scripod.com
Inside Anthropic: How Claude Actually Gets Built | Alex Albert

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Inside Anthropic: How Claude Actually Gets Built | Alex Albert

Behind the Craft

3 DAYS AGO
Behind the Craft

Behind the Craft

3 DAYS AGO
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

Alex is a research PM at Anthropic building the next Claude model. He gave me a rare inside look at how the research team operates — including how they turn user feedback into model training, how they decide which capabilities to prioritize, and how they i...

Highlights

In this podcast, Alex Albert, a research PM at Anthropic, provides an inside look at how the team develops the next Claude model, covering everything from product development and user feedback integration to character training and consciousness research.
00:00
Claude models are developed like products
04:58
Context about the user influences whether Claude should think hard
09:41
AI tools drastically reduced the cost and time to prototype.
11:15
One-way doors require more time and thought.
15:40
AI challenges assumptions to sharpen decisions
17:05
Creating test cases to identify weaknesses
21:02
Character is both quantifiable metrics and human intuition.
32:46
Research improves trustworthiness for autonomous tasks

Chapters

How Anthropic treats each new model as a product
00:00
Building adaptive thinking into Claude
04:58
Why Claude is starting to dream
07:07
"If it's not a one-way door, then it's essentially free"
11:15
How Alex uses Claude Cowork to pressure test his docs
15:40
Inside Anthropic's eval process for new models
17:05
How Anthropic trains Claude's character
21:02
The consciousness question Anthropic is quietly working on
32:39

Transcript

Alex Albert: I was definitely the first prompt engineer at Anthropic. Might have been the first in the world. We treat the model as if it's a product to some degree. With every new model, we are speccing out exactly what do we want this model to be good at...