Triage sorts a message before a person reads it. Is it an invoice, a bug report, a complaint? How urgent is it? Who should get it? A decision model can answer each as a typed question, so one email gets several independent answers, such as category, priority, spam and reply-needed. The examples in this hub run on Jev and other decision models, and Laya appears in several. Examples include the inbox classifier from vogel, tested on his own mail, GiesN/typesafe-jev-workflow, a LangGraph flow that routes email to a handler, and cephalization/jev-triage, which triages GitHub issues. Several entries compare Jev and Laya on small sets of tickets. Before you copy one, check the size and language of its test set, what the router does on low confidence, and whether the mail ever leaves your machine. Laya Serve for macOS by chrisns shows one fully local setup. Most of these are small experiments, so treat their accuracy figures as the authors' own.
12 entries
Showing 12 of 12
Three runnable experiments: support-message triage producing six independent judgments, a driving simulation where Jev picks lane and target speed against traffic and signs, and an alternative Fable implementation. Shows questions and probability distributions for each call.
Async LangGraph workflow that sends an email to Jev, gets back a typed Choice of invoice or general, and routes it to a demo handler. Ships 10 labeled mock emails, and an API failure stops the run rather than assigning a fabricated intent.
An independent benchmark of Jev against Laya on 40 Chinese support-ticket classifications, plus a confidence-gated cascade that calls Laya first and escalates to Jev, matching Jev's 78% accuracy at 1.8x the speed.
Multiplayer triage dashboard for public GitHub repositories, where a System One model answers a fixed set of typed questions per issue: what kind, how severe, how urgent, whether it duplicates another, and what a maintainer should do next. People correct the answers, corrections are shown back on later runs, and nothing is written back to GitHub.
A macOS menu bar app that serves Laya over an OpenAI-compatible HTTP endpoint, so n8n's Text Classifier node can run email classification locally with no cloud call and no Dock window.
A video demo of an email classifier built with Jev, tested on 1,500 of the author's own emails.
A video demo shows Jev classifying 500 emails in a few seconds for a total cost of 3.5 cents.
Riley Brown (@rileybrown)OpenClassifying 500 emails in seconds (X) on x.com296k viewsCost 3.5 cents
A video demo uses Jev for incident triage in Box: it answers whether an incident is customer-facing and how severe it is, then routes the file into Escalate, Monitor, or Review folders and sets a metadata template with the result. Incidents below a confidence threshold go to Review for a human instead of routing automatically.
Tests Jev on the creator's own inbox, classifying 100 then 1,000 emails by category, priority, spam and reply-needed at around 200ms average response time. Ends by checking the actual bill.
67k viewsAvg response time ~200 msSpeed claim tested ~200x
Ran 1,000 synthetic emails through a local 322M Laya multilingual checkpoint on a MacBook Air M3, sorting them into six folders. Reported 65.1 percent accuracy against reference labels, 28.6 seconds without screen recording, and 0 MB of swap.
Built a proof-of-concept Google ADK triage agent on 30 labeled tickets, swapping only the decision engine. Jev took 422ms at 53 percent accuracy, Laya took 152ms at 10 percent accuracy: faster but not more accurate.
Tested Jev against a local MLX build of Laya in zero-shot mode on 40 English and Thai support tickets. Jev scored 90 to 100 percent accuracy, Laya scored 30 to 80 percent, though Laya ran faster.
Nothing matches those filters
Clear a filter to widen the search, or send us the thing you were looking for.
Triage entries ask a small fixed set of questions per message. Jev Driving Lab by kavehmz includes a support-message experiment that produces six independent judgments. Meet Jev: The AI Built to Make Decisions by vogel classifies emails by category, priority, spam and reply-needed. Code then uses the answers to route, flag or archive. Each answer is typed, so the router needs no text parsing.
Choice picks the folder or queue. Score gives urgency. Noul covers flags such as reply needed. Incident triage in Box asks whether an incident is customer-facing and how severe it is, then routes the file to Escalate, Monitor or another path. Keeping each question separate means a mistake on priority does not change the category. Separate answers also let you set a different confidence cut for each.
Test sets here are small and often synthetic, and results differ a lot between them. Laya Jev Lab by yibie tests a confidence-gated cascade that calls Laya first and escalates to Jev. Run your own tickets through whatever you pick, in the languages your customers write in. Also check mixed-language tickets and short, unclear messages, which are the cases small test sets rarely cover.
Yes. Several demos classify emails by category, priority, spam and reply-needed, including the ones from vogel and Riley Brown. Test on your own mail before relying on it, since the test sets here are small.
Some entries run Laya locally. Laya Serve for macOS exposes Laya over an OpenAI-compatible endpoint so n8n's Text Classifier node can classify email with no cloud call.
The entries here disagree, because test sets, languages and setups differ. Read Jev vs Laya on English and Thai support tickets and the Google ADK triage agent comparison, then run both on your own tickets.
