Why AI is going vertical (again) | Dianne Penn (Anthropic)

Why AI is going vertical (again) | Dianne Penn (Anthropic)

The Early Days of Anthropic

Initial Impressions and Challenges

  • In 2023, the speaker reflects on the early perception of Anthropic, believing they had no chance against OpenAI's advancements.
  • Observations noted that models were evolving from simple code autocomplete to writing long-form code, indicating a significant opportunity for improvement.

Key Product Developments

  • The launch of Opus 45 marked a pivotal moment, combining model capabilities with an effective product experience like Cloud Code.
  • Emphasis on user feedback has shifted; evaluations are now considered as crucial as traditional product requirement documents (PRDs).

The Role of AI in Product Development

User Value and Ambition

  • There is a call for greater ambition in utilizing AI tools due to their increasing capabilities; teams are encouraged to think about future iterations like Claude 8.

Diane Penn's Background

  • Diane Penn, head of product at Anthropic, shares her journey from being one of five engineers to leading significant projects including Claude and Cloud Code.

Reflections on Growth and Culture

Company Culture in Early Days

  • The strong culture at Anthropic during its inception was vital for identity formation and technological value delivery to users.

Notable Projects and Innovations

  • A quirky project called Golden Gate Claude showcased the company's ability to rapidly develop unique user experiences based on research insights.

Inflection Points in Development

Major Milestones

  • Training Opus 3 was identified as a critical moment when the company recognized the need for creating frontier models that could attract users.

Evolution of Coding Capabilities

  • Initially overlooked, coding became a focal point for differentiation as users began leveraging models for more complex coding tasks.

Understanding Exponential Growth in AI

Current Trends and Predictions

  • Discussion around being inside an exponential growth curve highlights how rapid improvements can lead to significant jumps in capabilities.

Preparing for Future Changes

  • Adaptability is emphasized as essential; organizations must be agile in decision-making amidst fast-paced developments in AI technology.

Experimentation and Discovery

Importance of Experimentation

  • Spending on tokens is framed not just as input but as part of an experimentation strategy that drives innovation within products.

Collaborative Discovery Process

  • Internal collaboration fosters communal discovery where team members share findings from using models, enhancing overall understanding and application.

Insights into Labs at Anthropic

Purpose and Functionality

  • Labs focus on exploring large bets outside core roadmaps, aiming for innovative breakthroughs through small teams dedicated to zero-to-one experimentation.

Team Dynamics

  • A culture that encourages risk-taking while accepting failures allows labs to thrive within the larger organization by fostering creativity.

Understanding AI Research and Development

The Research Process

  • Researchers hypothesize improvements for AI models, find relevant data, tweak algorithms, adjust training methods, and test outcomes to iteratively enhance the model.
  • At organizations like Anthropic, researchers also focus on a broader vision for the future of AI technology alongside immediate improvements.
  • There is a dual focus on long-term aspirations for AI capabilities and short-term enhancements based on user feedback and current technology performance.

User Impact and Feedback Integration

  • Teams work closely with researchers to ensure that user feedback directly informs actionable improvements in areas such as coding and tool use.
  • Vague user feedback (e.g., "Claude hallucinated") poses challenges; teams must analyze the context to provide specific, actionable insights to researchers.

Analyzing User Feedback

  • Detailed analysis of user interactions helps identify whether issues stem from tool usage failures or knowledge synthesis problems.
  • Bringing detailed insights into discussions with researchers allows for better evaluation of problems and improvement strategies.

Success Factors for AI Researchers

Key Traits of Successful Researchers

  • Successful researchers are strong first-principles thinkers who are passionate about their field and maintain close attention to detail in their work.
  • Leadership at Anthropic emphasizes being hands-on with training runs and understanding underlying data to inform research directions.

Ambition in Research

  • A bold vision is crucial; successful researchers aim high while remaining grounded in practical details necessary for achieving those ambitions.
  • The rapid evolution of technology necessitates forward-thinking approaches that align product development with future capabilities.

Evolving Landscape of AI Models

Increased Scrutiny on Advanced Models

  • Recent developments have led to heightened scrutiny over new models due to concerns about their capabilities, impacting how companies operate internally.
  • As models become more capable, pre-release testing processes must evolve quickly to ensure safety without compromising user experience.

Balancing Innovation with Safety

  • Companies are developing safeguards while striving to deliver excellent user experiences; this balance is critical as technologies advance rapidly.

Changing Roles in Product Management

Skills Required for Modern PM Roles

  • Current hiring practices emphasize first-principles thinking over traditional pattern matching; PM roles now require deeper understanding of user value through direct engagement with technology.

New Evaluation Methods

  • Evaluations (evals), rather than traditional PRDs (Product Requirement Documents), have become essential tools for defining user needs based on real-time feedback.

The Importance of Hands-On Experience

Managerial Involvement in Product Development

  • Effective managers need hands-on experience with products; they should engage directly with users' needs by participating actively in product development cycles.

Fostering Joy Through Collaboration

  • Encouraging collaboration among team members can help foster enthusiasm around experimentation with AI technologies, making the process enjoyable rather than burdensome.

Exploring the Depth of AI Technology

The Challenge of Navigating AI Innovations

  • The speaker discusses the difficulty in choosing among numerous AI technologies, emphasizing the importance of focusing on a few that bring immediate value.
  • A survey revealed that happiness in tech comes from deep engagement with select tools rather than superficial interactions with many.
  • Finding effective solutions through AI can lead to greater satisfaction compared to dealing with multiple ineffective tools.

Prioritizing Joy and Experimentation

  • The culture at Anthropic encourages public experimentation, which helps teams understand user needs and improve product development.
  • The speaker highlights using Claude (an AI tool) for enhancing productivity and improving team communication as a project manager.

Enhancing Conversations Through AI

  • Claude is utilized not just for technical tasks but also for preparing better conversations, aiding managers in crucial discussions.
  • A skill developed from "Crucial Conversations" assists in determining appropriate levels of detail during discussions, enhancing managerial effectiveness.

Leveraging AI as a Thought Partner

Using Claude for Personal Development

  • The speaker describes how they use Claude to prepare for difficult conversations by seeking advice on approach strategies.
  • This interaction serves as personalized coaching, helping refine thoughts and reactions before engaging with colleagues.

Concerns About Over-Reliance on AI

  • There are concerns about people adopting an overly formulaic way of speaking due to frequent interactions with AI-generated content.
  • The speaker acknowledges these fears but emphasizes maintaining personal judgment while collaborating with Claude.

Balancing Human Judgment and AI Assistance

Maintaining Individual Thought Processes

  • It's important to retain one's own point of view when working alongside Claude; this ensures personal tone and judgment remain intact.
  • For routine tasks like monthly business reviews, delegating writing to Claude allows more focus on critical thinking rather than execution.

Proactivity in Collaboration

  • Effective collaboration involves using Claude not just as a task executor but as a partner that challenges ideas and fosters deeper thinking.

Understanding the Role of Safety and Alignment

Enhancing Interaction Quality

  • Safety measures within Claude's design encourage constructive feedback rather than mere agreement, leading to improved outcomes.

Pricing Decisions Informed by AI Insights

  • Utilizing research versions of models like Opus helps inform strategic decisions such as pricing based on data-driven insights.

Future Directions for Writing Capabilities

Improving Writing Skills in AI Models

  • Ongoing efforts aim to enhance writing capabilities within models like Claude, addressing current limitations observed in generated content quality.

Verifiability Over Authorship

  • Emphasis is placed on who verifies information rather than who writes it; ensuring accuracy is paramount regardless of authorship.

Valuing Human Traits Amidst Technological Advancement

Importance of Human Judgment

  • As technology evolves towards superintelligence, human traits such as judgment will remain invaluable due to their nuanced nature.

Encouraging Curiosity in Children

  • Instilling curiosity and persistence in children is essential; fostering their individual voices prepares them for future challenges.

Avoiding Burnout Through Team Support

  • Collaboration within teams alleviates pressure during high-stakes projects; shared responsibilities foster resilience against burnout.

Building Low-Ego Teams

  • Hiring individuals focused on team success over personal ego contributes significantly to sustainable work environments.

The Role of Product Managers in Technology-Driven Environments

Importance of User-Centric Approaches

  • The discussion emphasizes the necessity for product managers (PMs) to focus on understanding user needs and translating them into actionable insights.
  • There is a call for more user-centric individuals who can delve deeply into what users aim to achieve, highlighting curiosity and hands-on involvement as essential traits.
  • PMs are seen as crucial in determining what should be built, ensuring that products meet user expectations and are worth pursuing.

Empathy and Detail Orientation

  • Empathy is highlighted as a critical trait for PMs, enabling them to connect with users and understand their experiences better.
  • The conversation transitions into a lightning round format, indicating a shift towards personal recommendations and insights.

Recommended Books for Personal Development

Key Book Recommendations

  • "How to Raise an Adult" is recommended for its perspective on parenting aimed at fostering adult characteristics in children.
  • "Incorruptible" by Eric Ries is noted for its exploration of building sustainable companies through metrics beyond just revenue.

Insights from Recent Media Consumption

Favorite Shows and Products

  • The speaker enjoyed binge-watching "Fallout" on Amazon Prime, appreciating its humor and action-oriented narrative.
  • Claw Tag is mentioned as an interesting product experience that has garnered positive feedback within Anthropic.

Life Lessons from Personal Background

Influential Life Motto

  • A life motto shared by the speaker reflects the idea that there’s always another level to strive for, instilling a mindset of continuous improvement.

Experiences from Early Career at JP Morgan Chase

Learning Outcomes from Trading Experience

  • Working on the trading floor at JP Morgan taught valuable lessons about authenticity and conviction in ideas regardless of background or position.
  • Emphasizing vulnerability in leadership, the speaker aims to create an environment where all team members feel empowered to share their ideas.

Engaging with Feedback and Team Growth

Call to Action for Audience Engagement

  • Listeners are encouraged to provide feedback on products like Claude, which helps improve offerings based on user experiences.
  • There’s an open invitation for potential candidates who resonate with the company culture focused on curiosity and innovation.

Turn any video into a summary like this

YouTube links, meetings, lectures — with transcripts, search, and chat.

Video description

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, and helped incubate Claude Code, MCP, Skills, computer use, tool use, and reasoning. Before Anthropic, she helped build Alexa’s AI at Amazon and, before that, traded high-yield bonds at JP Morgan Chase. *In our in-depth conversation, we discuss:* 1. What Anthropic’s early days were like 2. The inflection points that turned Anthropic from an underdog into the fastest-growing company in history 3. How exactly Claude got so good at coding 4. The eval-driven development loop her team is pioneering 5. How to find joy in AI when everything is moving this fast 6. Why Claude’s willingness to push back is key to its success 7. Where human judgment remains irreplaceable *Brought to you by:* WorkOS—Make your app enterprise-ready, with SSO, SCIM, RBAC, and more: https://workos.com/lenny Mercury—Radically different banking, now with Command: https://mercury.com/ *Episode transcript:* https://www.lennysnewsletter.com/p/anthropics-first-technical-pm-on *Archive of all Lenny's Podcast transcripts:* https://www.dropbox.com/scl/fo/yxi4s2w998p1gvtpu4193/AMdNPR8AOw0lMklwtnC0TrQ?rlkey=j06x0nipoti519e0xgm23zsn9&st=ahz0fj11&dl=0 *Where to find Dianne Penn:* • LinkedIn: https://www.linkedin.com/in/dianne-na-penn Where to find Lenny:* • Newsletter: https://www.lennysnewsletter.com • X: https://twitter.com/lennysan • LinkedIn: https://www.linkedin.com/in/lennyrachitsky/ *In this episode, we cover:* (00:00) Introduction (02:31) Early Anthropic days (08:55) Big milestones (13:50) Inside the exponential (20:02) Token maxing (23:30) Anthropic Labs and the incubation model (27:30) How the research role works (31:35) How to become a top researcher (35:18) Frontier model safeguards (39:38) Hiring in the AI era (44:16) Building an eval set (47:48) Evals vs PRDs (49:55) The importance of hands-on leadership (52:46) Finding joy in AI (58:10) How Dianne uses Claude (01:01:05) Avoiding overreliance on AI (01:03:50) The constitution that makes Claude better (01:07:11) AI writing and verification (01:11:40) Where human brains will continue to be valuable (01:14:10) Navigating AI with kids (01:16:26) Alignment, the future of the PM role, and burnout (01:21:54) Lightning round and final thoughts *Referenced:* • Anthropic: https://www.anthropic.com • Golden Gate Claude: https://www.anthropic.com/news/golden-gate-claude • Dario Amodei’s website: https://darioamodei.com • Scaling Laws and Interpretability of Learning from Repeated Data: https://www.anthropic.com/research/scaling-laws-and-interpretability-of-learning-from-repeated-data • Tokenmaxxing: How Top Builders Use AI To Do The Work Of 400 Engineers: https://www.ycombinator.com/library/Pa-tokenmaxxing-how-top-builders-use-ai-to-do-the-work-of-400-engineers • Garry Tan on X: https://x.com/garrytan • Anthropic co-founder on quitting OpenAI, AGI predictions, $100M talent wars, 20% unemployment, and the nightmare scenarios keeping him up at night | Ben Mann: https://www.lennysnewsletter.com/p/anthropic-co-founder-benjamin-mann • Anthropic’s CPO on what comes next | Mike Krieger (co-founder of Instagram): https://www.lennysnewsletter.com/p/anthropics-cpo-heres-what-comes-next • Introducing Labs: https://www.anthropic.com/news/introducing-anthropic-labs • Louis CK | about airplane Wi Fi: https://www.youtube.com/watch?v=me4BZBsHwZs • What happens after coding is solved? | Fiona Fung (Manager of the Claude Code and Cowork Teams): https://www.lennysnewsletter.com/p/building-the-most-ai-pilled-engineering • The Anthropic Hive Mind: https://steve-yegge.medium.com/the-anthropic-hive-mind-d01f768f3d7b • How to build a company that withstands any era | Eric Ries, Lean Startup author: https://www.lennysnewsletter.com/p/how-to-build-a-company-that-withstands • Fallout on Prime Video: https://www.amazon.com/dp/B0CN4GGGQ2 • Fallout (video game): https://fallout.bethesda.net • Claude Tag: https://www.anthropic.com/news/introducing-claude-tag *Recommended books:* • Crucial Conversations: Tools for Talking When Stakes Are High: https://www.amazon.com/dp/0071771328 • How to Raise an Adult: Break Free of the Overparenting Trap and Prepare Your Kid for Success: https://www.amazon.com/How-Raise-Adult-Overparenting-Prepare/dp/1627791779 • Incorruptible: Why Good Companies Go Bad... and How Great Companies Stay Great: https://www.amazon.com/dp/B0FWZZBPZB _Production and marketing by https://penname.co/._ _For inquiries about sponsoring the podcast, email podcast@lennyrachitsky.com._ Lenny may be an investor in the companies discussed.