#100 - Good Faith
The hundredth Flocode newsletter, a free Python Intermediate course for engineers, and an open newsletter archive.
This is the hundredth Flocode newsletter. I didn’t plan for that to mean anything when I started, but it does, so I want to use it to tell you where things stand, where they’re going, and to make a few announcements.
How This Started
Flocode began in mid-2023 with a simple thesis: professional engineers should know how to write code. Not to become software developers. To extend their reach. Python was (and remains) the obvious vehicle. It’s free, it’s open source, and even a couple of months of focused learning lets you glue together tools that previously required expensive specialty software or sub-consultants that can hammer your project budget.
I’d been pushing this internally, where I work, for a while. Data science tasks that used to sit outside our scope were suddenly accessible. Calculations or model runs that took hours could be scripted and automated. The capability was just sitting there, waiting for someone to pick it up. This is obvious today in 2026 because AI tools have permeated white collar work. But in 2023, these things were still taking shape.
So I started writing about it in the open. I didn’t have a content strategy. I had a conviction that this stuff mattered and a suspicion that other engineers were thinking the same thing but didn’t have anyone showing them how.
One hundred articles later, I still don’t have a content strategy but I do have a body of work that reflects this progress.
What Flocode Actually Is
I’ve been asked this question more times than I can count, and my answer keeps evolving. Here’s where I’ve landed.
Flocode is what happens when you try to navigate an impossible transition in the open. The engineering profession is in the middle of a shift that none of us fully understand yet. AI is reshaping how we work, how we communicate, and what’s expected of us. Everybody knows it. Everybody is using these tools in some form. And everybody can tell the difference between AI-assisted work that’s genuinely better and AI-generated slop that wastes everyone’s time.
The debate about whether AI belongs in engineering is over. It's here. The question I keep returning to is narrower and harder: how do we make sure it makes us better at the parts of this work that actually require judgment?
I’ve been doing Brazilian Jiu Jitsu for years, and one thing that training has drilled into me is that there are many ways to make a technique your own. A submission that works beautifully for one body type can be completely wrong for another. The details matter, and they’re personal. Computational tools work the same way. The right workflow depends on your role, your projects, your personality, the problems you’re solving. There is no universal prescription. There’s only the discipline of figuring out what works for you, understanding why it works, and adapting when things change.
That’s what Flocode is. It’s the R&D. It’s testing boundaries, sharing what works, being honest about what doesn’t, and trying to help other engineers build the judgment to navigate this for themselves.
Where My Thinking Is on AI in Engineering
I’m currently leading our global AI initiatives at Knight Piésold, and the work has forced me to think carefully about categories of knowledge that I’d previously taken for granted.
There’s individual knowledge: what does a specific engineer know, what’s their experience, what projects have they worked on, what’s their availability? That’s your resource base.
There’s project knowledge: inspection reports, drawings, image archives, correspondence, design decisions made years ago by people who’ve since moved on. Every project accumulates this, and every project organizes it slightly differently. The signal-to-noise ratio is a genuine problem. You can spend more energy trying to synthesize institutional data than the synthesis is worth.
Then there’s reusable reference knowledge: design codes like ACI 318-25, standards, specifications. These are high-value, high-frequency resources where having a precise retrieval system pays off immediately. Cross-checking a report conclusion against the relevant code provision is exactly the kind of task where AI tools earn their keep.
And finally, there’s the work itself. Meetings, emails, Teams messages, document formatting, report assembly. The daily activity that consumes an enormous amount of cognitive bandwidth on tasks that are, frankly, not the best use of an engineer’s mind. You’ve only got so much RAM. Every minute spent wrestling with table borders in Microsoft Word is a minute not spent on a problem that actually requires your judgment.
The tools we're building span retrieval-augmented generation for institutional knowledge, custom skills and automations for routine engineering tasks, and a mix of on-site hardware and cloud-based services matched to different objectives. The architecture is deliberately model-agnostic. The pace of development from Anthropic, Google, OpenAI, and the smaller model providers is relentless, and over-investing in any single provider is a risk we're not willing to take. Maintaining security across that landscape, in an environment that shifts every few weeks, is one of the harder problems we're solving. We're deep in R&D with working prototypes, and we'll soon expand to alpha testers within the organization. The overall objective is simple: we already have great problem solvers. We want to give them more tools to solve more problems in different ways.
Two Announcements
The Flocode Intermediate Python Course is live, and it’s free. Completely free, for everyone. Six modules covering developer foundations, the standard library, modularization, data science with pandas, visualization, and object-oriented programming. If you’re getting started with Python or looking to move beyond the basics, there’s enough material here to keep you working for a year or more. You can find it on GitHub, along with the Essentials course if you need the foundations first.
The entire Flocode archive is now open. Every article I’ve ever written is free to read. No paywall. No locked content. If you’ve been a paid subscriber, nothing changes for you except that you now have my deeper gratitude for supporting work that’s available to everyone.
This decision is based on a simple principle: financial barriers should not prevent access to knowledge. If Flocode is worth something to you and you’re in a position to support it, the paid subscription is $8 a month, and every ounce of that support is genuinely appreciated. I’ve got two young kids at home in Vancouver, and the people who’ve chosen to pay have my heartfelt thanks. But if you can’t, or you’re not sure yet, everything is here for you regardless.
This is a good-faith model. I’m betting that the work speaks for itself, and that the people who benefit from it will support it when they can.
Thank You
Because of Flocode, I’m getting to work on problems I find genuinely exciting. The newsletter opened doors to the AI leadership role at Knight Piésold. It’s connected me with engineers across disciplines and geographies who are thinking about the same challenges. People reach out regularly, and the conversations are consistently thoughtful. It’s a community I’m proud to be part of.
To everyone who’s been reading since the early days, and to those who found this last week: thank you. I’m going to keep sharing what I learn. I’m going to keep being as honest as I can about what works, what doesn’t, and where the uncertainty sits. That’s the only version of this I know how to do.
This is the first big milestone. There will be more.
Be good out there.
James 🌊
Congrats James on reaching this milestone! Looking forward to the next hundred 🚀
Amazing work Jimmy!