Writing

Ideas, written down.

Leadership, execution, technology, and applied AI — from someone who runs the play, not just the commentary.

Newsletter

Operator notes, straight to your inbox.

Occasional, no-noise notes on leadership, execution, and applied AI — from the field, not the sidelines.

Articles

15 Jul 2026 · 8 minen

The Fastest Way to Lose Quality Is to Treat Suppliers Like a Contract Problem

Supplier quality rarely collapses because the contract is missing a clause. It collapses because you never built a shared operating system for how work gets done, measured, escalated, and fixed together.

14 Jul 2026 · 7 minen

Digital Twins That Earn: Turn the Model Into a Control System in 90 Days

A digital twin that produces “insights” is a workshop artifact. A digital twin that changes weekly decisions becomes a control system, and that is where P&L shows up.

10 Jul 2026 · 8 minen

AI Leadership Is Workflow Leadership: Own the Handoffs

AI creates value when a real team owns the full workflow: inputs, approvals, exceptions, and failure handling. If nobody owns the handoffs, you do not have an AI system. You have a demo.

9 Jul 2026 · 7 minen

Pricing Is Product Management in Disguise

Pricing is not a spreadsheet exercise. It is product management with margin accountability, and if you do not run it with discipline, your sales team will do it ad hoc in the field.

8 Jul 2026 · 6 minen

The Operator’s Playbook for Making Grid-Scale Storage Bankable

Grid-scale storage doesn’t become bankable because the chemistry improved. It becomes bankable when you de-risk interconnection, price revenue like it can disappoint, and engineer availability with an industrial operating system.

7 Jul 2026 · 7 minen

AI Architecture Isn’t a Diagram. It’s an Operator’s Checklist.

Most AI programs stall because leaders treat architecture as a picture and success as a demo. Scale happens when you run AI like operations: contracts, evals, SLOs, controls, and incident response—owned like any other production system.

7 Jul 2026 · 6 minen

Connected Sensors Don’t Buy You Uptime. Operating Discipline Does.

Connected sensors are cheap compared to downtime, but most companies still do not get the uptime they expected. The gap is not the hardware, it is the operating system you build around the data: ownership, economics, and closed-loop execution.

2 Jul 2026 · 7 minen

AI in Manufacturing Doesn’t Fail on Models. It Fails on the Data Contract.

Most manufacturing AI pilots die quietly, not because the model was wrong, but because nobody owned the data as an operational contract. The fix is less about data lakes and more about CTQs, exception standards, and governance that is funded like uptime and yield.

29 Jun 2026 · 7 minen

Agentic Treasury Needs a Kill Switch: Design Autopilot Like a Plant

Agentic AI in treasury will only be safe once we treat it like a production line, not a chatbot. The winning pattern is simple: gated autonomy, deterministic replay, and a kill switch that actually works under stress.

24 Jun 2026 · 9 minen

MiCA 2.0 Will Be Won in Ops Liability, Not Issuance Rules

Regulation becomes real where money gets stuck: complaints, reversals, and cross-border liability. The winners in the next phase of regulated tokenized money will be the ones who own the messy middle and can prove it in operations, not PowerPoint.

16 Jun 2026 · 6 minen

Tokenized Equities: The Tech Is Easy. The Control Plane Is the Product.

Tokenized equities will not scale on “better rails” alone. The winning platforms will treat the control plane, market controls, reconciliation, and operational assurance as the primary product.

27 May 2026 · 7 minen

LLMs Belong in the Narrative Layer, Not in the Financial Truth Layer

Language models can accelerate financial reporting, but only if you keep them out of the financial truth layer. Treat them as narration engines constrained by an immutable evidence pack, and your reporting gets faster without becoming unauditable.

20 May 2026 · 7 minen

AI Adoption Is Mostly Workflow Ownership, Not Models

Most “AI adoption” programs fail because they treat AI like a demo, not like a process. The operator move is to define handoffs, error budgets, and escalation paths, then measure outcomes in cycle time, leakage, and complaint rate.

13 May 2026 · 7 minen

Agentic Ransomware Isn’t New Risk. It’s Faster Risk.

Agentic attacks compress time-to-impact. The response is not more optimism or bigger dashboards, it is hard control primitives you can audit, rehearse, and recover from.

29 Apr 2026 · 7 minen

Stop Measuring AI by Hours Saved

Hours saved is a comfort metric. If you want AI to move the P&L without creating new operational risk, measure cycle time, error rate, leakage, and margin per constrained headcount, then change the operating model to match.

15 Apr 2026 · 8 minen

MiCA Plus EMI Makes Stablecoin Payments Sellable, Not Winnable

Regulatory rails can turn stablecoin payments into something a real finance team can approve. But the winners will still be decided by distribution and the unglamorous machinery of disputes, guarantees, and reconciliation.

1 Apr 2026 · 8 minen

A Debit Network Is Not a Network. It Is the Checkout Control Plane.

In payments, the most valuable asset is not the rail. It is the control plane that decides where volume flows, what it costs, and who gets to build economics on top of the checkout moment.

25 Mar 2026 · 7 minen

Tokenization Is the Wrapper. 24/7 Stablecoin Settlement Is the Rewrite.

Tokenization gets the headlines, but the structural shift is 24/7 stablecoin settlement. Always-on settlement turns margin, defaults, and liquidations into software loops, and it moves brokerage economics from distribution to risk and controls.

18 Mar 2026 · 7 minen

AI Moderation Needs an Error Budget and a Kill Switch

AI moderation fails in the same way automation fails on a factory line: when you ship it without stop conditions, a measurable error budget, and fast recourse. Treat wrongful actions as production defects and build the control plane before you scale the model.

4 Mar 2026 · 7 minen

When AI Gets Cheap, Execution Becomes the Only Margin

As model costs compress, the only defensible margin left is operational execution. The winners will budget AI like an ops improvement program, price outcomes, and run AI with reliability discipline, not demo discipline.

25 Feb 2026 · 7 minen

An “HFT Layer-1” Is Not Faster Crypto. It Is a New Exchange Rulebook.

When a chain removes the public mempool and replaces it with streamed transactions and preconfirmation, it stops being a pure throughput story. It becomes market structure, with a new ordering policy, new risk gates, and new rent collectors unless governance makes allocation transparent and auditable.

18 Feb 2026 · 8 minen

AI Value Isn’t in the Org Chart. It’s in the Stop Button.

Most AI programs fail the same way: they add teams and tooling without changing who owns the decision, the metric, and the rollback. Real value shows up when operations owns a small set of decision rights tied to bottlenecks, with finance-grade definitions and a hard stop button.

4 Feb 2026 · 7 minen

Small Models Are Not a Compromise. They Are the Only Sane Way to Ship AI to the Edge.

Edge AI is not a model-size debate. It is an operating model: latency budgets, connectivity risk, device QA, secure updates, and service economics decide whether your deployment scales or dies in pilot purgatory.

21 Jan 2026 · 7 minen

eSIM Isn’t Connectivity. It’s the Trust Anchor Your IoT P&L Forgot to Budget For.

Most IoT businesses budget for radios and data plans, then act surprised when provisioning, ownership transfer, and credential rotation turn into field hacks. eSIM changes the operating model, but only if you run it like an operating system, not a component.

8 Jan 2026 · 8 minen

Ruggedization Isn’t a Checkbox. It’s a Reliability Contract You Pay For and Enforce.

Ruggedization is not a label you apply to a product. It is a costed reliability contract that starts with an explicit mission profile and ends with controlled suppliers, governed testing, and a learning loop from field returns.