Theodor Spiro
Independent researcher · Vaika Inc. (aging-research nonprofit, Gudkov lab) · Based in Tel Aviv, Israel ·
ORCID 0009-0004-5382-9346 ·
tspiro@vaika.org
Emergent Ventures grant recipient (2026)
Currently centered on perceptome — a framework treating cellular signaling pathways as a perceptual repertoire, with a Python toolkit (44 modules, 9-PC eigenspace, cancer attractor reference). Alongside: aging biomarkers across substrates (EEG, ECG, transcriptome), the comparative biology of neural networks (DFE and epistasis applied to LLM training), and a developing framework for doing science in active collaboration with AI.
About
I’m an independent computational researcher based in Tel Aviv, Israel. My background is in biophysics (Lomonosov Moscow State University, Faculty of Physics), where my early research was computational modeling of cellular signaling — a thread that runs directly into the cellular-perception work that now anchors the portfolio. I am a 2026 Emergent Ventures grant recipient.
Since 2024 I have been affiliated with Vaika Inc., a not-for-profit aging-research organization associated with the laboratory of Andrei Gudkov (Roswell Park Comprehensive Cancer Center); my aging manuscripts are developed and reviewed within that affiliation. The independent research program itself runs from 2020.
The work organizes into six directions:
- Cellular perception — perceptome (toolkit + framework + a growing family of papers) and related work on oscillatory signaling and cancer
- Comparative biology of neural networks — DFE, epistasis, and population-genetics tools applied to trained transformers
- Aging research / biomarkers — cross-population replications on cardiac, transcriptomic, and EEG substrates
- Cognition, education, experiments, social projects — instruments and studies of human cognition
- Methods & cross-substrate work — bridges, methodological contributions, and honest negative results
- AI-collaborative research methodology — the framework behind all of the above, applied as worked examples
I publish under one canonical name everywhere: Theodor Spiro (ORCID 0009-0004-5382-9346). Manuscripts and code are released on GitHub at github.com/mool32 and on Zenodo / arXiv.
Currently
perceptome v0.3 is the active center of the work — a Python toolkit for cellular perception analysis with a 9-PC eigenspace, capacity-floor predictor, validity scorecard, and an 8-cell cancer attractor reference. A family of papers around the framework is in preparation. In parallel: functional-differentiation DFE manuscript writeup, the epistasis Tier-2 extension on Pythia 410M / OLMo-2 1B, and the clonal-crystallization cross-substrate bridge preparing for submission.
News
- Jun 2026 Oscillation as a constitutive substrate for perception — a multi-phase pre-registered dissection on a Stuart–Landau triad and a trained GRU triad. A trainable triad self-organizes a coupling-maintained 120° rotating limit cycle and self-sustains it with the loss off, but it is NOT a memory store (the coupling that holds the cycle suppresses capacity — the founding hypothesis refuted). It IS a phase regulator with a Hopf-universal temporal code, emergent in the trained GRU. The earned law: an oscillation’s existence is cheap; its directed rotation — the breaking of time-reversal symmetry — is the one irreducibly-active ingredient.
- Jun 2026 Neural-network fitting as a structure sensor — a pre-registered negative. An autoencoder’s unsupervised fit does not beat matched linear PCA at recovering donor age from immune scRNA-seq (GSE233321, 167 donors), and training erodes the signal; a synthetic positive control pins the exact regime where the method would win. The latent independently re-derives the perceptome eigenspace (CCA p = 0.001) but not its aging axis.
- May 2026 Awarded an Emergent Ventures grant (2026) in support of the research program.
- May 2026 perceptome v0.3 released — Python toolkit for cellular perception analysis (44 signaling modules, 9-PC eigenspace from 154 HPA cell types, capacity-floor predictor, validity scorecard, 8-cell cancer attractor reference). Zenodo DOI 10.5281/zenodo.20113468. 73/73 tests passing.
- Apr 2026 Universal statistical signatures of evolution in AI architectures — preprint deposited on arXiv (2604.10571). 935 ablation experiments from 161 ML publications show that the DFE of architectural modifications matches biological DFEs.
- Apr 2026 Spectral exponents of the twelve-lead ECG submitted to bioRxiv. 412,730 recordings across three continents; CLBBB vs CRBBB AUC = 0.982 with cross-population invariance.
- Apr 2026 Transcriptomic noise accumulates within tissue identity — post-review v4 archived (Zenodo 10.5281/zenodo.19944444); GTEx + Tabula Muris Senis + Calico rat + macaque.
- Mar 2026 Waveform asymmetry as a biomarker of neural aging submitted to Frontiers in Aging Neuroscience (LEMON N=215, Dortmund N=608, 5-year longitudinal subsample N=208).
Selected work
-
perceptome — a Python toolkit for cellular perception analysisA Python toolkit treating 44 cellular signaling pathways (NF-κB, mTOR, UPR, p53, nuclear receptors, …) as a cell's perceptual repertoire. Three capabilities other tools don't directly provide: (1) a 9-PC eigenspace derived from 154 normal human cell types (Human Protein Atlas) where any cell — yours, a tumor, a drug-treated — can be located on the same map; (2) a capacity-floor predictor — given a cell type and a pathway, will it ramp under stimulus or is it already saturated; (3) a validity scorecard with three null controls (random-gene panel, housekeeping, cell cycle) that catches common perturbation artifacts before interpretation. The framework's headline empirical result is an 8-cell attractor cluster (gastric chief, pancreatic acinar, ..., gastric progenitor) that 11 cancers from 11 organ systems converge toward during transformation — independently replicated on the Sun et al. 2021 paired HCC cohort. Result of a multi-year, pre-registered research program; the tool implements only what those studies validated and deliberately excludes operations that were tested and falsified.
-
Spectral exponents of the twelve-lead ECG reveal the anatomy of cardiac conduction disorders and a bifurcation between aging and diseaseSpectral exponent β extracted from 412,730 twelve-lead recordings across three continents (PTB-XL, Chapman-Shaoxing, CODE-15%). β is diagnostic of cardiac conduction anatomy: CLBBB vs CRBBB AUC = 0.982 (Germany), 0.982 (China), 0.979 (Brazil). Aging flattens the spectrum; disease steepens it. Honest null on independent mortality prediction (HR = 1.02, p = 0.83).
-
Transcriptomic noise accumulates within tissue identity across human agingThree-level variance decomposition on bulk transcriptomes from 263 GTEx v8 donors plus single-cell data from Tabula Muris Senis, Calico rat caloric-restriction atlas, and a rhesus macaque cross-species atlas. Tissue identity is preserved across forty years of aging; the signature is systemic noise, not selective accumulation. Caloric restriction acts as a noise filter, not a structure restorer. Cross-species lifespan scaling (α = −1.02, R² = 0.90).
Additional selected work
-
Universal statistical signatures of evolution in artificial intelligence architectures935 ablation experiments compiled from 161 publications. The DFE of architectural modifications matches biological DFEs (heavy-tailed Student's t; 68% deleterious / 19% neutral / 13% beneficial for major ablations); architectural diversification follows logistic dynamics with punctuated equilibria; 14 architectural traits were independently invented 3–5 times in parallel with biological convergences.
-
Waveform asymmetry as a biomarker of neural aging: spatial degradation of oscillatory cycle shape across two independent cohortsPeak-trough asymmetry of EEG oscillations across five frequency bands in 215 (LEMON) and 608 (Dortmund Vital Study) adults, with 208-subject 5-year longitudinal follow-up. Beta-band asymmetry decreases with age (r = −0.326 LEMON, r = −0.314 Dortmund); effect size exceeds classical alpha slowing. Theta spatial entropy predicts memory independent of age.
-
Clonal crystallization as a shared signature of bone-marrow aging and neural-network trainingA two-metric framework (Gini + effective N) applied jointly to mouse cell-type proportions and to Pythia-410M head-importance distributions. Pythia moves through (ΔGini, Δeff_N) by (+0.145, −45.5) over 143k steps; bone marrow on FACS and Droplet platforms moves in the same quadrant. Caloric restriction in rat bone marrow rescues 64% of the Gini drift. A specific substrate-independent signature, narrower than full DFE universality.
-
Functional differentiation generates universal fitness-effect distributions in neural networksAblation studies of 144 identity-tracked attention heads, 24 transformer layers, and 30 distributed-noise perturbations at each of 8 checkpoints (step 512 → 143,000) of Pythia 410M-deduped — 1,584 ablations. The DFE shape evolves from delta-peak-with-outliers to heavy-tailed Student's t (df ≈ 2.3, β ≈ 0.6 within biological range). 39% of eventually-critical heads emerge from noise; only 4% are born critical. L8H9 undergoes phase-transition-like specialization between steps 4k–8k.
-
The Oracle's Fingerprint: correlated AI forecasting errors and the limits of bias transmissionThree studies on whether the wisdom of crowds survives consultation of a small set of LLMs. GPT-4o, Claude, and Gemini show pairwise error correlation r = 0.78 on 568 resolved binary Metaculus forecasts (Study 1). Community forecasts move in the LLM-predicted direction (r = 0.20, p = 0.007), but the shift is fully explained by rational updating toward ground truth (Study 2). Pre-ChatGPT human biases already strongly resembled the LLM pattern (r = 0.87); post-ChatGPT the resemblance weakened (r = −0.28) — LLMs inherited human biases rather than introducing novel ones (Study 3).
-
21,000 attempts to think differently: a large-scale Russian adaptation of the Divergent Association TaskRussian-language adaptation of the Divergent Association Task (Olson et al., 2021). 21,159 submissions; Cronbach's α = 0.899; split-half reliability 0.696; no measurable practice effect. Strongest predictor: semantic category diversity (r = 0.47). Theoretical ceiling 110.5 (best human attempt: 104.8). Live instrument runs at mool32.github.io/dat-ru.