> _
Hey. I'm Bo.
I'm trying to answer one question: how do we make AI as reliable as infrastructure?
I spent twenty years training for this without knowing it.
It started with Duke Nukem and Zelda, then WoW addons where I spent hours tweaking config files and installing mods to make the game work exactly how I wanted. That was my first taste of bending computers to my will, and I was hooked.
WoW raids until 3am, Dark Souls bosses where you die a hundred times and keep going anyway, leading a guild of twenty people through encounters where one mistake wipes the group. Early reps in systems thinking, pattern recognition, and staying calm under pressure.
Then I found vibe coding with Claude Code, and I haven't touched a game since.
It turns out to be the same flow state and the same "one more run" energy, but now the output builds something real.
// the_day_job.sh
My work focuses on making AI reliable in environments where failure carries real consequences. During the Ukraine-Russia war, I helped install NATO's SECRET network at the Pentagon. Today, I manage AI infrastructure for the intelligence community—over 100 GPUs powering 50+ production applications that support active operations worldwide.
Eleven years ago, I started with TS/SCI clearance and a polygraph, pulling cables in data centers. Now I'm deploying frontier AI models on classified networks, having worked directly with Google engineers to bring Gemini and Gemma into environments they couldn't access themselves.
That's where reliability transformed from an abstract metric into something I feel viscerally. Infrastructure teaches us about observability, deterministic deployments, and graceful failure handling. AI systems desperately need these same patterns.
> I didn't stop competing. I found a marathon worth running.
// the_construct.init
After hours, this is where I live. I'm apprenticing myself like a master craftsman, honing skills through deliberate practice and answering the questions in the codex.
Neo downloads kung fu in seconds, and I'm not that fast, but vibe coding compressed years into months. The builds are proof the skills loaded, and each session compounds into the next. The green text isn't decoration; it's a reminder that the rules are different here.
Things I'm Building
If you're working on AI agent reliability, or thinking about how to make these systems work in production, I'd like to hear from you.