AttackPathMan
Back to the library

128 commits, five characters and a Head of IT. What could possibly go wrong?

Dear adventurer,

Last Sunday, I tried to answer a simple question: Why this? Why now? Just... why?

Five days later, I've got a castle, five characters, a soundtrack, a movie storyboard and 128 commits since this adventure began.

And a growing suspicion that I've seriously underestimated what I've got myself into.

First, I promised you a bard.

An old friend, Deejay Nic, joined the adventure and together we created Glitch In The Matrix, the official AttackPathMan soundtrack.

Then I spent an unreasonable amount of time making a teaser fit exactly 117.96 seconds of music.

Apparently, creativity now comes with decimal places.

You can hear the result at attackpathman.com.

Then things got a little out of character.

Or rather, entirely into character.

What started as a few ideas turned into five personalities, complete with concept art, cartoon sprites and RPG cards.

Some look heroic. Others look suspiciously like people who would schedule a meeting to discuss why the castle is on fire.

Storyboard, The first meeting that closed anything: a cinematic close-up test, labelled not canon, in four pixel-art panels. 1: Mr Crumble, mid-croissant: “As I flagged, we have the incident entirely under control.” Crunch. 2: Amra: “Then why is the access still active?” 3: Mr G fills a lit doorway: “Close the confirmed route.” Crumble: “Could we close the meeting instead?” 4: AttackPathMan’s glowing eyes: “Oh, splendid. IT finally closed something. My path.” (opens the full-size picture)

I even had four AI sessions working in parallel to bring them to life.

Which brings me to another discovery.

Managing AI agents is remarkably similar to managing people.

They misunderstand instructions, occasionally invent things, and can spend an impressive amount of time agreeing with each other.

Unlike people, they can be wrong at industrial speed.

We even discovered that one of our automated tests was confidently reporting successful downloads when the downloads hadn't actually completed.

The AI had tested the test. The test had passed.

The file, unfortunately, hadn't.

An independent review eventually caught it.

Lesson learned: four confident AI agents don't necessarily make one correct answer. Sometimes you've just automated a committee.

And somewhere along the way, I changed my mind.

The original direction was too constrained by reality. I wanted room for creativity without inventing facts about real people or organisations.

So I rewrote the brief, reconsidered the story and gave the fictional world permission to be fictional.

Quite liberating, actually.

Which brings us to today.

A week of writing, designing, experimenting, reviewing and occasionally questioning my life choices.

There's something wonderfully ironic about creating a cybersecurity adventure and discovering that the biggest attack path so far runs straight through my own assumptions.

Still, I'm enjoying every minute of it.

The castle has inhabitants. The bard has delivered. The AI committee is under supervision.

And as for what awaits behind the castle gates?

Well, where's the fun in giving away the adventure before you've even started?

Until the next adventure,

Kay

P.S. The castle now has a Head of IT. Nobody knows what he does yet, but I'm told he's very senior. So that's reassuring.

Back to the library

Join the newsletter to hear when the next article is published.

Join the newsletter