deep-learning
an archive of posts in this category
| Jun 07, 2026 | stateful-agent: an agent that actually remembers |
|---|---|
| Jun 07, 2026 | A VLA from scratch: 29% tokens, 0% grasps, and a GRPO that wouldn't budge |
| May 23, 2026 | A haiku VLM: SFT did the work, KTO collapsed at λ=1.0 |
| May 22, 2026 | A 7B math fine-tune on 8× H100: SFT +6.4, DPO +0.6 |
| May 17, 2026 | Eight A100s, $61, and 124M parameters |
| Apr 15, 2026 | Tiny Shakespeare, tiny GPT |
| Apr 08, 2026 | makemore: from counting bigrams to a WaveNet |
| Apr 01, 2026 | micrograd: a scalar-valued autograd engine |
| Mar 01, 2026 | A transformer that reads C++ and writes Python |