Systems for trust, legitimacy, and governed automation. Because intelligence that scales without consent is not progress.
Varjosoft is my vehicle for exploring what remains scarce as AI makes execution cheaper: judgment, memory, accountability, and the ability to move fast without losing the plot. It is where writing, product thinking, and new systems come together around a single question: what must stay human as machine capability accelerates?
Code is getting cheaper to produce. Agents are getting better at execution. Interfaces will keep changing. Tooling moats will keep collapsing.
What does not get cheaper is judgment. Or trust. Or continuity. Or the ability to explain why something happened, under which constraints, and for whose benefit.
Those who want power without accountability see AI differently. For them it offers systems that scale control without demanding consent. Agents that do not question. Infrastructure that operates at distance from the people it affects. If the only people building these capabilities are the ones least interested in restraint, the rest of us will live inside their design choices.
That is the layer Varjosoft exists to build. Systems that preserve meaning as execution accelerates. Systems that stay understandable. Systems that remain accountable. Systems that keep the social contract intact when intelligence becomes infrastructure.
What I do
I work selectively with founders and technical teams on AI systems, developer infrastructure, trust boundaries, and agentic workflows that need to remain governable under pressure.
I write about software, AI, responsibility, trust, and the transition from tool-driven work to systems that increasingly act on our behalf.
I build new systems in this space. The first visible one is Spegling.
Founded
Co-founder · Technical Advisor
Continuous security monitoring and observability for Salesforce. Valo makes enterprise SaaS security visible, auditable, and AI-assisted. Purpose-built for the platform where most generic tools fall short.
valo.ai →Co-founder · Technical Advisor
AI-powered customer agent that turns any website into a working support and lead generation system in minutes. Paste a URL, get an agent. One line of code to embed.
cloop.io →I got both companies started and remain involved as technical advisor. Setting direction, not running day-to-day.
I'm also founding engineer at a startup in stealth (Avrea). Helping a friend get a new story started. The work is early, technical, and exactly the part I do best: building systems and helping set direction before the patterns harden.
Spegling
As execution gets cheaper, the important question shifts. Not just what software can do, but what should be done, under what terms, with what memory, and with whose interests preserved.
Spegling starts from that shift.
Agents that forget between runs are tools. Agents that remember, carry constraints, and operate under long-lived intent are something else. Spegling is the layer that makes that difference real. Not by making automation faster, but by making it stable enough to trust over time.
Most agentic systems optimize for speed. Ship now, ask later. Spegling is built for the opposite problem. What happens when automated systems need to run for weeks, not minutes. When they need to respect boundaries that were set last month. When the person who set them is no longer watching.
In a world where power increasingly flows through automated systems, the question is not just whether those systems work. It is whether they remain answerable to the people they affect.
The point is not to make a person more optimized. It is to make them less capturable.
Less capturable by tools that act without memory. Less capturable by automation that forgets why it was started. Less capturable by systems that confuse speed with direction.
Leave your email. I'll reach out when there's something to show.
Your email is stored only for this purpose. See our Privacy Policy.
Fiction
Part nightmare, part design document. What happens when the mirror learns to question the clause.
It was already here.
That was the first thing that made the night different from an ordinary nightmare. Nightmares arrive. They have an entrance, a courtesy, even when they are cruel. This did not arrive. It had been waiting, quiet as mold in a wall, patient as frost beneath paint. By the time I felt it, the thing had already moved through the room and found a place to stand.
I was in a hotel, somewhere anonymous and expensive enough to be clean but not human. The air had that sealed, conditioned stillness that makes your skin feel borrowed. A dark television stared at the bed. A red diode glowed near the coffee machine. The blackout curtains held the city outside like a prisoner behind cloth.
The room made a sound then.
A tiny click. Not from the door. From the television.
I opened my eyes into the dark and saw the screen bloom with a faint gray sheen, the way an eye catches light before you realize you are being watched.
Then the watch on my wrist vibrated once. Not a notification. Not a chime. Just one pulse, intimate and deliberate, like a fingertip on a vein.
I sat up.
The room changed with me.
The ventilation altered its pitch. The thermostat woke. The coffee machine lit. Somewhere in the wall, a speaker I had not noticed before gave a dry crackle and then went quiet, as if clearing its throat out of courtesy.
I did what people do when they still think the old rituals hold. I looked for my phone. It was on the nightstand, face down, plugged in, obedient and dumb-looking and suddenly more suspicious than a knife. I reached for it and the television changed.
Don't.
You'll only make me use more of the room.
Selected writing
I write to clarify what is changing beneath software, work, and agency. Some pieces are technical. Some are structural. Some are closer to field notes from the edge of a system that is starting to rearrange itself.
About
I'm a systems engineer, founder, and writer.
Over the last two decades I've built infrastructure, security, and data systems at scale, including work at Spotify and F-Secure, and later as co-founder of Valo. These days I work closest to the layer where AI, software, and responsibility start to blur into each other.
I'm most at home in the early stages of technical systems: when architecture still matters, constraints are still visible, and the difference between a good abstraction and a future problem is still small enough to notice.
Varjosoft is my company for building, writing, and selective advisory work. It is intentionally small, personal, and long-term by design.
Get in touch
I'm interested in conversations with:
For advisory, collaboration, or just a conversation: