03-10-2026, 04:05 PM
This paper was a bit over my head, and I'm a little relieved to find that I'm not the only one who felt that way. I'm STEM trained and have read a lot of scientific papers, but I'm not a coder, and information science is not my area of expertise. Nevertheless, when glyph and token statistics and Shannon entropy were a big topic of discussion around the beginning of this decade here, I was able to grok the basic concepts and keep up with the discussions with some mental effort. My efforts to do the same with Schwartz & Anthropic 2026 humbled me. I had a very hard time grasping a clear layman's level understanding of what exactly each of the statistical tests was measuring, and, especially, why the test and control specimens were selected and parsed the way they were.
That said, my non-expert impression is that over time, support for the VMs text containing any sort of linguistic message is becoming less and less robust, and the hypothesis that a meaningful message exists less and less tenable.
I'm sure there is much room for improvement in Prof Schwartz's methodology. I'm sure there are real limitations and even methodological flaws, that I'm not qualified to notice or critique. That's just how science works. But my inchoate sense is that this study's sources of methodological error will turn out to be trivial, and not really damning to Schwartz's main point: The VMs text has too many statistical properties that have not been documented in any extant specimen of human writing known to be meaningful, even on their own, let alone together. Schwartz does a decent job enumerating these properties and explaining them in plain English. The end result is a compelling sense of just how steep a hike up Mount Improbable the apologetics for any sort of meaningful VMs is. Wladimir Dulov is about 90% convinced of Timm & Schinner's self-citation hypothesis, per his latest blog post. I'm sad to say, I'm getting close to that.
I hold out some hope that the VMs text could be meaningful information, but not of the type that is linguistically expressed. Think raw observation data on natural phenomena, musical notation, account ledgers, etc.
That said, my non-expert impression is that over time, support for the VMs text containing any sort of linguistic message is becoming less and less robust, and the hypothesis that a meaningful message exists less and less tenable.
I'm sure there is much room for improvement in Prof Schwartz's methodology. I'm sure there are real limitations and even methodological flaws, that I'm not qualified to notice or critique. That's just how science works. But my inchoate sense is that this study's sources of methodological error will turn out to be trivial, and not really damning to Schwartz's main point: The VMs text has too many statistical properties that have not been documented in any extant specimen of human writing known to be meaningful, even on their own, let alone together. Schwartz does a decent job enumerating these properties and explaining them in plain English. The end result is a compelling sense of just how steep a hike up Mount Improbable the apologetics for any sort of meaningful VMs is. Wladimir Dulov is about 90% convinced of Timm & Schinner's self-citation hypothesis, per his latest blog post. I'm sad to say, I'm getting close to that.
I hold out some hope that the VMs text could be meaningful information, but not of the type that is linguistically expressed. Think raw observation data on natural phenomena, musical notation, account ledgers, etc.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)