feat: upload blog diagram diagram_curriculum

feat: upload blog diagram diagram_training_loop

feat: upload blog diagram diagram1_timeline

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

chore: delete colab/ folder (superseded by eval/)

fix(links): restore colab link to eval/arm_gym_grpo_colab.ipynb

feat: upload eval/arm_gym_grpo_colab.ipynb

fix(readme): colab notebook link

fix(links): colab notebook path colab/arm_gym_grpo_colab.ipynb

fix(readme): plot URLs point to eval/plots

fix(ui): plot URLs now point to eval/plots (correct images uploaded)

fix(plots): upload eval/plots/before_after_kernel.png

fix(plots): upload eval/plots/training_loss.png

fix(plots): upload eval/plots/correctness_rate.png

fix(plots): upload eval/plots/reward_curve.png

fix(ui): correct plot URLs — point to colab/results/plots which actually exist

docs(readme): full story with training curves, research lineage, how it works

docs(readme): remove em dashes, clean prose

feat(ui): restore Space landing page (from eval/space-src)

fix(docker): install full aarch64 cross toolchain in final image

fix(reset): copy gcc-cross headers in Dockerfile, add proper error handling to reset/step

feat(blog): embed mermaid PNGs for timeline, training loop, curriculum

fix(ui): rewrite why-cards, metric strip, hero sub, remove em dashes from captions

fix(hero): remove tag pills from hero section

feat(readme+blog): narrative rewrite, em dash cleanup, remove hero badge

feat(blog): full narrative writeup — research lineage, philosophy, honest results, future roadmap

fix(space): vertically center figcaption text in plot cards

fix(ui): plots uncropped — object-fit:contain, height:auto, no forced aspect-ratio

feat(ui): V11 frontend — real plots, updated metrics/diagram, model links

feat(notebook): storytelling layer — origin story, research lineage, honest speedup framing, future vision

feat: add arm_gym_grpo_colab.ipynb — V11 primary, V10 comparison, 5 plots pre-embedded

docs: restore flowchart LR layout with br breaks and full Mermaid styles

docs: fix Mermaid diagrams for HF/GitHub (graph LR, ASCII labels)

chore: drop .gitattributes LFS rule (plots stay normal blobs on GitHub)

docs: blog post; HF sync workflow; kaggle prompt/reward + tests

feat(hf): add v7_train.py — tuned lr, beta, epsilon, completion, grad_accum

docs: record MVP50 job ids and curl fix

fix(hf): MVP job launcher without curl; add analysis/hf-mvp-50-runs docs

chore(hf): add job_run_mvp50_github_v6.sh for Jobs (MVP + main v6)

chore(hf): add publish_v6_to_dataset.sh for arm-gym-pkg sync

feat(hf): export manifest, MVP save@50, long checkpoints@50-200

chore(hf): set num_generations to 4 to avoid OOM on A10G-class GPUs