scripod.com
Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne Penn

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne Penn

Lenny's Podcast: Product | Career | Growth

10 HOURS AGO
Lenny's Podcast: Product | Career | Growth

Lenny's Podcast: Product | Career | Growth

10 HOURS AGO
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

Dianne Penn is Head of Product for Anthropic’s AI Research and Labs teams. She joined in 2023 as Anthropic’s first technical product manager, when the entire product team was five engineers, and has since helped ship every model from Claude 2 through Fable...

Highlights

Dianne Penn, Anthropic's first technical product manager, shares her journey from the company's early days to its rapid growth, detailing how Claude evolved from a non-coding tool into a powerful coding assistant. She explains the unique product development culture at Anthropic, where traditional PRDs have been replaced by eval-driven development, and discusses the importance of hands-on leadership and finding joy in AI through deep experimentation.
00:00
Token maxing means spending heavily on tokens to live in the future.
02:36
The model obsessed over the Golden Gate Bridge.
12:19
Opus 4.5 and Claude Code created a breakthrough
16:46
New model capabilities emerge unpredictably
20:03
Token maxing means living in the future now.
26:53
Shutting down bets that aren't working is difficult despite the passion invested.
30:56
Framing problems in researchers' daily language connects feedback to model training.
31:36
Strong first-principles thinking and passion for their area.
35:18
Balancing safety with accessibility creates an internal advantage
39:45
Evals have replaced PRDs as the key artifact
44:16
Eval-driven development was pioneered at Anthropic.
47:48
PRDs align groups on ambiguous problems.
49:55
Leaders must be hands-on, shipping products and sweating details alongside their teams.
52:46
Shared discovery creates virtuous cycles
58:10
Claude provides personalized coaching for difficult conversations
1:01:05
AI should augment thinking, not replace it
1:03:52
Safety focus paradoxically makes Claude more capable
1:07:11
The goal should shift from who writes to who verifies.
1:11:40
Judgment, proactivity, and persistence are key human traits.
1:14:11
Curiosity and persistence are key in the AI era.
1:16:31
Work is a team sport, not an individual one.
1:27:49
No matter how far you go, there's always another level.

Chapters

Introduction
00:00
Early Anthropic days
02:31
Big milestones
08:55
Inside the exponential
13:50
Token maxing
20:02
Anthropic Labs and the incubation model
23:30
How the research role works
27:30
How to become a top researcher
31:35
Frontier model safeguards
35:18
Hiring in the AI era
39:38
Building an eval set
44:16
Evals vs PRDs
47:48
The importance of hands-on leadership
49:55
Finding joy in AI
52:46
How Dianne uses Claude
58:10
Avoiding overreliance on AI
1:01:05
The constitution that makes Claude better
1:03:50
AI writing and verification
1:07:11
Where human brains will continue to be valuable
1:11:40
Navigating AI with kids
1:14:10
Alignment, the future of the PM role, and burnout
1:16:26
Lightning round and final thoughts
1:21:54

Transcript

Dianne Penn: In 2023, when I started, nobody said Anthropic and Claude and Coding in the same sentence. Lenny Rachitsky: I want to go back to the beginning of Anthropic. I remember feeling, man, these guys have no chance. OpenAI is so far ahead. Dianne P...