Article
CommunityJev Practices & Patterns
A Thermometer, Not an Analyst: A Text Judge in a Trading Lab
An 11-page working paper on placing a small commercial text judge beside a live trading lab: six tests for a text signal, what the judge may and may never touch, and the plumbing failures that corrupted its readings. No profitability claims.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Alex Vidovich and Zivana Zerjal
- Last verified
- Oct 2026
Works forJev for Evaluation
Used inJev for Finance & Fintech
Highlights
- The judge is small, commercial, pinned to one version, and answers only typed questions about headlines beside one live account.
- Stated readings so far: one fact read well, one small effect below the bar, several effects that vanished; the margin over a headline count is borderline.
- The paper lists what the judge may touch (stamps written after the close, counted forward) and what it may never touch (any number in the trading system).
- The six tests were assembled over three days, most of them after a failure, and are presented as a checklist to apply up front.
- Working paper, 11 pages, October 9, 2026, doi 10.2139/ssrn.7589261; research and educational content only.
Watch out
Grounded on the paper's abstract and citation block captured from alexvidovich.com on 2026-10-15, with the digest text truncated after the reference block; the page text does not name Jev or TypeSafe, so the Jev connection here is that the item comes from a Jev-focused community list and the paper describes a small commercial judge answering typed questions. The paper makes no claim about either strategy's returns, and neither does this entry.
Questions about this entry
Short answers drawn from the linked project or page.
A text model sitting next to a live trading system the way a thermometer sits next to a patient: it reads, it never prescribes. The authors follow a small commercial judge, pinned to one version, into a lab running an intraday and an overnight strategy on one account, and describe six tests for a text signal.
The captured page text does not mention Jev or TypeSafe. It describes a small commercial judge pinned to one version that answers only typed questions about headlines, which matches the decision-model pattern; the item reaches this directory through a Jev-focused community list, so treat the connection as the list's classification, not the paper's.
Read a plain fact correctly against a real label; beat a simple count of the judge's own headlines; show no memory of famous past events; give the same answer twice; always see the decision-time date; and check each forward verdict rule against random flags. They were assembled over three days, most after a failure.
Readings are small: one fact read well, one small effect below the judge's bar, several effects that vanished, and a margin over a simple headline count that is borderline. The fact was read under one setup, and the authors present the checklist as what they would now apply up front.
It may touch stamps written after the close, counted forward. It may never touch any number in the trading system. The paper also reports the plumbing failures that corrupted its readings and compares the judge with a general chatbot on blind rows.
Alex Vidovich and Zivana Zerjal, both independent researchers, with links to alexvidovich.com and zivanazerjal.com. It is an 11-page working paper dated October 9, 2026, posted on SSRN under doi 10.2139/ssrn.7589261.
No. It states it is a paper about checking data and the instrument, not about profit, and makes no claim about either strategy's returns; the page footer adds that nothing on it reports the performance of any live system. This entry makes no profitability claim either.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: A trading loop with Jev and GPT-6 Astra
X: News sentiment in a trading terminal
X: A Jev trading bot on Telegram
X: A Jev and Grok copy-trading bot
This post couldn't be loaded here — read it on X.https://x.com/Mnilax/status/2102760642872508510More Jev resources like this
A social-scheduler founder's accuracy test of Jev: four yes/no questions over 522 captions, then the same questions for Fable on a random 100. Agreement reached 94/100 on promotion, and all 520 scored captions cost $0.011.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
Unofficial laptop study of parallel constrained decoding: prefill once, share the KV cache across fields, and decide all schema fields in one batched forward pass without writing JSON.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
rorshopping/jev-on-a-laptopPython
rorshopping/jev-on-a-laptopPythonVendor-neutral harness that tests whether Jev reranking improves a small RAG stack, using frozen hybrid BM25 plus BGE-M3 candidates and isolating retrieval, rerank, and generation stages.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
erendikmenn/jev-rag-benchmarkPython
erendikmenn/jev-rag-benchmarkPythonPython browser agent that reads one lean page view per state and answers with a single JSON action, using any OpenAI-compatible model with no decision API or browser daemon.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
idovmamane/dejevuPython
idovmamane/dejevuPythonReproduction kit that rebuilds a frozen typed-decision benchmark suite from public sources and runs any engine through it with resume, scoring, and index computation.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
apolinario/decision-indexPython
apolinario/decision-indexPythonBenchmark harness that tests whether models can forecast federal motion-to-dismiss outcomes from the written record, scored per claim-defendant pair with micro-Brier and clustered intervals.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
johnhughes3/LegalForecastBenchPython
johnhughes3/LegalForecastBenchPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.