# preferencemodel/karotte

> A framework for building isolated RL environments where AI models run tasks inside a VM, with each run's transcript saved for analysis.

- Magnitude: 2.3 out of 10 — Steady
- Stars: 24 total · +15 stars measured 2026-10-07, 18:45–20:26 UTC, ≈ 17 by evening
- Star trust: star growth looks organic
- Category: AI agents · Language: Python · License: MIT · Created: 2026-09-15 · Last push: 2026-10-07
- GitHub: https://github.com/preferencemodel/karotte · Page: https://gitnova.dev/en/r/preferencemodel/karotte

## Useful for

- Create a custom RL environment from a template and run its example task
- Run a task with a Claude model and inspect the transcript in the dashboard
- Add your own tasks to an environment for training aligned AI

## Why it’s here

- Star-counter measurements on 2026-10-07 (UTC), 18:45–20:26: 9 → 24 stars (+15). This is the change over that interval.
- Estimated end-of-day forecast: about +17 stars, using observed gains and the previous day.
- The repository is 22 days old and already has 24 stars.
- Hacker News: “Show HN: Karotte, a framework for building robust RL envs” — 8 points, 2 h ago.

## 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: 2
- Issues and pull requests: 58
- Watchers: 0
- Average over the last week: 3 per day
- Usual pace: 0 per day
- Stars in the last hour (measured): 9

## Stars per day, last 25 days (oldest → newest, today is partial)

2026-09-13 … 2026-10-07: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 15

## Hacker News

- Show HN: Karotte, a framework for building robust RL envs — 8 points, 0 comments: https://news.ycombinator.com/item?id=49996497

## Spotted in now

- Spotted on Hacker News

## More in this category

1. **morluto/rea** — 9.3 · Breakout · AI agents · TypeScript · +4,798 stars measured 2026-10-07, 00:19–20:20 UTC, ≈ 5,565 by evening
   A reverse-engineering tool that lets AI agents inspect apps and native binaries: it connects agents to Hopper or Ghidra via MCP and CLI, helping explain how a feature works and rebuild it in your own project.
   Full card: https://gitnova.dev/en/r/morluto/rea.md
2. **lexmount/moli** — 7.7 · Breakout · AI agents · Rust · +1,253 stars measured 2026-10-07, 00:19–20:21 UTC, ≈ 1,466 by evening
   Moli is a headless browser written in Rust for AI agents, combining a full browser runtime with on-demand layout and rendering to keep resource use low. It fetches and extracts web pages, searches the web, and automates browser tasks via…
   Full card: https://gitnova.dev/en/r/lexmount/moli.md
3. **thedotmack/claude-mem** — 6.9 · Breakout · AI agents · TypeScript · +456 stars measured 2026-10-07, 00:19–20:21 UTC, ≈ 532 by evening
   Persistent memory system for Claude Code and other agents: it captures agent actions via lifecycle hooks, compresses them into semantic summaries, and injects relevant context into future sessions.
   Full card: https://gitnova.dev/en/r/thedotmack/claude-mem.md
4. **odysseus-dev/odysseus** — 6.8 · Breakout · AI agents · Python · +462 stars measured 2026-10-07, 00:19–20:21 UTC, ≈ 547 by evening
   Self-hosted AI workspace combining chat, agents, deep research, documents, email, notes and calendar in one Docker deployment. Aimed at users who want local and API models plus personal data on their own server.
   Full card: https://gitnova.dev/en/r/odysseus-dev/odysseus.md
5. **nykooi1/vibe-wise** — 6.8 · Breakout · AI agents · Python · +378 stars measured 2026-10-07, 00:19–20:21 UTC, ≈ 448 by evening
   A Claude Code plugin that makes the AI ask for your approach first, discuss tradeoffs, and explain unfamiliar concepts, writing code only after your confirmation. For anyone who wants to learn design while AI writes the code.
   Full card: https://gitnova.dev/en/r/nykooi1/vibe-wise.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 2026-10-07 20:36 UTC, updated every 30 minutes.