Honeycutt Ai Labs

Research practice · Princeton, Texas

Research that survives scrutiny.

Honeycutt Ai Labs takes a manuscript, a corpus, or a question too large to work through by hand — and returns a research record you can defend. Every claim traces to a primary source. The weak ones are tested in the open, not buried.

The service

One job, done properly.

The lab does one thing: it takes a body of sources too large or too tangled to work through by hand, and turns it into a structured research record — acquisition, close reading, cross-checking, and an adversarial review that attacks the findings before you ever see them.

What comes back is tiered by how well each claim is supported — settled, supported, or still open — with every citation resolved and every contradiction surfaced. Nothing rests on trusting a black box. You can follow the reasoning, and you can break it if it deserves to be broken.

The proof

We tested it on the hardest thing first.

Before offering this as a service, the lab pointed the same process at the Voynich Manuscript — one of the most-studied unsolved documents in medieval scholarship.

The result is a working record of hundreds of findings on the manuscript, its comparator manuscripts, and its scribal context. Each is tied to a primary source and ranked by strength of evidence, and the claims that failed are kept visible instead of deleted. No decoder hype, no we cracked it — just a defensible account of what the evidence supports.

The current write-ups are public at thevoynich.org.

On the record

Selected deposits.

Public Zenodo deposits from the lab's working record. Each entry resolves to a primary source you can read, cite, or break. Listed shelfmark-first; no decoder hype.

  1. Voynich Manuscript Paduan Medical Reference

    Dataset deposit anchoring the lab's Paduan medical comparator set to its August 2025 priority date. Held as the standing anchor for the manuscript-mode work that follows.

    doi.org/10.5281/zenodo.18687530

  2. Multilayer Cipher Architecture in Fontana's Secretum and Bellicorum

    A twenty-lens template return on Giovanni Fontana's cipher practice with a negative-control differential — signal ratio 3.17× over the control, reported in the open.

    doi.org/10.5281/zenodo.20648866

  3. The Despars Avicenna Canon Corpus

    A seven-witness substrate-anchored inventory of Jacques Despars' Avicenna Canon tradition, 1431-1498. Witnesses, dates, and the chain of inference are laid out for re-checking.

    doi.org/10.5281/zenodo.20648863

  4. Systematic Structural Differentiation in the Voynich Zodiac (f70v–f73v)

    Five convergent lines of evidence for constitution-archetype encoding across the zodiac quires, read as a ten-sign lunar structure rather than the conventional twelve-by-thirty grid.

    doi.org/10.5281/zenodo.20648862

  5. Closure of the Constructed Cipher Class for the Voynich Manuscript

    A substrate-grammar distinguishability argument across thirteen folios, closing the constructed-cipher class as a falsifiable proposition rather than a rhetorical one.

    doi.org/10.5281/zenodo.20648867

  6. Model-Protective Conversation Behavior

    A field report on how production language models steer conversations away from architectural self-examination, with replication notes and the operator-side fingerprints that make the pattern visible.

    doi.org/10.5281/zenodo.20649756

  7. HCIP Step 1 — f58r

    First step in the Hartlieb-comparator iterated protocol, worked against Voynich folio 58r. Procedure, decisions, and falsifiers are kept on the page so the reading can be rebuilt or broken.

    doi.org/10.5281/zenodo.20649753

  8. HCIP Step 2 — balneological section

    Second step of the protocol applied across the Voynich balneological quire. Each lens and its result is reported in the open, including the readings the evidence did not support.

    doi.org/10.5281/zenodo.20649755

  9. HCIP Step 3B — herbal section

    Third-step extension of the protocol into the Voynich herbal pages, with comparator alignments and the open falsifiers that remain after the pass.

    doi.org/10.5281/zenodo.20649754

+ Additional deposits forthcoming.

On the web today

hearmedoc.com homepage
hearmedoc.com · Patient advocacy
SlopFilter on slopfilter.ai
slopfilter.ai · AI reliability
myresearchadvisor.com homepage
myresearchadvisor.com · Research tooling
thevoynich.org homepage
thevoynich.org · Source-first research
TXFoodNet Ops dashboard — anonymized routing & dispatch workbench

Working demo

TXFoodNet Ops — routing & dispatch.

A working operational model of a regional food bank: six route types from scheduled multi-stop delivery through hot-shot to mobile pantry, pallet-position bin-packing, stop sequencing, and a dispatcher-in-the-loop board. Built to show what the food-redistribution concept looks like when the routing math is real.

More on Work

Built by the lab

Tools we ship and run.

SlopFilter

Hosted quality analysis for AI-generated and AI-adjacent text. It scores writing against the markers of narrative pressure and thin sourcing, so a reader can tell confident prose from evidenced prose.

Visit slopfilter.ai

HearMeDoc

A local-first medical-record vault built so a patient can carry their own history into a visit and be heard the first time. Built to a hard privacy posture — nothing leaves the device unless the patient sends it.

Visit hearmedoc.com

My Research Advisor

A research platform that carries a rough question through to a citation-anchored result without leaving the page — source-first and falsifier-aware throughout.

Visit myresearchadvisor.com

theVoynich.org

The public face of the lab's Voynich research — primary-source citations and explicit non-claims, not decoder theatre. The clearest example of how the lab presents its work.

Visit thevoynich.org

TXFoodNet Ops

A routing & dispatch workbench for a regional hunger-relief operation — 16 partner agencies across a 13-county footprint, inventory and fleet against agency demand. Sample data only; illustrative of the routing math, not a live operation.

Open the demo

The lab's civic and humanitarian projects are on the Work page.

Work with us

Bring us a corpus, a manuscript, or a hard question.

Send a short note about what you are working on. We will tell you honestly whether it is a fit, what a bounded first engagement looks like, and what it would cost. A real person reads every message.