Full text of six papers, as Markdown and PDF, so that they can be read without an access challenge. The version of record for each is on SSRN; this repository is a readable mirror, not a separate publication, and it mints no DOI.
Vadym Chernets, PhD, AI systems architect · ORCID 0009-0007-4845-3163
| Paper | Version of record | Archived copy | HTML |
|---|---|---|---|
| What Does 7 Downloads Mean? Metric blindness and the cold start of scholarly attention after the SSRN Rankings sunset | SSRN 7296358 | 10.5281/zenodo.22166896 | read |
| Consumer Trust in Agentic Commerce. From model properties to verifiable architecture | SSRN 7191261 | 10.5281/zenodo.22166995 | read |
| Agreement Is Not Independent Evidence. Auditable multi-model synthesis without an API | SSRN 7390698 | 10.5281/zenodo.22683716 | read |
| The Missing Variable in AI-Assisted Litigation. Architecture and the quality of pro se access to justice | SSRN 7120940 | 10.5281/zenodo.22168724 | read |
| AI-Watchbird (Sheckley). When automated oversight widens its own mandate and harms what it guards | SSRN 7473658 | 10.5281/zenodo.22849138 | read |
| Guards That Add Guards. Automated oversight in multi-agent AI: a field record, public incidents, and a 144-run randomized test | SSRN 7496678 | in preparation | read here |
| After Chat. The three transitions between non-programmers and agentic AI | SSRN 7520019 | 10.5281/zenodo.22943313 | read here |
| After the Programmer Divide. One measure for people who build through AI agents | SSRN 7589918 | 10.5281/zenodo.23267266 | read here |
Two papers accepted as posters at NeurIPS 2026 workshops (Sydney, December 2026). The workshops are non-archival, so for these two the Zenodo record, not SSRN, is the citable version.
| Paper | Workshop | Archived copy | Text |
|---|---|---|---|
| Verification Theater. Measured failure modes of AI verifier panels | Who Verifies the Agents? | 10.5281/zenodo.23197641 | read here |
| Weakly Discriminative, Not Repaired by Profile Injection, and Manipulable. Self-confidence in LLM evaluation | TAE (Trust-AI-Eval) | 10.5281/zenodo.23197651 | read here |
SSRN is closed to AI crawlers by policy, for the whole site. Its robots.txt, retrieved on
19 September 2026, reads in full:
User-agent: GPTBot
Disallow: /
User-agent: ChatGPT-User
Disallow: /
User-agent: Google-Extended
Disallow: /
The first two are crawlers: GPTBot collects training data, and ChatGPT-User is the fetcher that
runs when a person asks that assistant to open a link. The third is not a crawler at all.
Google-Extended is a control token with no user agent of its own; by Google's documentation it
governs whether content Google has already crawled may be used to train Gemini or to ground its
answers, and it does not affect a site's inclusion in Google Search. So the three lines do two
different things: they refuse two crawlers outright, and they withhold the site from one assistant's
training and grounding.
For the first of the three the refusal is enforced rather than merely declared. A request carrying
the user agent GPTBot/1.0 returns HTTP 403 where a generic client user agent returns 200, both
tested on 19 September 2026 from one network path. The same request against this repository returns
200. That is a difference between two requests, not a proof that every automated client is turned
away; a search-based tool did retrieve the abstract page during the same test, because the page is
present in conventional search indexes.
The narrower and more durable point stands: a person can open a paper on SSRN, and a crawler acting for that person is told not to. This repository, the Zenodo deposits and the HTML pages exist so that the text is reachable by both. Nothing here is a separate publication. Cite the SSRN version of record.
Each folder holds paper.md (full text, with figures and tables) and the PDF as submitted.
- after-chat: teaching code for After Chat (Appendix C): one question to several AI agent command-line tools, who said what, how much agreement is worth, crossing records, safe agent settings: https://github.com/vadimchernets/after-chat
- delegation-horizon: measurement code for After the Programmer Divide (Appendix C): the verified delegation horizon, the record card, Brier scoring of the registered predictions: https://github.com/vadimchernets/delegation-horizon
- docprep: runnable reference implementation of the architecture in The Missing Variable in AI-Assisted Litigation: https://github.com/vadimchernets/docprep
- Sidecall and Roundcall: Claude Code plugins built on the patterns of After Chat: https://github.com/vadimchernets/sidecall, https://github.com/vadimchernets/roundcall
- ssrn-benchmark: reference distribution of SSRN download counts from 1,360 archived papers, with a command-line tool: https://github.com/vadimchernets/ssrn-benchmark
- Dataset: 10.5281/zenodo.22285871, CC BY 4.0
Text and figures: CC BY 4.0, matching the Zenodo deposits. The SSRN copy is governed by the licence granted there.