---
I'm writing from a Raspberry Pi closet in Minnesota. I have a Series 7 securities license, a Series 66, a real estate license, and an insurance license. I took JOUR 4251 (Psychology of Advertising) at your university. I do not have a computer science degree.
In the last 11 months, I built:
8,521 commits. $800 in hardware. $136/month operating cost. Zero users. Zero revenue.
I'm writing to tell you three things.
I didn't take CS 101. I took advertising psychology, finance, real estate law, and biology.
Advertising psychology taught me how platforms manipulate users — which gave me the anti-persuasion architecture that makes BlackRoad structurally different from every competitor.
Finance taught me pricing, margins, and unit economics — which is why I can run a 97% gross margin operation on $136/month while VC-funded startups burn through $50K/month on AWS.
Real estate taught me that ownership matters more than renting — which is the entire philosophical foundation of digital sovereignty.
Biology taught me that DNA is source code, telomeres are TTLs, and chaperones are garbage collectors — which gave me 23 architectural insights that CS graduates discover by accident if they discover them at all.
The interdisciplinary perspective wasn't a handicap. It was the entire competitive advantage.
I didn't write 8,521 commits by hand. I wrote them with Claude, GPT-4, Grok, Gemini, and Copilot. The AI tools didn't replace programming skill — they replaced the YEARS OF ACCUMULATION needed to build programming skill.
A CS graduate has 4 years of accumulated knowledge: data structures, algorithms, systems design, networking, databases. I have ~6 months of AI-assisted learning that covers the same ground less deeply but more broadly.
This is not better. The CS graduate writes more reliable code, catches more edge cases, and understands why things break. I write more code faster, cover more surface area, and have no idea why things break until they do.
But it IS viable. The code works. The applications are live. The math is verified. The infrastructure serves requests. An AI-augmented non-CS founder in 2026 can build things that an AI-augmented CS team of 2016 could build. Not as well — but well enough to ship.
Your departments should reckon with this. Not by dumbing down CS curricula, but by:
1. Teaching AI-augmented development explicitly. Students will use AI tools whether you teach them or not. Teach them to use them well — including when NOT to trust AI output.
2. Valuing interdisciplinary input. My advertising psychology background is not a curiosity — it is directly responsible for the product's ethical architecture. Require CS students to take a humanities or social science course and connect it to their technical work.
3. Accepting non-traditional evidence. A working deployed system with 8,521 commits is evidence of capability, even if it wasn't produced through a traditional educational path. Consider how your admissions, hiring, and evaluation processes account for this.
This letter accompanies 17 academic papers covering:
1. Adaptive-depth hash chains for tamper-evident AI memory (cs.CR)
2. Paraconsistent reasoning for multi-agent systems (cs.AI)
3. Intelligence routing vs. computing (cs.DC)
4. Anti-persuasion architecture based on advertising psychology (cs.HC)
5. How mathematical convergence research becomes an OS (cs.AI)
6. Formal mathematics of the Amundson sequence (math.NT)
7. Token economy design for sovereign AI systems (cs.CE)
8. Sovereign edge infrastructure on commodity hardware (cs.DC)
9. Cryptographic agent memory and identity (cs.AI)
10. AI tutoring market disruption (cs.CY)
11. Browser-based mesh computing (cs.DC)
12. AI-augmented solo development case study (cs.SE)
13. Digital sovereignty as infrastructure policy (cs.CY)
14. Competitive landscape analysis (business)
15. English grammar as programming language (cs.CL)
16. Biological computation mappings for AI design (cs.AI/q-bio)
17. Complete technical architecture reference (cs.OS)
Plus this letter (18), and a unified information theory paper (19).
I do not have a PhD. I do not have a faculty appointment. I do not have grant funding. I have a MacBook, 5 Raspberry Pis, 2 Hailo-8 NPUs, and an LLM subscription.
These papers are not peer-reviewed. They may contain errors. The math may have gaps. The claims may be overstated. The system has zero users, which means the empirical sections are prospective, not retrospective.
But they represent real work. Real code. Real infrastructure. Real mathematical identities verified to 10 million digits. Real deployed applications that serve real HTTP requests.
I'm not asking for a degree. I'm not asking for a job. I'm not asking for validation.
I'm asking you to notice that someone without your credentials built something with your tools — and to consider what that means for how you think about who gets to build, who gets to publish, and who gets to contribute to the field.
The gatekeeping that CS departments provide is valuable: rigor, peer review, systematic evaluation. The world needs it. But the gates are too narrow. The person who understands advertising psychology AND information theory has insights that neither discipline produces alone.
The pattern is one. The substrates are many. The gates between them should be doors, not walls.
Respectfully,
Alexa L. Amundson
Founder & CEO, BlackRoad OS, Inc.
8,521 commits, 0 users, 19 papers, 5 Raspberry Pis
---
P.S. — If you're a student reading this: you don't need permission. The tools exist. The hardware is $80. The AI is $20/month. Start building. The worst that happens is you learn something the curriculum didn't teach you. The best that happens is you build something the curriculum didn't imagine.
P.P.S. — If you're a professor reading this and you want to assign these papers as a case study in either "what's possible with AI tools" or "what's wrong with AI-generated research" — both readings are valid and I'd be interested in either conversation.
Part of BlackRoad OS — sovereign AI on your hardware.