This is your work, valued

New York, NY

brendan hogan

Expert
@brendanhogan

AI/ML Research @ Morgan Stanley PhD Cornell University

loophole. Python

395

hermitclaw. Python

327

DeepSeekRL-Extended. Exploring Applications of GRPO

252

picoDeepResearch. Python

69

2025_advent_of_small_ml. Python

22

completion_tree_view. HTML

15

nano-grpo-envs. Python

10

modded-nanogpt. NanoGPT (124M) quality in 8.2 minutes

4

nbn-gpt. Python

3

forum_chat. Python

3

rainbow_dqn. Python

3

base-model-agents. Python

2

simple_rlm. Python

2

pass-the-turing-test. A reality-TV game show for language models. Six LLM agents each believe they're the only AI in a room of humans. They have to pass for human or get voted off.

2

popcorn. Local tool for processing a reading list into a biweekly post, with Claude-generated summaries, screenshot OCR for tweets, and per-entry chat. Inspired by Karpathy's LLM Wiki idea.

2

nbn-agentic-podcast. Python

1

auto-harness. Python

1

autopcb. The Karpathy ratchet, applied to physical hardware design. Autonomous LLM loop that optimizes PCB component placement.

1