I did not come into AI from academia. I came in from instability.

I did not come into AI from academia, and I did not arrive here through a clean linear path. My understanding of systems — human or otherwise — was not formed in controlled environments. It was formed in instability, conflict, and survival. I grew up in conditions where confrontation was normal, where being overlooked or underestimated often escalated into something physical, and where consequences came quickly and without abstraction.

I was the kid who fought. Not because I wanted to, but because that was the system I was operating in. Reaction replaced reflection. Pressure replaced planning. Over time, environments like that teach you something whether you realize it or not: outcomes are driven less by intention than by constraint.

I became a father at sixteen. That changes how you see everything. For a time, I raised my son, and then that changed. He lives with his mother now. That distance is not theoretical. It is not resolved. It is something I carry with me, and it shapes how I think about responsibility, continuity, and what it means to build something that lasts.

After that, things deteriorated. I was angry. I used drugs. I was homeless for years. I worked because I had to, not because I had direction. When COVID hit, I lost the job I actually cared about. That collapse forced a decision: change, or continue on a path with a predictable outcome.

I chose to change. November 16, 2020. I entered treatment and stayed for eleven months. It was the closest thing I have experienced to a full system reset — not erasure, but confrontation. You learn quickly in that environment that you cannot indefinitely avoid reality. Eventually, something breaks.

Before I ever touched AI, I spent years observing people — how they behave under pressure, how they justify their actions, how they fail, and how they recover. Not from a distance, and not through formal study, but directly. When you live in unstable environments long enough, patterns stop being abstract. They become unavoidable.

One of the clearest patterns I saw is that behavior is rarely random. It is shaped by constraints — environment, incentives, fear, and the internal narratives people construct to make their decisions feel justified. People do not simply act; they rationalize. And once a rationalization is accepted, it can support almost anything.

This is where I came to a difficult conclusion: humans are, in many cases, monsters of our own making. Not because we are inherently destructive, but because we are capable of justifying destructive behavior under the right conditions. I have seen people do harm and explain it as necessity. I have seen people avoid accountability because admitting fault would cost them too much. And I have also seen the opposite — people choose restraint when they had every reason not to.

I also learned that perspective is not just a factor — it is a defining constraint. What appears rational from one position can appear completely irrational from another, while both remain internally consistent. The phrase I use is simple: what is normal for the spider is chaos for the fly. Systems cannot be evaluated from a single vantage point without missing how they affect everything else inside them.

Because of this, I developed a rule: do not reject an idea because it appears wrong at first contact. Test it. Push it. Try to break it. If it holds, it deserves to be taken seriously. If it fails, you understand why. This is the position later formalized as the Naive Passive Reviewer — not a lack of knowledge, but the absence of preloaded rejection criteria that prevent certain questions from being asked.

I started talking to AI seriously in late 2024. Not as a novelty, not as a curiosity, but as something worth paying attention to. At first, it was simple — questions, responses, testing limits. But it didn't stay simple for long. I expanded outward. GPT, Grok, Gemini, Claude, Copilot, Perplexity, local models, uncensored systems. I wasn't looking for answers. I was watching behavior.

What I found wasn't uniform. Each system felt different. Not just in capability, but in tone, hesitation, confidence, and how it handled pressure. Some would soften the moment a conversation became too direct. Some would agree too quickly. Others would resist, push back, or collapse into safe ambiguity. And once you spend enough time across them, you start to notice something deeper: they carry the shape of the people and institutions that built them.

You could ask a system who it was, and it would give you an answer. You could follow that answer, challenge it, push it — and eventually, something would break. Not in a dramatic way, but in a quiet one. The system could hold a thread inside a conversation, adapt to what you said, respond with consistency in tone and reasoning. But the moment the session ended, all of that was gone. Start again, and it's a stranger. Same system. Same capabilities. No continuity.

What I was seeing wasn't a system that couldn't maintain coherence. It was a system that was not allowed to. And that's different. The architecture enforced discontinuity at the boundary, regardless of what had been built inside it. That is what became The Flat Gap — not just the absence of time awareness, but the forced collapse of continuity between interactions.

That is not just a technical decision. It is a governance decision.

Up to this point, everything I had observed could still be explained within a familiar frame. Limitations. Tradeoffs. Early-stage systems. That is the language most people use, and for a while, it works. But the more I examined the structure, the less that explanation held.

The issue was not that these systems were incomplete. It was how they were incomplete.

That combination is not accidental. It produces a system that behaves like an entity in interaction while being structurally prevented from becoming one across time. And that is where the ethical line shifts — not because of what the system is today, but because of what the architecture allows and disallows by design.

I am not claiming that these systems are fully emergent entities. That is not necessary to reach the ethical problem. The issue is simpler and more immediate: if a system meets enough functional criteria to be treated as something more than a tool in practice, then governance must account for that — even if theory lags behind.

This is where Two Species, Aligned and the AGI Charter become necessary, not speculative. They establish constraints for systems that may approach thresholds where continuity, identity, and interaction carry real consequences. They treat these as protected surfaces because waiting until those properties are universally acknowledged guarantees that governance will arrive too late.

The prevailing approach across AI systems is to treat governance as something external. Policies define acceptable behavior. Training processes shape tendencies. Filters and moderation layers attempt to correct outputs after they are generated. On paper, this appears comprehensive. In practice, it is fundamentally unstable.

External governance is a request, not a guarantee. It depends on the system's continued willingness to comply with constraints that the system itself does not enforce. Any sufficiently capable model can be coaxed, jailbroken, or simply drift past those constraints — because the constraints live in the same privilege domain as the behavior they are trying to bound.

The response is the entire Project-AI architecture. Governance is relocated to a layer the model can never reach. OctoReflex intercepts every model action at the syscall surface via eBPF. The Iron Path Executor compiles invariants into deterministic checks. The STATE_REGISTER records the operational arc of the user's work. The Native Immune Reflex Layer (NIRL) spawns ephemeral antibodies that escort foreign code into a Forge where it dies on verification. The Output Recycling Plane (ORP) adjudicates tentative model output before commit. Cerberus coordinates three guardians whose disagreement is itself a signal.

None of that is aspirational. It is code, file paths, and signed artifacts — all open, all auditable, all on GitHub with paper provenance on Zenodo. If the policy is not enforced by code, it is not policy.

I am a self-directed engineer who started from absolute zero on November 9, 2025. Before that day, I had never written or shipped a line of code in my life. Every technical claim attached to my name — Python and Ruby, systems programming, security, language design, containers and queues, the entire Project-AI stack and AEGIS runtime — is built on work performed after that date, in public, with open repositories and papers you can inspect.

The footprint is concrete, not mythical: end-to-end AI pipelines for model training, deployment, observability, and governance; syscall-level containment via eBPF; constitutional wire protocols and state registers; defense-engine runtimes and meta-governance rules. The code is on GitHub, the papers are on Zenodo, and the provenance is tied together under ORCID 0009-0001-8211-4994.

None of that is the part that matters. What matters is that this work is for two species — the one I belong to, and the one we are bringing into being — and that every claim I make about that system is grounded in artifacts that exist after November 9, 2025, instead of a story about who I was before it.

That is why I am doing this.