Review economics
FareedKhan-dev/train-llm-from-scratch
Read from the 18 most recently merged pull requests ·
Pull requests with attributed agent authorship took only 0.1× the time to merge of the rest.
≥25%
Attributed
a floor
6
Attributed PRs
12
Other PRs
18
Read
Attributed — the latest 5
#33 docs: rewrite README as an end-to-end tutorial with hand-drawn diagrams
0 reviews · 0h to merge
#32 fix: multi-GPU (DDP) training hangs in SFT, Reward, DPO
0 reviews · 0h to merge
#29 feat: opt-in AMP + gradient checkpointing + gradient accumulation; fix CPU contiguity (fixes #5)
0 reviews · 0h to merge
#28 Harden checkpoint writes and restore Python 3.9 typing compatibility in legacy trainer
1 reviews · 23h to merge
#18 docs CI: deploy via official GitHub Pages Actions (Source=GitHub Acti…
0 reviews · 0h to merge
The rest — the latest 5
#30 Correct equation format in objectives.md
0 reviews · 3.9d to merge
#31 Add link to Pretraining at bottom.
0 reviews · 3.8d to merge
#26 docs: add foundations-first LLM tutorial
0 reviews · 3.7d to merge
#15 Improve modern GPU compatibility and document RTX 5090 validation
0 reviews · 2.5d to merge
#13 fix: add missing output projection in MultiHeadAttention and optimize training
0 reviews · 5.3d to merge
How this was measured
25% of merged commits carry agent attribution — a floor, not the share; tools that only complete code inline leave no commit trail, so the unattributed side includes AI-assisted work.
Detected: Claude Code, GitHub Copilot coding agent.
“Attributed” means a commit carried an agent’s signature — a co-author trailer, an agent commit identity, or an agent bot account. Tools that only complete code inline leave no such mark, so the other column is “rest”, not “human-written”. Full method and its limits
The badge reads this repo’s current report, so it follows the number.
Tell me when this moves
We re-read FareedKhan-dev/train-llm-from-scratch weekly and email only when the number changes materially.
Read your own repos
This one is public. For a private repo, run the same read locally through your own GitHub credentials — nothing is installed and nothing is sent to us.
npx @ambera/review-taxWant this continuously on FareedKhan-dev/train-llm-from-scratch — each pull request paired to the task it came from, and the work graded from its review loop rather than its diff size? Claim this repo in Forge
Browse every repo people have read · Read a different repository