The PM Logic Has Fundamentally Changed

Compiled by Peilin

Click to Listen to the Podcast Version

The AI industry is redefining the role of product managers.

Over the past decade, product managers in internet companies have typically focused on writing PRDs, creating roadmaps, coordinating cross-departmental resources, and launching features. The norm has been six-month plans and annual strategies, emphasizing process, collaboration, and stable delivery.

But this logic is rapidly being disrupted in the AI era.

Recently, Cat Wu, Head of Product for Anthropic Claude Code, revealed in a multi-hour public discussion the inner workings of AI-native companies: Features can launch in a week, engineers directly shape product direction, PRDs are significantly streamlined, organizational structures are decentralized, and boundaries between product, engineering, and design are rapidly disappearing.

More importantly, the definition of a good product manager in AI companies has fundamentally changed.

Previously, a top PM’s key strength was coordinating resources, managing timelines, and driving complex organizations to deliver. But now, with AI drastically reducing the cost of coding, the true scarcity lies not in execution capability but in deciding what to build.

This means product managers are not disappearing—they are being thrust back into the core of AI companies. However, this role is no longer the traditional “internet product manager.”

AI Companies Are Phasing Out Traditional PM Logic

Cat Wu noted that many product managers she interviews still approach AI product roles with internet-era logic.

This, she believes, is the biggest issue.

One of the most significant changes in the AI era is the extreme compression of product development cycles.

Traditional internet products typically followed quarterly or even annual plans. Feature development, cross-departmental coordination, testing, and launch each required substantial time, making alignment across teams a top priority for product managers.

AI-native companies operate entirely differently.

Model capabilities shift every few weeks, and user behavior continuously evolves. Many features that were impossible today may become fully viable months later after model upgrades.

The entire industry has entered a state of extreme rapid iteration.

Cat Wu mentioned that at Anthropic, many features go from idea to launch in just a week—or even a day. Numerous features are initially released as “research previews,” allowing users to test them first, with iterations driven by feedback.

This means long-term roadmaps are increasingly irrelevant in AI companies.

The core responsibility of product managers has shifted: Coordination was once paramount; now, the priority is shortening the distance from idea to launch.

She explicitly stated that the most critical skill for an AI product manager is rapidly defining what a product most urgently needs to build.

In the AI era, code itself has become cheaper. What’s truly valuable is judgment.

With tools like Claude Code enabling engineers to write code at high speed, the competitive focus has shifted from “Can it be built?” to “Should it be built?” and “What defines the right product experience?”

Product taste has suddenly become one of the most important skills in the AI industry.

Cat Wu noted that while they receive tens of thousands of GitHub requests, the real challenge isn’t execution but deciding which requests are worth pursuing—and how.

This skill isn’t exclusive to traditional PMs.

At Anthropic, engineers develop product sense, designers participate in implementation, and product managers must understand code. Many engineers can now take user feedback all the way to product launch with minimal PM involvement.

Role boundaries in AI companies are rapidly dissolving.

AI Product Competition Has Become a War of Organizational Speed

Many attribute Anthropic’s rapid growth to its powerful models.

But Cat Wu argues that while models matter, true speed comes from transforming the organization into a high-velocity release machine.

She repeatedly emphasized that one of Anthropic’s top internal goals is eliminating all barriers to launch.

For example, once engineers deem a feature ready, it enters the release process immediately. Documentation, product marketing, and developer relations teams engage instantly, with launches sometimes happening the next day.

The organization’s goal is no longer perfection but reaching users as quickly as possible.

This contrasts sharply with traditional internet companies.

In the past, large companies emphasized process integrity, stability, and consistency, often subjecting features to lengthy approvals. AI companies now accept a new reality: Many products launch imperfectly, with bugs inevitable.

But that’s fine—if core user value is present, the product goes live, and issues are resolved through rapid iteration.

Cat Wu admitted she used to anxiety (anxious) about launching buggy features. Now, she accepts this state, as feedback arrives quickly and fixes follow suit.

What’s truly changing is the industry’s understanding of “product.”

In the internet era, products resembled finished buildings; in the AI era, they resemble ever-evolving organisms.

“Speed” now trumps “consistency.”

Cat Wu noted that Anthropic even intentionally weakens product consistency. Strict uniformity would slow the entire organization.

This is why many AI products feel like they update daily: Features multiply, interfaces shift, and workflows refactor (reconfigure) constantly, driven by rapid model evolution.

Many features originally designed to compensate for model limitations become obsolete with new model releases.

She cited a typical (typical) example.

Early Claude Code included a to-do list feature to prevent model task omissions, having AI complete modifications item by item. Later models automated these steps, rendering the feature unnecessary.

As models grow stronger, products become simpler.

This is one of the AI industry’s most unique aspects: Product managers must continually reinterpret model capabilities, as shifts in these capabilities can overturn entire product logics.

AI Is Breaking Down Role Boundaries

Cat Wu made a crucial observation in the interview: “Roles are somewhat overrated.”

This represents the deepest layer of change in the AI industry.

Traditional large companies had rigid role boundaries: Product managers handled X, designers handled Y, engineers handled Z. Everyone had fixed responsibilities.

AI companies increasingly distrust this model. As model capabilities evolve rapidly, the organization’s greatest enemy becomes “waiting”—waiting for others to finish, for approvals, for cross-team collaboration.

Thus, AI companies favor people who fill gaps immediately.

Cat Wu mentioned their preference for those who “solve problems as soon as they see them” rather than strictly adhering to job descriptions.

Consequently, many Anthropic product managers have engineering backgrounds; designers code frontends; engineers possess product judgment.

The industry now values people who solve problems across boundaries over those who excel in single roles.

This reflects how AI is amplifying knowledge work leverage.

In the past, individual workload limits necessitated hyper-specialization. Now, AI boosts personal productivity.

Cat Wu gave numerous examples: Using CoWork to auto-generate presentation slides; sales teams creating customized proposals from client data; team members auto-organizing meeting materials via Slack, Gmail, and Google Drive.

Tasks that once took hours or days now complete in minutes. One person can handle increasingly more. What limits output now is the ability to identify problems, define directions, and act swiftly.

This is why Cat Wu repeatedly emphasizes “agency”—the capacity to drive outcomes proactively. She considers it one of the AI era’s most critical traits.

As future changes accelerate and role boundaries blur, truly scarce talent won’t wait for instructions but will identify team gaps and fill them immediately.

In a way, AI hasn’t diminished human value—it’s reselecting for those with judgment, execution, and decision-making capacity amid chaos.

Article Source: How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code), Lenny's Podcast

THE END

Copyright and Disclaimer

1. Content Copyright: Except for quoted public data, policies, and cases, this article is original. Professional data comes from authorized databases and government websites; cases are compiled from real events.

2. Image Licensing: Some images are proprietary or officially licensed; others are AI-generated. Uncredited network images remain copyrighted to their original authors and will be removed upon request.

3. Republication Rules: Unauthorized reproduction is prohibited. Republished content must retain full attribution and authorship.

4. Liability Disclaimer: This article represents the author’s observations on business figures and industry commentary based on public information. It is for reference only and does not constitute professional advice. Risks from its use are borne by the user. Channel Partner retains final interpretation rights.