# jaredpalmer/kev

> kev is a LoRA adapter with a small readout head on top of Qwen2.5-0.5B that answers many typed questions about a document in a single forward pass, returning calibrated probabilities instead of text.

- Magnitude: 2.2 out of 10 — Steady
- Stars: 9,789 total · +18 stars measured 2026-11-03, 00:37–13:54 UTC, ≈ 30 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Python · License: Apache-2.0 · Created: 2026-09-17 · Last push: 2026-11-03
- GitHub: https://github.com/jaredpalmer/kev · Page: https://gitnova.dev/en/r/jaredpalmer/kev
- GitNova flagged it 2 days before it took off on Hacker News.

## Useful for

- Route support tickets by department, urgency and customer tone in one request
- Point typesafe-sdk at a local kev server by changing base_url
- Train and run a document classification model on a MacBook without a GPU

## Why it’s here

- Star-counter measurements on 2026-11-03 (UTC), 00:37–13:54: 9771 → 9789 stars (+18). This is the change over that interval.
- Estimated end-of-day forecast: about +30 stars, using observed gains and the previous day.
- The repository is 47 days old and already has 9,789 stars.
- Top new repositories in the last 4 months: #70.
- Recent forks include notable developers: @dexhunter (404 followers).

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 670
- Issues and pull requests: 276
- Watchers: 33
- Average over the last week: 27 per day
- Usual pace: 67 per day
- Stars in the last hour (measured): -1
- Latest release: kev-1.0 (2026-10-01)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-10-05 … 2026-11-03: 77, 100, 62, 85, 97, 58, 50, 62, 70, 57, 54, 53, 26, 26, 39, 56, 39, 39, 22, 29, 28, 47, 53, 37, 38, 25, 17, 19, 25, 18

## Magnitude by day, last 14 days

10-21 2.5, 10-22 2.4, 10-23 2.2, 10-24 2.0, 10-25 2.0, 10-26 2.3, 10-27 2.8, 10-28 2.9, 10-29 2.6, 10-30 2.5, 10-31 2.0, 11-01 1.8, 11-02 1.9, 11-03 2.2

## Hacker News

- Kev: Tiny Jev-like family of decision models built on top of Qwen3.5 — 462 points, 205 comments: https://news.ycombinator.com/item?id=49783999

## Spotted in now

- Top new repositories in the last 4 months: #70

## Similar by description

1. **deepopen-com/deepopen** — 0.0 · Cooling · Language models · Python · Not enough measurements today · unusual star pattern (heuristic)
   Open-source non-autoregressive System 1 decision engine for typed decisions: returns typed answers (choice, score, noul) in a single forward pass across 100+ languages without generating text. Suited for classification, routing and…
   Full card: https://gitnova.dev/en/r/deepopen-com/deepopen.md
2. **NandhaKishorM/laya** — 2.7 · Steady · AI agents · Python · +38 stars measured 2026-11-03, 00:34–13:52 UTC, ≈ 67 by evening
   Laya is a non-autoregressive decision engine that answers typed questions (choice, score, noul) over any state (text, email, ticket, JSON) in a single forward pass (~35 ms on GPU) with calibrated probabilities.
   Full card: https://gitnova.dev/en/r/NandhaKishorM/laya.md
3. **spinfinc/spinf-serve** — 0.8 · Steady · Language models · Python · +0 stars measured 2026-11-03, 00:41–14:02 UTC, ≈ 0 by evening
   A serving engine for bulk classification and decision inference on a single 24 GB GPU: it reads content once and scores many questions about it in the same pass, exposing an OpenAI-compatible API. Built for high-throughput calibrated…
   Full card: https://gitnova.dev/en/r/spinfinc/spinf-serve.md
4. **mizorewww/laya-mlx** — 0.7 · Steady · Language models · Python · +0 stars measured 2026-11-03, 00:40–14:02 UTC, ≈ 0 by evening
   Native MLX runtime for Laya typed decision models: returns probabilities for choices, scores or truth values without text generation, running locally on Apple Silicon.
   Full card: https://gitnova.dev/en/r/mizorewww/laya-mlx.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-11-03 14:05 UTC, updated every 30 minutes.