Approach

How I work

Every engagement is built around the same question: where is the cost of a unit of work about to drop by an order of magnitude, and how do we deploy that before the competition does. In satellite communications that unit was a byte in transit. In business systems it was an integration. Today it is knowledge work, and the answer is AI.

So the current approach is simple. We work together to get the most out of the latest innovation. I bring a set of agents and a way of running them where your team can see every step. Agents do the volume. Your people do the judgment. Neither works alone.

The work happens in the open, in shared channels, not in private chats with a model. Your team learns the tools by watching. The agents learn your company from the corrections. When I leave, the room keeps working.

What I bring into the room

Agents with tool access scoped to your systems, read-only until you say otherwise, every call logged where the team can see it
A persistent memory of the engagement so Tuesday's lessons are still true on Thursday
A shared channel per problem, where reviews happen in public and become instructions for next time
A handover plan from day one. The goal is that the workshop outlives the consultant

inspired by: Tobi Lütke, Learning on the Shop Floor

A closer look at my capabilities, the writing behind them, and where they are useful.

AI systems that actually ship

I build agentic systems, retrieval pipelines, and memory layers, and I write about what breaks when they meet production.

The work

built
F.R.E.D., the chat on this site
Retrieval over every post, streaming responses, hardened to stay on topic. Ask it something about me.
post
Building a Persistent Memory2026
An agent memory that learns what matters and when to forget.
post
Building the Self-Driving Entity2025
A blueprint for agents that operate, not just answer.
post
The MCP Reality Check: An Implementer's Perspective2025
What Model Context Protocol looks like after you ship it.
post
The AI Always Takes the Shortest Path. That's the Problem.2026
Field report on how models corrupt architecture at scale, and the countermeasures.
post
Design Patterns in Action: Building an Intelligent Agent Workflow2025
Classic design patterns assembled into a human-in-the-loop agent.
post
Stop Coding Logic, Start Prompting Models: A New Paradigm for Complex Applications2025

Working together: Scoping and building an AI feature that has to survive real users: retrieval, agents, tool calling, evaluation, and the guardrails around them.
For investors & founders: I have opinions about which AI claims hold up because I have shipped the failure modes. Useful for diligence on AI-native pitches.

more: ai (22) · agents (3) · automation (11)

Architecture and engineering velocity

Systems that deploy in minutes, are observable, and can be changed without fear. I have opinions on how, and a track record of writing them down.

The work

post
My Development Non-Negotiables: Shipping Fast, Seeing Clearly, and Letting AI Assist2025
The four rules behind shipping large volumes of code daily.
post
Splitting the Monolith: Critical Patterns for Microservices Migration2024
Strangler fig, parallel run, and the other patterns for a migration that finishes.
post
Time Travel for Your Data: Unlocking the Power of Event Sourcing2025
Event sourcing as a debugging and AI superpower.
post
Your Fear of Friday Deploys Is Killing Your Product2025
Why frequent shipping beats perfect architecture.
post
The 90% Mindset: Why Senior Developers Think Backwards2026
Building is 10% of the job. Design for the 90%.
post
The Parable of the Mislabeled Granary2025
Naming is a contract. A short story about what breaking it costs.
post
Clean Code Is the Best Documentation — Especially for AI Agents2025
built
This site
Next.js on AWS via SST, static content pipeline, vector index rebuilt from the posts. Small, but built the way I tell clients to build.

Working together: Untangling a system that has become slow to change: migration planning, deployment pipeline, observability, and the team practices that keep it fast.
For investors & founders: I can tell within a conversation whether an engineering team is set up to compound or to stall. Technical diligence and fractional CTO-style review.

more: architecture (9) · deployment (3) · best-practices (7)

ERP and business systems for midsized companies

My day job is the seam where business logic meets execution: ERP, integrations, and the data contracts underneath them.

The work

project
Adteco
The consulting practice. Technology strategy and implementation for small and midsized businesses.
post
Embracing APIs as the Language of the Business with Robust Source Control Integration2024
Treat the API as the language of the business, with source control to match.
post
Creating Changes That Are Embraced By Your Team2025
Most system fixes fail on people, not technology. How to make change hold.
post
The Art of Efficient Documentation: What to Document and What Not to Document2024
What to document, what to skip, and why documentation is a cost.
post
The Advantage of Open Source in the Age of AI for Enterprises2024
Why open source is the safer enterprise bet in the AI era.

Working together: ERP selection, cleanup, and integration work where the requirement is business outcome, not a feature list. I have been on both the vendor and the operator side.
For investors & founders: I understand how midsized companies actually buy and run software. Useful context for B2B and vertical SaaS theses.

more: api (2)

Company building and lean strategy

I have started a handful of companies. Some fizzled, one hit. I apply engineering discipline to the ones I build and advise.

The work

post
The Geek's Guide to Startup 'Code Smells'2025
Organizational smells, diagnosed the way an engineer reads code.
post
What Clayton Christensen Would Tell the SaaS Industry Right Now2026
The Innovator's Dilemma applied to SaaS when build costs fall 90%.
post
Ditch Perfection, Embrace Iteration: Why Launching Early and Often is Your Secret Weapon2024
Launch early. The argument, with the risks named.
post
AI Moonshots: Building Revolutionary Technology with Lean Strategy2025
Lean strategy for revolutionary bets, not just incremental ones.
talk
Deploying Technology the Smart Way
The satcom-era talk on total cost of technology and cost per byte. Same lens, new domain.

Working together: Operator-level advice for founders and owners: what to build, what to buy, when to slow down, and how to keep the team aligned with the outcome.
For investors & founders: I have carried a company from zero through the part where technology decisions either compound or kill you. I invest and advise with that scar tissue.

more: lean-strategy (2) · innovation (4)

Investing and decision-making

I split my time between engineering and investing. Each sharpens the other, and I write to keep myself honest.

The work

post
The Arrogance of Investing: Why Market Inefficiencies Still Exist in the Information Age2025
Why market inefficiency persists in the information age, and what it demands of you.
post
The Illusion of Effortless: Why We Dismiss the Secrets We're Given2025
Why we dismiss the simple secrets that actually work.
post
Tesla's Neural Network Revolution: How Full Self-Driving Replaced 300,000 Lines of Code with AI2025
Reading a technology inflection from the outside: 300,000 lines of code replaced by a model.
project
An investment fund
No website, no solicitation, by design. It absorbs a lot of my thinking time.

Working together: A capital-allocator view of technology spend. I will tell you what not to buy.
For investors & founders: Co-investor or technical partner on deals where the engineering claims need a second set of eyes.

Writing and explaining hard things

Forty-odd posts, a conference talk, and a habit of pairing every deck with a memo. Clear writing is how I find out whether I actually understand something.

The work

post
Beyond the Bullets: Why Your Presentation Deck Needs a Partner in Crime (a.k.a. The Detailed Memo)—Now With Interactive Data!2025
Why a deck needs a memo, with interactive data to prove it.
post
Why Did LSTM Break Up with BERT? (A Gentle Intro to Neural Network Memory)2025
Neural network memory explained as a breakup. Proof I can make a hard topic land.
post
Engineering Signal from Noise: The Art of Digital Minimalism for Mental Clarity2024
Digital minimalism for engineers who need to think.
post
5 Secrets to Writing Content People Actually Want to Read2025
built
The whole blog
Writing to think, in public, since 2024.

Working together: Board-ready memos, architecture decision records, and the kind of documentation a team will actually read.
For investors & founders: You can audit how I think. It is all here.

more: documentation (4) · productivity (13)

Tools & technologies

The tools and disciplines I work with.

Languages: TypeScript, Python, SQL
Web: Next.js, React, Tailwind, MDX
AI: LLM agents and tool calling, retrieval pipelines, vector search, MCP, evaluation
Infra: AWS, SST, serverless, CI/CD, observability
Data: Event sourcing, API-first design, ERP integration
Earlier life: Satellite communications, network economics, cost per byte

Currently exploring: Where agent memory and evaluation stop being research and start being infrastructure.

Get in touch

ask F.R.E.D. /chat explore ideas from 39 published posts
email fred@fredpope.com
linkedin linkedin.com/in/fredpope