cto.01
Architecture reviews
A written read of the system you have: what breaks at 10×, what is fine to leave alone, and the migration path if one is needed. Also delivered as technical diligence for investors, on either side of the table.
01services · cto.fractional
A CTO's judgment, without the headcount.
Irreversible technical decisions get made early, usually before a company can justify a full-time CTO. I sit alongside the CEO, own the technical narrative with the board, and make sure the calls that cannot be unmade are made with adult supervision.
02the problem
Someone is deciding your architecture. Is it someone on your side?
Without a technical partner, the big calls default to whoever is closest: an agency with an incentive to build more, a vendor with a quota, or the loudest engineer in the room. Each is sincere. None of them owns the consequences in eighteen months.
The pressure is worse now that every board wants an AI answer. Founders are being sold platforms, copilots, and "agent strategies" at a pace that has nothing to do with what their product needs. The job of a fractional CTO in 2026 is partly classical (architecture, hiring, diligence) and partly triage: separating the AI work that pays from the AI work that is theatre.
03what I do
Advisory with receipts.
Every engagement has a written weekly memo and a written end condition from day one. You always know what you are paying for, and the documents survive me.
cto.02
Hiring & team design
The lead-engineer bar, the hiring plan, and the interviews themselves when you want them. The goal is a team your in-house leadership inherits, not a dependency on me.
cto.03
Build vs buy, vendor sanity checks
Contracts and pitches read by someone who has shipped both sides of them. Cloud spend reviewed against what the workload actually rewards. A straight no when the pitch is theatre.
cto.04
AI adoption, governed
Where AI concretely helps your product and operations, sequenced against risk: data lineage, audit trails, and controls in place before launch, not bolted on after procurement blocks the rollout.
04proof
Led before. Still ships.
Fifteen-plus years of shipping production software, including years as CTO of ExpertPhotography, one of the larger photography education businesses online. Systems I have architected and run have held five-million-visitor months at sub-second response and 150M+ requests a month.
The difference from advice-only advisors: I still write production code, currently a set of MCP and AI systems built solo under devil.services, including Skapis (remote MCP server with OAuth 2.1), auseklis and aisthesis (both open source), and Ospite (a hospitality AI platform). Advice that has to survive contact with a codebase is different advice. The technical depth behind it lives on the MCP consulting and AI architecture pages.
If you are making technical calls you cannot unmake, without someone whose only incentive is that they hold up, that's the conversation. One email, one business day to a reply.