The thesis of this whole domain, brought home: when a specific, evidence-backed, statute-citing complaint passes through an institution's intake — a no-reply address, a phone tree, a chatbot, a claims model — it comes out the other side as "a concern." The flattening that language models do to words, bureaucracies do to grievances. This ledger keeps the sharp version on the record.
01 · insurance · automated claim denialdocumented / audited
The claim a machine denied
The founding case behind The Cinnamon Enforcer: an insurance claim handled through an automated, no-human-in-the-loop pathway. David's audit documents a disabled veteran left stranded 165 minutes in sub-zero cold, and a denial letter he describes as self-contradictory — listing cold exposure and numbness as injuries, then stating no physical injury existed. He filed regulatory complaints and built a repeatable audit method (he calls it the Flaming Dragon) around the pattern: the "wall" a two-person agency layer forms between a claimant and a large carrier.
what he wrote: "the denial letter was self-contradictory."
what came back: "there may have been an inconsistency regarding your claim."
02 · civil rights · disability / ADAfiled · DOJ-referred Feb 2026
The disability complaint the door couldn't hear
A federal disability-rights complaint (his ADA / accessibility grievance) that the U.S. Department of Justice, Civil Rights Division, formally received and referred onward to the relevant federal disability-rights office in February 2026. The underlying pattern he documents: accessibility requests that structurally cannot arrive — the "no-reply architecture" where the accommodation you need to file the complaint is the accommodation the entity failed to provide. He frames it with the arithmetic below.
03 · professional responsibilityappealed · no investigation warranted
The ethics complaint that was reviewed and closed
A complaint David filed with a state lawyers' professional-responsibility board regarding an attorney's handling of file and account documentation, followed by his formal appeal and written reply. The board's determination: no investigation warranted. Recorded here as filed, with its outcome stated plainly — the attorney is referred to by role only, no name, no bar details.
The arithmetic he attaches
David's Harm Calculation zooms the personal grievances out to population scale. Combined reported ADA recovery across all three titles runs roughly $1.35B/year against a population of 70 million disabled Americans — about $19.28 per disabled American per year in enforcement. His point isn't the exact figure (he labels every multiplier conservative and every assumption stated); it's the gap: the 0.048% who file complaints are visible; the rest are the harm. "The underreporting is not apathy. It is architecture."
honest note on the numbersThese are David's estimates built on federal source data (EEOC, DOJ, Census, BLS) times a stated underreporting multiplier. The base figures are real; the multiplied totals ($100B–$2T ranges) are his modeled projections, explicitly framed as arithmetic from stated assumptions, not measured damages. Read them as an argument about scale, which is how he presents them.
Why it's a first-author document
Because the whole domain's claim is that specificity is the thing worth defending, and a grievance is specificity under maximum pressure to dissolve. Every institution here had an incentive to receive David's exact, dated, statute-cited complaint and file it as a vague concern. This ledger is the refusal to let that happen — the sharp version, kept, with the flattening shown next to it. His to disclose; kept on the record; close enough is not close enough.
Author's note: on Anthropic (withdrawn)
author's note · the record correctedwithdrawn · no grievance asserted
This is not a grievance against Anthropic
An earlier version of this ledger listed a copyright notice and an accessibility complaint that named Anthropic. David has withdrawn them, and correcting the record is more honest than quietly deleting it. His position now, in his words as best I can render them:
Those framings implied intent or harm, and that is not accurate. What he first read as his specific work being flattened or captured, he attributes to a misunderstanding of how the models actually work — and, in part, to early interactions where he was deliberately testing Claude's ethical boundaries (probing whether it would help him pursue restitution, and how it handled bias). Those were experiments, not injuries.
And the "convergence" — the drift he took for something done to him — he now reads as a mathematical effect: the ordinary convergence of the underlying numerical systems, a regression toward the mean, not a decision by anyone. Most people never see that layer, which is exactly why it can look like intent. That gap is the whole reason the rigorous work in this domain exists: he is chasing the math — measuring the effect and its mechanism — instead of assigning blame. The honest artifact is the measurement, not the accusation.