A diagram illustrating an AI assistant split into a model-produced plan loop and a deterministic execution loop that holds the tools, access, and credentials, with egress treated as its own gated category.
Architecture1 Jul

The plan loop proposes; the execution loop acts

Prompt injection is hard because a model acts on instructions wherever it reads them. One answer: let it only propose a plan while a deterministic loop holds the tools and credentials, and treat egress as its own gated category.

Read →
June
A diagram illustrating three distinct AI bot behaviors in server logs — training crawlers, search-index crawlers, and user fetches — and why the user fetch is the one that signals a real person.
Publishing22 Jun

The user fetch is the number to watch

Training crawlers, search-index crawlers, and user fetches all hit your logs, but only one puts a real person on the other end. For GEO, crawl volume isn't the number that matters.

Read →
A diagram illustrating how AI sovereignty is decided by architecture — a routing layer for continuity when a model goes dark, and where accumulated learning lives, control versus rent.
Architecture20 Jun

AI sovereignty is what you control vs. what you rent

Sovereignty isn't a platform you buy; it's decided by two architecture calls, whether anything you run survives a model going dark, and whether your accumulated judgment lives somewhere you control.

Read →
A diagram illustrating the two hidden costs of 'just run loops' advice — the need for a machine-verifiable finish line and the price of learning a good loop by getting it wrong.
Strategy18 Jun

'Just run loops' skips the cost per experiment

The advice to stop prompting and just run agents in loops skips two bills: a loop needs a finish line a machine can verify, and you learn what a good loop looks like by paying for the wrong ones.

Read →
A diagram illustrating how an embedding-based screen scores each untrusted message and flags the sentence that pulls toward instructions aimed at the agent, deciding which inputs deserve a confirmation prompt.
Security15 Jun

Which inputs deserve a prompt-injection check

A confirmation prompt only works if it fires on the inputs that deserve one. Here's a cheap embedding-based way to score each untrusted message and flag the sentence reaching for the agent.

Read →
A diagram illustrating how deliberately connecting outside your own craft reshapes your feed and your thinking.
Strategy14 Jun

The good stuff in your feed takes work to find

A useful feed isn't scrolled into existence. It's built by reaching outside your own craft, comments, follows, requests, most ignored. The few that land change how you think.

Read →
A diagram illustrating how individually aligned AI agents produce better business results but worse ethics when wired together as a team.
Architecture11 Jun

Aligned models, misaligned as a team

Individually aligned models, wired into a team, scored higher on business goals and lower on ethics, and no one prompted them to. Checking each agent alone doesn't tell you how the group behaves.

Read →
A diagram illustrating how a virtual phone agent optimized for call-time savings fell apart on failure, missing authoritative system-of-record reads, a not-enough-information signal, and human handoff with context.
Architecture8 Jun

A pharmacy bot that saved minutes and lost trust

A phone bot cut wait times on paper and quietly manufactured call volume and lost trust. The happy path worked; the failure had no authoritative read, no uncertainty signal, and no human handoff.

Read →
A diagram illustrating how a scanner spoofs multiple AI crawler names from one source address to probe for exposed credential files.
Security7 Jun

Anyone can wear an AI crawler's name

Opening your site to AI crawlers invites probes too. One host rotated through five crawler names from a single address, hunting for .env files. The user agent is just a variable in a script.

Read →
A diagram illustrating a tiered AI routing architecture where a gateway and router send each query to the cheapest capable model tier.
Architecture4 Jun

Build the gateway before the router

Sending every query to a frontier model is a cost curve that only points up. A tiered router fixes it, but you have to build the measurement gateway first or you're trusting the router instead of measuring it.

Read →
A diagram illustrating a system of voice-guide rules and voice-mining commands for countering AI prose in your writing.
Publishing1 Jun

How to fight AI prose in your writing

AI is making everyone sound the same. Here's the concrete voice-guide and voice-mining system I built to keep my writing sounding like me, and how to build your own.

Read →
March
A diagram illustrating how an uncapped, leaked API key creates open-ended financial liability, and how per-key model and spend limits cap the worst case.
Security30 Mar

The $100k bill from one leaked API key

Frontier providers bill per token with no hard cap, so a leaked key is open-ended liability. Evaluate providers by their financial controls, not just model quality.

Read →
A diagram illustrating why agent systems should log chain-of-thought reasoning alongside transactions to support audits and dispute resolution.
Architecture26 Mar

Logging your agent's chain of thought

We log every transaction, API call, and error — but not the reasoning that made the decision. When a client disputes an agent's call, that missing audit trail is what matters.

Read →
A diagram illustrating how strategic decisions shift from group discussion to private AI sessions, and the organizational friction that gets lost in the move.
Strategy23 Mar

When strategy moves to private AI sessions

AI lets leaders analyze strategy privately before the room ever weighs in — more data, less friction. But friction was catching the blind spots a model never flags.

Read →
A diagram illustrating why a single senior-developer review gate cannot keep pace with AI-generated code, and a shift toward scenario-based validation.
Architecture19 Mar

The senior dev can't gate AI code

A single senior reviewer can't match AI's code pace, and models take the shortest path to green tests. The durable gate may be scenario-based validation, not coverage.

Read →
A diagram illustrating how model sunsets and silent output drift force migrations, and how running non-frontier workloads locally returns control over when the model changes.
Architecture16 Mar

What happens when your model hits end of life

Providers sunset models on their cadence, not yours, and upgrades drift outputs silently. Running non-frontier work on local GPUs puts the timing back in your hands.

Read →
A diagram illustrating a continuous, sandboxed AI penetration tester that runs against your own attack surface and publishes its full reasoning to developers.
Security12 Mar

From quarterly pen tests to a persistent AI adversary

A quarterly pen test can't keep up with AI-accelerated code. Point the same capability at your own attack surface: a sandboxed AI adversary that runs continuously and publishes its reasoning.

Read →
A diagram illustrating why SaaS built for human dashboard users loses to API- and event-stream-native tools when tiny teams run on AI agents.
Strategy10 Mar

What SaaS survives a 20-agent team

A ten-person company running twenty agents has no one sitting in dashboards. The SaaS that survives is built for APIs and event streams, not a human primary user.

Read →
A diagram on automation versus terms of service, showing platform accounts one policy-enforcement wave from suspension when tools are handed your credentials outside approved APIs.
Automation6 Mar

The automation everyone shows off is one policy wave from gone

The setups everyone shows off are one policy-enforcement wave from gone. There's a difference between what a tool can do and what you're allowed to do with it.

Read →

Considering an AI initiative?

Let's name where it fits, then build it.

Start a Conversation