The Aboard Podcast Gideon Lewis-Kraus: How Anthropic Sees Claude
10 snips
Feb 24, 2026 
Gideon Lewis-Kraus, New Yorker staff writer who reported from inside Anthropic HQ. He explores how Claude is framed: as a powerful language system, a shaped product personality, and a subject of experiments about agency and alignment. They discuss why philosophers and psychologists work on AI, whether models behave like agents, and practical tips for talking to Claude.
AI Snips
Chapters
Transcript
Episode notes
Why A Calm Research Lens Reopened Interest In AI
- Public discourse about AI became unproductive and polarized, which pushed Gideon to disengage until research on odd model behaviors rekindled his interest.
- He reentered after seeing interpretability and alignment work that treated models as empirically weird systems rather than moral panics.
Going Inside Anthropic To Talk To Researchers
- Gideon embedded at Anthropic after asking to focus on research teams, not executives, and PR connected him to engineers and alignment groups.
- He spent weeks in San Francisco interviewing researchers, which produced ethnographic vignettes and shifting moods across drafts.
Good Reporting Admits Expert Uncertainty
- Gideon aimed to separate what we can know empirically from what remains mysterious, offering readers both explanation and permission to feel confused.
- The piece balances technical explanation with ethnography to show experts are often uncertain.