Issue #1 · August 5, 2026
Issue #1: I shipped one tool, abandoned another, back on stage
This is the first issue, so here is what it is: once a month I write up what actually happened. What I found, what I tried, what I built, what I killed. A monthly statement of account on my work and on the things I am making, not a list of links I liked.
July was a month where a few things genuinely changed direction, and you can see it in the two posts below. They point opposite ways, and that is the point.
New on the blog
First I had to admit I was building a tool I was not using. That was hard to see and harder to write down, because Worky was not the thing I actually wanted to be working on. So I stopped: not because it failed, but because I had no intention of carrying it forward, and pretending otherwise would have been the incoherent move.
I Built a Tool I Don't Use
Worky took a weekend to build and the workshop it was for never happened. On what AI made cheap, what it did not, and letting a project stay an experiment.
Then something else held my attention properly. Argus sits exactly where I think the interesting work is right now: security and AI, and specifically the overlap between the two. It is pre-1.0 and there is a lot left to do, but as a first pre-release I am happy with it, and I get to spend the next months making it more reliable and more useful to people other than me.
Introducing Argus: The Code Security Colleague Your Team Doesn't Have
Most AI agents are a language model with a shell pointed at your repo. Argus is my bet on a different mix: real scanners as structured tools, a human in the loop, and a model reasoning across both.
Where to catch me
I also spent a full year off stage. That is over: three events said yes, and all three are about security and AI. The focus is deliberate. For a long time I talked about whatever I happened to be doing that year, and now there are two threads instead of ten, plus everything that happens where they meet. Giving the work a direction turned out to make it much easier to decide what to say yes to.
./GoLab 2026 / Securing Go for Production: Best Practices, Supply Chain, Testing
./reactjsday 2026 / AI Won't Design Your System. You Have To.
One thing I'm still unsure about
I do not have a take on this one, which is exactly why I am asking. How do you onboard a junior now that the AI is already in the room? I learned by walking into a codebase with nothing there to help me and figuring it out the hard way, and that route is basically gone. Maybe the order flips: first learn to read and understand the code that already exists, then write your own in small side projects. Honestly, I do not know.
If you are onboarding juniors right now, hit reply and tell me what you are doing. I am collecting answers before I write anything about it.
That is July. Later this month I am writing up the agent skills I actually use for security review, which is the practical follow-up to Argus.