Articles by galsapir
1

A bitter lesson for medicine, or a benchmark problem? (sparsethought.com)

40

Can LLMs Beat Classical Hyperparameter Optimization Algorithms? (arxiv.org)

2

Gemma 4 E4B as a primary local LLM (replaced Qwen) (digg.com)

3

PEEK: Give Your Agent an Orientation Cache (MIT CSAIL, Khattab group) (zhuohangu.github.io)

1

Hyperagents (Meta Research) (arxiv.org)

2

The Unreasonable Effectiveness of HTML (claude.com)

2

The Comparator in Clinical AI (sparsethought.com)

13

Borges' cartographers and the tacit skill of reading LM output (galsapir.github.io)

2

Best read of 2026 so far was written in 1880 (galsapir.github.io)

1

Anthropic launched community ambassador program (claude.com)

1

LLMs as nudging research towards luke-warm middle (nature.com)

2

How do you evaluate a foundation model before you know what it's for? (galsapir.github.io)

1

Ask HN: Anyone using Claude Agent SDK in production?

4

Data Activation Thoughts (galsapir.github.io)