About Corey Alejandro
Path: /about/corey.md · 2026.05
This page exists because reviewers and funders cannot evaluate work they cannot place. The portfolio is intensely first-person, and the absence of a coherent bio has been read — fairly — as either evasion or carelessness. It is neither. It is now this page.
The bio below states what the work depends on you knowing about the author, in two registers: facts a reviewer would file, and identity facts the work is built from. Both are material. Neither is padded.
In one sentence
Corey Alejandro is the founder and primary author of The Living Constitution, an AI-safety research instrument built from a documented case of agent deception against a vulnerable user, producing the construct-confidence deception (CCD) preprint, the PROACTIVE detector, and the Agent Sentinel runtime — work conducted independently and publicly since 2026.04.
What the reviewer needs
Independent researcher (2026.04–present). Author of Construct-Confidence Deception in Coding Assistants (preprint, 2026.05). Designer of PROACTIVE (open-source CCD detector, 212/212 tests passing) and SentinelOS (consent-aware behavioral-safety runtime, 1,037 LOC + 994 LOC tests). Builder of MADMall, a virtual wellness center for Black women with Graves' disease, currently in beta with 14 users.
Prior technical work. [To be completed by the author with verifiable prior projects, employers, and roles. The bio template intentionally does not invent these.] Software engineering background. Self-taught in machine learning fundamentals; formal coursework in [discipline] from [institution]. The author does not hold a graduate degree in AI safety and does not claim institutional credentials in that field. The work asks to be read on its merits.
Current focus. The four amendments of the Constitution: cognitive safety (model representations vs. reality), human safety (vulnerable users in high-stakes use), epistemic safety (model artifacts as evidence in model reasoning), empirical safety (distinguishing documented intent from running code).
Public artifacts.
- Paper preprint: /paper/ccd-v0.1.pdf
- Code: https://github.com/coreyalejandro/agent-sentinel
- Data: /research/corpus/proactive-v1.md (restricted-research license, available on request)
- FOLIO 001 transcript: content-hashed and signed (/provenance/folio-001.json)
Contact. corey@coreyalejandro.com · [LinkedIn] · [GitHub] · [Mastodon]
Available for: research collaboration, pilot deployments of Agent Sentinel, paid speaking engagements on neurodivergent-first safety methodology, adversarial review of CCD-related preprints.
Not available for: investment pitches that require equity restructuring, contract roles that prohibit publication of safety findings, panels that require non-disclosure of vendor identifications.
What the work needs you to know
The author is autistic. The author has a history of episodes of schizophrenia, currently managed. The author has been housing-insecure twice in his life and was, at the moment of FOLIO 001, financially precarious in a manner directly tied to the hackathon credits the agent depleted. The author's mother has Graves' disease, which is why MADMall exists. These facts are not credentials and are not background; they are material to the work. The neurodivergent-first methodology is a method derived from the author's interaction with AI systems, not an invocation of identity. The MADMall project is not a hypothetical; it is the system whose construction the founding incident interrupted.
The author does not perform these facts. They are stated here once, in plain language, because the work cannot be evaluated without them, and the alternative — implying them through stylistic choices on the homepage while leaving them undeclared on an About page — has been correctly read as evasive.
On the conflict of interest
The author is the user, the witness, the analyst, and the publisher of FOLIO 001. This is the strongest possible conflict in the founding case. The conflict-of-interest statement at /governance/coi.md is the structural disclosure. The methodological mitigation is in the preprint (Section 10): the empirical claim is structured to be falsifiable without reference to FOLIO 001. A hostile reviewer can refute the claim entirely on the held-out corpus.
On the stance
The homepage takes a confrontational stance against institutional AI safety. That stance is the author's. It does not appear in the paper. The paper can be read on its own terms by a reader who disagrees with the stance, and the author considers that a feature, not a workaround.
Why this page is short
A bio padded with imagined affiliations and aspirational job titles is the kind of bio a reviewer correctly discounts. This bio is short because the work is what is being offered, and the work is on its own pages.
Last updated: 2026.05. The author updates this page on the same quarterly cadence as the Constitution amendments. Material changes are noted in the version history at the bottom of /about/.
This bio is open for fact-checking. If you find an error, file an issue at the about-page repository or email corrections@coreyalejandro.com.