Notes from the middle of real work.

Why 1 Year of Agentic AI Production Is Already the Bar

One year of agentic AI in production is now a hiring filter, and most people who qualify learned by doing while organizations expected full delivery output and no room to experiment.

Forward Deployed Engineering Is Not Product Engineering on Location

FDE ships in weeks what product teams ship in quarters, but the difference is not speed. It is scope, ownership, and what happens after the work is done.

Forward Deployed Engineering Existed Before the Title Did

Forward Deployed Engineering is a new label for work that's existed for decades. The best technology leaders embedded with operators, stayed through production, and moved between strategy and implementation. AI increased the leverage; the model stayed the same.

The Token Trap

We stopped measuring engineers by lines of code, then built leaderboards around token usage. The metric changed. The incentive problem didn't. The engineers who will matter are still asking whether the code should exist at all.

AI Can't Staff a Project

Skills and schedules live in systems, but the factors that actually decide who should staff a consulting project—client fit, continuity, trust—usually do not. AI inherits fragmented data; people still route hard staffing calls through the expert who holds the full picture.

AI Infrastructure Is the New Bottleneck

AI progress is no longer constrained only by models. Data centers, permits, power, components, and chip supply now set the pace of deployment.

Broad Models, Narrow Problems

General LLMs and most business workflows are a poor fit. Narrow the problem before you call the model—RAG, fine-tuning, or smaller models—instead of paying in tokens and retries.

You Can Ship Software Without Understanding It

AI-assisted delivery can make a product look finished before the team understands how it fails. Speed helps, but a working demo is not the same as a trustworthy system.

From Punch Cards to AI-Assisted Coding

AI-assisted coding is part of a long arc: tools reduce friction so teams can spend more energy on problem framing, solution design, and real-world outcomes.

A Roadmap Is Not a Plan — It's an Informed Guess With Dates

Roadmaps look precise, but many dates depend on decisions not yet made. Naming what's uncertain keeps them useful instead of fragile.

Agents Get the Attention. Workflows Get the Work Done.

The industry jumped from prompts to agents and skipped the conversation about AI workflows. Most problems need predictable steps, not open-ended autonomy.

The Constraint Didn't Go Away. It Moved.

AI can make coding feel unlimited, but every system still has a bottleneck. Speeding up the wrong step often piles more work in front of the real limit.

What Are We Paying For Now?

As AI compresses delivery time, clients are testing whether consulting fees still map to value. The firms that hold up will make judgment unmistakable, not effort defensible.

AI Adoption Depends on the CTO

The same technology plays out very differently as companies grow: from tool to capability to business lever.

Architecture Is an Organizational Problem

Team structure shapes how systems evolve more than any diagram or tool. If you want to move faster, start with structure before technology.

Decisions That Don't Stick

Most teams are not slowed down by code. They are slowed by direction that keeps changing and decisions that get reopened.

AI Is Making It Easier to Execute, and Harder to Grow

When execution is always smooth, the stretch where judgment forms can quietly disappear. Be deliberate about when to accelerate and when to stay in the problem.

AI Didn’t Replace Stack Overflow

AI changes how engineers get unstuck, but it does not remove the real work: judging relevance, risk, and fit. Answers got cheaper; responsibility did not.

The Gap Between Estimated and Actual Effort Is Growing

For engineers, estimated time, actual time, and reported time do not always align. AI anecdotes can widen that gap and distort what the team expects.

We Get a Second Chance with AI

AI is moving faster than most policies and habits. That gives teams another chance to set standards before consequences arrive after the fact.

What Happens When the Context Walks Out the Door

A consulting relationship that depends on one person’s memory and trust is fragile. Strong partnerships distribute context so turnover does not reset the work.

When the Timesheet Is the Scoreboard

Pricing is not neutral. When time is the unit, hours become the scoreboard and the incentive system quietly tells teams what “good work” means.

Answers Got Faster. Responsibility Didn't.

AI makes answers cheap. It doesn't make them safe. The work shifts from retrieving information to judging relevance, risk, and fit.

Pricing Is an Incentive System

Billing models shape behavior. When hours are the scoreboard, decisions start to optimize for billable work, not the right work.

We Get Another Shot at This

AI is moving faster than most guardrails. That makes judgment the real advantage, at work and at home.

When Knowledge Leaves With the Person

If one person holds the context, the relationship is fragile. The structure should carry the work even when people change.

Beyond the Pilot: Why Day-to-Day Is the Real Test

Pilots are controlled. Real life isn't. The hard part is ownership, support, and how the work holds up once the spotlight moves on.

Built to Demo, Not to Hold Weight

AI-assisted builds can look nearly finished before the important intent is specified. The risk shows up under load, where missing rules and weak assumptions start to matter.

The Next Era of Tech Leadership Is Closer to the Work

Strategy without contact with the work breaks down. Leaders don't need to code full time, but they do need real context.

The Right Partner Changes Everything

The best partnerships reduce blind spots and speed up decisions. What I've learned building and delivering side by side.

Explaining Tech to My Mom (and My CEO)

If you can't explain it simply, you probably don't understand it yet. The best leaders make the complex clear without dumbing it down.

How to Spot a Good Tech Consultant

You can usually tell before you sign. Look for curiosity, plain language, and proof they can connect strategy to delivery.

AI Is Everywhere. But It's Not What You Think

AI can deliver real value in focused spots. But it's still messy, adoption is uneven, and the headlines are ahead of reality.

Where AI Helps and Where It Doesn't

Most AI pilots can demo something. Fewer change outcomes. The difference is a real problem, usable data, and clear guardrails.

A Different Way to Build

Part memoir, part field guide. What I learned about building big things without burning out in the process.

Lean Consulting Holds Up Better

Big teams can slow decisions. A small senior team stays closer to the work and moves faster, with fewer handoffs and less waste.