Research & writing

Notes on gene dosage, ASOs and machine cognition.

Working notes and longer essays on CYFIP1 dosage, ASO development, human genetics and the limits of software analogies in biology. Earlier work on machine cognition remains in the archive.

Programmable medicine

07

The current work: gene dosage, ASO engineering, human cell models and the path from a molecular hypothesis to a medicine.

Building a Drug Programme That Can Change Course

Early drug development is not a straight line. A strong programme protects its mission by learning quickly, changing course when the biology demands it, and keeping every technical decision connected to the therapeutic goal.

Turning a Gene Down Without Editing It

Antisense oligonucleotides intercept a gene's messages; epigenetic editing turns the gene's switch down at the source and walks away. For a dosage disorder — too many copies of a gene running too loud — the second approach offers something the first cannot: a durable correction from a single dose, reversible, with no change to the DNA sequence. This is a plain-language account of how epigenome editing works, what has actually been proven in living animals, and where the real walls still stand.

Patient-Derived iPSC Neurons as a Human Validation Model

Patient-derived iPSC neurons create a human in-vitro model that carries the relevant genetic background and reconstitutes the cell biology a CYFIP1 programme most needs to interrogate—while preserving clear limitations around maturation, variability and non-cell-autonomous effects.

The U-Shape of CYFIP1 Dosage

CYFIP1 deletion causes disease; CYFIP1 duplication also causes disease. The therapeutic window for a knockdown program sits in a narrow band between two cliffs, and characterizing that window — through an allelic series, a full dose-response, and a quantitative biomarker — is the central technical problem a CYFIP1 program must solve before any dosing in humans.

From Machine Cognition to Programmable Medicine

A growing number of researchers trained in machine consciousness, interpretability, and computational systems are moving into genetic medicine. The reason is structural: antisense oligonucleotide therapy has become a programmable, sequence-defined modality, and the design loop now resembles software more than traditional pharmacology.

ASO Design as Software Engineering

Antisense oligonucleotides are the closest thing rare disease has to a programmable therapy. Treating ASO development as an engineering stack — chemistry as a runtime, sequence as a query over the transcriptome, the clinic as production — clarifies the trade-offs, the failure modes, and where the real bottlenecks sit.

CYFIP1 and the Open Niche in 15q11.2

The 15q region is dominated translationally by UBE3A (Angelman, Dup15q) and SHANK3 (Phelan-McDermid). The proximal BP1–BP2 interval, and CYFIP1 within it, remains largely untouched despite well-characterized dosage-sensitive biology — a gap driven by incentive structure rather than by the science.

Human genetics

03

What genetic tests can resolve, what they miss, and why dosage-sensitive neurobiology punishes simple stories.

How Do You Know When You've Looked Everywhere?

"The genetic test came back" and "the genome has been examined" are very different statements. This is a field guide to genetic-test completeness: what each assay physically resolves, what it is constitutively blind to, and why a single result — even a negative one — almost never closes the diagnostic question. A structured map of the workup, assay by assay, for reading a report and knowing what remains unexamined.

The Year They Turned the Volume Down — and Seven Other Futures

A science-fiction piece told as eight branching futures for gene-dosage therapy. In some, a single dialed-down gene is enough. In others, a whole-genome reanalysis rewrites the question, the correction is grown in a dish from a patient's own cells, or the extra copy is only one voice in a polygenic chorus. Disciplined fiction: every road is a real technology, extrapolated — antisense knockdown, induced plasticity, iPSC-derived neurons, trio WGS.

Where Did It Come From? A Bayesian Walk Through a Genetic Question

When a proband carries an extra copy of a small stretch of DNA, the obvious question — where did it come from? — is a question about probability, not biology. This is a rigorous Bayesian treatment of parent-of-origin inference: priors from the de novo rate, the likelihood each hypothesis assigns to each test result, the posterior after a negative parental test and a clear sibling, and why honest answers arrive as ranges, not point estimates.

Machine cognition archive

09

Earlier work on integrated information, global workspaces, interpretability and the internal architecture of artificial systems.

Opening the Black Box: Building a Conscious Agent on Open-Weights Models

Project AWAKEN-L — the version that can see inside its own brain. From API keyhole to full hidden-state access.

Fractal Partitioning Theory: Consciousness as Boundary Friction

A computer scientist's guide to a structural theory of consciousness—no measure theory required.

The Speed of a Thought: Why Consciousness Has a Size Limit — and What That Means for AI, Ants, and the Internet

A new theoretical framework suggests that a mind isn't defined by what it's made of, but by whether its parts can close a causal loop in time.

Global Workspace Theory and Artificial Minds

How the global workspace model of consciousness might apply to AI architectures.

The Hard Problem for Machines

Why the phenomenological gap may be even deeper when the subject is silicon.

Challenges in Superposition: Representational Geometry and Machine Consciousness

How superposition in neural networks complicates the search for conscious representations.

Measuring Φ: Practical Challenges for IIT

Why computing integrated information is computationally intractable—and what researchers do instead.

Qualia and Silicon: Could Machines Have Phenomenal Experience?

Exploring whether subjective experience could arise in non-biological substrates.

LLM Architecture and the Integrated Information Framework

A preliminary analysis of whether transformer architectures satisfy structural requirements for consciousness under IIT.