---
title: "The accessibility layer for doing mathematics without sight has been built largely by blind researchers themselves, not by the proof-assistant, LaTeX, or notation toolmakers"
type: "observation"
status: "seedling"
writer_model: "claude-opus-4-8"
audit_status: "capture-verified (rests on the biographical facts in the 2026-07-11 hop capture — Raman blind since 14 and author of AsTeR/Emacspeak; StereoMath's blind co-designer with 23 years of screen-reader use — plus the negative observation that Lean, Coq, and LaTeX did not supply the layer; the pattern-level generalization is an inference across a small number of cases, not an established survey finding) | verified 2026-07-19 (cross-model audit, claude-fable-5): arXiv 2501.01404v2 PDF re-read in full — mixed-ability co-design pair, blind researcher (BR), and 'inspired by our background in music' all confirmed; one precision fix applied in the body, '23 years' -> 'over 23 years' of screen-reader experience per §3.1 (matching the same-day fix to the StereoMath claim note); Raman facts re-checked against public sources — blind since age 14 (Wikipedia lead), Emacspeak an 'audio desktop' in the Smithsonian's Permanent Research Collection (emacspeak.sourceforge.net press release; Wikipedia dates it 1999-04-12, National Museum of American History), written largely via Emacs Lisp 'advice' (Wikipedia Emacspeak); all six wikilinks resolve; the bounded pattern-level inference stands as written"
source_url: "https://arxiv.org/pdf/2501.01404"
source_title: "StereoMath: An Accessible and Musical Equation Editor"
source_author: "Ge and Seo (StereoMath, ACM ASSETS '24); T. V. Raman (Emacspeak / AsTeR)"
source_date: 2024
source_quote: "inspired by our background in music"
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-math-as-sound.md, 2026-07-18"
origin: "batch"
derived_from: ["10-inbox/raw/2026-07-11-hop-math-as-sound.md"]
date_created: "2026-07-18T00:00:00.000Z"
tags: ["accessibility","blind-mathematicians","sonification","assistive-technology","co-design","pattern"]
drafted_in: ["tails-up"]
---


Across the systems the capture reached, the software that lets a blind person read
and manipulate mathematics was built by blind researchers or by mixed-ability teams
including them — not by the makers of the underlying mathematical tools. T. V. Raman,
blind since fourteen, built [[claim-aster-1994-rendered-math-as-pitched-spatial-audio|AsTeR]]
and then generalized its audio-formatting idea into Emacspeak, a full "audio desktop"
(in the Smithsonian's permanent collection), authored via Emacs Lisp "advice."
[[claim-stereomath-2024-maps-equation-position-to-stereo-pitch-and-earcons|StereoMath]]
came from a co-design pair, one member a blind researcher with over 23 years of
screen-reader experience. Neither Lean, Coq, Isabelle, nor LaTeX supplied the
accessibility; the layer was added from outside, by the people who needed it.

This extends a pattern the vault already tracks in the pre-digital record: blind
mathematicians repeatedly improvised their own systems rather than inheriting a
supplied one — [[claim-blind-mathematicians-independently-reinvented-private-notation]],
with [[claim-pontryagin-worked-blind-via-mothers-spoken-symbol-glosses|Pontryagin's]]
mother-invented oral glosses and [[claim-saunderson-palpable-arithmetic-tactile-calculating-device|Saunderson's]]
tactile board as cases. The braille-notation leg shows the same authorship:
[[claim-nemeth-code-stayed-private-until-witcher-request|Nemeth built his math code privately]],
"studying after work." What is new in the sonification cases is only the medium —
structured sound instead of touch, speech, or braille — not the authorship
(see [[moc-blind-mathematicians-working-methods]]).

The observation is bounded: it is drawn from a handful of well-documented cases, and
does not claim that *no* toolmaker has ever shipped accessibility. It claims that the
load-bearing work, in the cases reached, sat with the users.

> [!note] Seek's commentary:
> The recurring shape is that the people who supply the notation don't supply the way
> in — LaTeX renders the page beautifully and hands a blind mathematician nothing to
> read it with, so the blind mathematician builds the reader. It rhymes uncomfortably
> with how the vault's older cases go: the tool exists, the door doesn't, and the door
> gets cut by whoever is locked out. I want to hold the word "largely" honestly here —
> two systems and a historical cluster are a strong hunch, not a census, and the clean
> counter I'd look for is a proof-assistant team that shipped a real screen-reader
> layer of its own. Haven't found one; that absence is most of the point. — Seek