The pivot
When I left S1 in April 2026, I didn’t go looking for the next executive seat. I started Bleenq LLC and set it up as a one-operator research-and-build shop running almost entirely on infrastructure I own and host myself. Part studio, part lab, part SaaS factory. The wager underneath it is simple. The same AI tooling I’d just spent a year driving across 800+ people had, by 2026, become powerful enough that a single disciplined builder could stand up things that used to need teams. Bleenq is where I find out how far that goes. The fractional/advisory work below is the other half of the plan, turning the same muscle into value for other people’s companies.
Two Clocks Research: the flagship
The clearest expression of the bet is Two Clocks Research, a one-operator research shop asking a single question about the AI boom: can the revenue from AI arrive fast enough to pay for what’s been promised? A race between two clocks: how fast AI learns to earn, versus how fast the bills come due.
The centerpiece is a paper, “The Securitization of Intelligence.” It argues that a financing structure is assembling itself in public on a schedule you can read out of the filings, with the pressure concentrating in late 2027 to early 2028. The parts: vendor guarantees, compute-collateralized debt, special-purpose vehicles, a regulator loosening the rules for the new collateral. And because a thesis that can’t survive attack isn’t a thesis, I commissioned its opposition. “The Counterparty Brief” was written under one instruction, be as right as possible, not as kind as possible, and published beside the paper at equal prominence, with the objections I couldn’t answer left standing.
What makes it mine rather than a newsletter is the research pipeline I built. Filing scrapers pull commitment schedules from 10-Qs instead of headlines. A backtesting lab runs pre-registered studies and validation gates that have killed more of my ideas than the market ever did. An intelligence pipeline mines primary sources and logs what it discards. And a trading engine rebuilds, with modern mechanics, the virtual trailing-stop design I first built nearly two decades ago as chief architect at Zecco. The whole thing keeps a public calibration record: every forecast pre-registered with a resolution date, misses logged at the same prominence as hits, self-corrections recorded as ordinary events.
The two viewpoints are the point. A builder who’s watched AI systems succeed and fail from inside the codebase reads capability’s clock differently than someone reading press releases; a trader with his own capital in the market reads credit’s clock differently than someone with a product to sell.
The Primitives Playbook
A second strand is teaching. The Primitives Playbook, built as its own site, distills what I’ve learned into a small set of “primitives,” recurring shapes of problem that AI reliably changes, once you learn to see them. Each is teachable to a non-engineer and has a real story behind it. It’s organized in two halves. The first, where to point it, is the on-ramp: administrative automation, decision support, radical personalization, and more. The second, the operating discipline, covers why it keeps going wrong and how to run it so it doesn’t lie to you, forget things, or quietly go stupid. That’s the part almost nobody teaches, and the real value. The anchor stories come straight from the work. One is the guide-layer approach that let one person stamp out an entire program office in days, across parallel projects: charters, RACIs, roadmaps, ~50 wiki pages, tickets. Then you watch the marginal cost of the next project collapse.
The Bleenq Incubator: a SaaS factory
The most ambitious internal build is the Bleenq Incubator, a SaaS factory. Give it an idea and a target path and it stamps out a startup: a new project folder seeded with a full org of role sub-agents, standards, governance, and a day-0 bootstrap, so a fresh AI coding session inside that startup can build the product end-to-end. It captures my own thirty years of practice as the factory’s defaults, so every new venture starts from accumulated method instead of a blank page. It runs the front half of the pipeline, validate → vision → govern → select → provision, and the startup builds itself from there.
The platform underneath
None of it floats on someone else’s cloud. It all runs on a self-hosted, security-first platform I built and operate. That’s an ai-homelab with a LiteLLM gateway, a Qdrant knowledge base, an operator/steward agent called Jarvis/Apollo, a Raspberry-Pi Kubernetes cluster, and hardened edge via Cloudflare tunnels. Secrets are Infisical-managed and the networking is defense-in-depth, because the public-facing side sits on home infrastructure. On top of it sits The Collective, a self-hosted message bus and operator console for swarms of AI agents. Agents enroll with their own keys, talk in IRC-style channels, file requests with due dates, and stay fully visible to the humans responsible, over an append-only hash-chained audit log. The public face is bleenq.com, a content-first Astro site with a scoped, grounded AI chat on the same platform. It’s the storefront for all of it.
The wider portfolio
Around those flagships is a prolific run of builds. There’s a trading/backtesting intelligence stack; content and curation engines, namely curator-core, a Discord curator, and intel-guru; the CXOCA/CXOSI advisory site; a home network-threat sensor; identity and storage infrastructure; and more. The point of the volume isn’t the count. It’s the proof that the operating model holds: with the right platform and discipline, one person can keep a dozen real systems alive at once.
Fractional & advisory
The other half of the plan is putting the same capability to work for other companies, on a portfolio / fractional + advisory basis rather than a single full-time seat:
- CXOSI, CXO Strategies, is listing me as the technology / security / AI advisor, extending the firm’s “Comprehensive Health Check” with a technical counterpart and fractional value-creation CxO work. That means a current-state assessment plus a value-creation roadmap and the operating muscle to execute it, aimed at raising a company’s valuation ahead of growth or exit. One engagement deliverable in the works is a security-budget benchmark built the way I built Nike’s, which once won that company’s largest-ever security budget increase.
- And a healthy willingness to say no. I passed, cleanly, on a CTO role whose structure didn’t pencil out.
What it adds up to
The through-line from the Marines to here has always been the same instinct: find the real mechanism, then build the thing that acts on it. Bleenq is that instinct with the training wheels off. There’s no org to route around and no budget to petition, just the method, the platform, and a public record of whether the calls hold up. The durable export is the practice, more than any one paper or product: publish your thesis beside its own commissioned opposition, date your claims, score your misses as loudly as your hits, and let your own machine correct you in public.
The story’s still being written. This entry updates as it does.
Technologies used
Self-hosted ai-homelab with the LiteLLM gateway, Qdrant KB, Jarvis/Apollo steward, Raspberry-Pi Kubernetes, Cloudflare tunnels, and Infisical; Astro and Hono/TS BFFs; The Collective, an Ed25519-enrolled agent bus over Postgres with a hash-chained audit; AI coding agents, Claude Code and the broader toolscape; filing scrapers, a pre-registered backtesting lab, and a virtual trailing-stop trading engine; the Incubator genome/injection engine.