The AI paradox: More automation, more humans, more work | Dan Shipper

Key Concepts

• AI Agents: Autonomous or semi-autonomous software entities that perform tasks, manage workflows, and act as "digital partners."
• Codex/Claude Code: Terminal-based or desktop-integrated AI environments that serve as the primary "operating system" for modern knowledge work.
• Vibe Coding: A colloquial term for building software by describing desired outcomes to an AI agent rather than writing manual code.
• Forward-Deployed Engineer: A specialized role focused on maintaining, "gardening," and optimizing AI agents to ensure they function correctly for the broader organization.
• The Allocation Economy: The shift in human labor toward managing AI agents rather than performing manual execution.
• SaaS Apocalypse (Rebuttal): The argument that AI agents will increase, rather than decrease, the demand for SaaS tools by acting as high-volume users.

---

1. The Future of Work: Bifurcation and Integration

Dan Shipper predicts that work will evolve into two primary modes:

• The Super Agent: Companies will move away from individual, fragmented personal agents toward a single, company-wide "super agent" (e.g., Shopify’s "River"). This agent acts as a central hub for data requests and organizational tasks.
• The New Work Surface: Professional work will increasingly shift away from traditional browsers and standalone apps toward integrated environments like Codex or Claude Code. These tools act as an "operating system" where the AI has visibility into the user's computer, terminal, and browser, allowing it to execute complex, multi-step tasks.

2. The "Gardening" Paradigm

A central argument presented is that "automation is a lie"—every agent requires a human to "garden" it.

• Maintenance: Agents are not "set and forget." They require human oversight to handle errors, provide context, and ensure the output remains aligned with company goals.
• The Human-in-the-Loop: Even as models become more autonomous, the human role shifts to that of a manager. The most successful individuals will be those who can effectively delegate to agents while maintaining a high-level view of the system's output.

3. The Role of SaaS in an AI-First World

Shipper offers a contrarian take on the "SaaS Apocalypse":

• Bullish on SaaS: He argues that agents will drive an exponential increase in SaaS usage because agents will interact with these tools at high volumes.
• Changing Business Models: SaaS companies should stop trying to build AI into their products and instead focus on making their products "agent-friendly." This means building robust CLIs and clean HTML structures that allow agents to navigate and manipulate the software efficiently.
• Token Economics: In this future, the user brings their own AI (and tokens) to the SaaS platform, which may actually improve the margins for SaaS providers.

4. Who Will Thrive?

• Product Managers (PMs): PMs who become "AI-pilled" are uniquely positioned to succeed. By pairing their product sense with the ability to use coding agents, they can ship features and prototypes without needing a full engineering team.
• Full-Stack Designers: Designers who can use AI to build their own interactions and pull requests will become significantly more valuable, as they can bypass the traditional "hand-off" friction between design and engineering.
• The "Generalist" Advantage: The ability to act as a generalist is becoming a superpower. AI allows individuals to touch multiple parts of the product stack, making small, agile teams more effective than ever.

5. Notable Quotes

• "What models do in general is they make yesterday's human competence cheap."
• "Automation is a lie. Every agent needs a human."
• "I am simultaneously extremely AI-pilled and very bullish on humans."
• "The edge of AI is wherever AI meets a real human doing something."

6. Actionable Insights for Success

• Ride the Models: Don't ignore new model releases. Treat them as new "powers" and constantly "turn over the rock" to see if they can now solve problems they couldn't previously.
• Adopt the Workflow: Shift your daily work into environments like Codex or Claude Code. Use the in-app browser to let the agent see what you are working on.
• Find Your "Moment of Joy": Don't approach AI out of fear or FOMO. Find a specific, enjoyable problem in your life or work that AI can solve, and use that as your entry point.
• Prioritize "Spaciousness and Strength": When feeling overwhelmed by the pace of change, approach the technology from a position of calm, strategic strength rather than reactive panic.

Synthesis/Conclusion

The future of work is not a total replacement of humans by AI, but a fundamental transformation of the shape of work. While the "AI job apocalypse" is largely a myth, the nature of professional roles is shifting toward management, system design, and the curation of AI-generated output. Success in this era requires curiosity, a willingness to experiment with new tools, and the ability to maintain a human-centric, "spacious" perspective while leveraging AI to handle the prosaic, repetitive aspects of knowledge work.