Researcher

Nathan Lambert

Machine learning scientist; author of Interconnects

Evals & benchmarks Claude Code Codex & OpenAI agents

ML research scientist (post-training, RLHF) who writes the Interconnects newsletter. His 2026 posts covered Claude Code, the "post-benchmark era" of Opus vs Codex and where agents go next.

X LinkedIn GitHub YouTube Website

In research (1)

Reasoning Patterns for Efficient AI Coding: An Empirical StudyVerdict: Incremental
Oct 26, 2026 · Writes widely on reasoning models, overthinking and inference-time compute, the topic of this study.

Sources

interconnects.ai/about
github.com/natolambert
interconnects.ai/p/claude-code-hits-different

Verified Oct 29, 2026