# RikiyaT/lm_understand-logs

> Complete logs of every experiment from the blog post 'LLMs can understand and write neural network weights', including prompts, model replies, sandboxed code runs, evaluations and the networks read or written.

- Magnitude: 6.8 out of 10 — Early signal
- Stars: 260 total · +81 stars measured 2026-11-03, 00:31–02:52 UTC, ≈ 354 by evening
- Star trust: star growth looks organic
- Category: Science & research · Language: Python · Created: 2026-10-29 · Last push: 2026-10-29
- GitHub: https://github.com/RikiyaT/lm_understand-logs · Page: https://gitnova.dev/en/r/RikiyaT/lm_understand-logs

## Useful for

- Reproduce an experiment from the post using the recorded prompts and run.json settings
- Inspect the model's reasoning summaries and sandboxed code output for a specific turn
- Reuse the evaluation scripts and answer keys to score weight-space tasks

## Why it’s here

- Star-counter measurements on 2026-11-03 (UTC), 00:31–02:52: 179 → 260 stars (+81). This is the change over that interval.
- Estimated end-of-day forecast: about +354 stars, using observed gains and the previous day.
- The repository is 5 days old and already has 260 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #106.
- About 101 forks a day — people are taking the code.

## 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: 29
- Issues and pull requests: 2
- Watchers: 1
- Average over the last week: 102 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 44

## Stars per day, last 6 days (oldest → newest, today is partial)

2026-10-29 … 2026-11-03: 0, 0, 0, 0, 260, 81

## Spotted in now

- Top new repositories this week: #106

## More in this category

1. **kornia/kornia** — 6.2 · Breakout · Science & research · Python · +34 stars measured 2026-11-03, 00:31–02:41 UTC, ≈ 247 by evening
   Kornia is a differentiable computer vision and geometry library built on PyTorch: over 1,200 GPU-accelerated operators for image processing, feature detection/matching, minimal solvers and camera models, all keeping gradients. It is aimed…
   Full card: https://gitnova.dev/en/r/kornia/kornia.md
2. **openai/math** — 2.3 · Cooling · Science & research · Lean · +4 stars measured 2026-11-03, 00:34–02:45 UTC, ≈ 25 by evening
   A collection of mathematical manuscripts and Lean formalizations produced by an internal OpenAI model while solving open research problems. Intended for mathematicians and researchers studying the results and verifying proofs.
   Full card: https://gitnova.dev/en/r/openai/math.md
3. **da-luce/astroterm** — 2.1 · Cooling · Science & research · C · +1 star measured 2026-11-03, 00:35–02:45 UTC, ≈ 10 by evening
   A terminal planetarium written in C: renders real-time positions of stars, planets, constellations and Moon phases as ASCII art right in the command line. Configurable by date, time and observer coordinates.
   Full card: https://gitnova.dev/en/r/da-luce/astroterm.md
4. **CuboiLeo/SuperMimic-Reproduced** — 1.7 · Steady · Science & research · Python · +5 stars measured 2026-11-03, 00:38–02:48 UTC, ≈ 20 by evening
   From-scratch reproduction of the SuperMimic paper (arXiv:2610.28480): a differentiable world model plus short-horizon BPTT actor-critic for Unitree G1 whole-body motion tracking, built on mjlab/MuJoCo Warp, with data pipeline, training,…
   Full card: https://gitnova.dev/en/r/CuboiLeo/SuperMimic-Reproduced.md
5. **google/or-tools** — 1.7 · Steady · Science & research · C++ · +1 star measured 2026-11-03, 00:39–02:47 UTC, ≈ 8 by evening
   Open-source suite by Google for solving combinatorial optimization problems: constraint programming, linear programming, routing and graph algorithms. Includes C++ core with Python, C# and Java wrappers.
   Full card: https://gitnova.dev/en/r/google/or-tools.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 02:52 UTC, updated every 30 minutes.