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

May 24, 2026 Episode Page ↗
Overview

Dan Shipper, CEO of Every, shares his contrarian predictions for how AI will transform work in the coming year. He discusses the rise of company-wide "super-agents," SaaS thriving with user-provided AI tokens, and how roles like PMs and full-stack designers will dominate, while the AI job apocalypse remains a myth.

At a Glance
9 Insights
1h 34m Duration
16 Topics
7 Concepts

Deep Dive Analysis

Introduction to Dan Shipper and his AI predictions

Every's unique AI-forward work environment

The future of work: Super-agents and AI as OS

Why companies will adopt a single "super-agent"

Codex and Cloud Co-work as the new operating system

How SaaS companies should adapt to AI agents

Why CLIs are becoming obsolete for general work

The power of two agents collaborating

Why Dan is bullish on SaaS stocks

Automation's true impact: More human work

The value of human oversight in AI-generated code

The changing shape of work and new job roles

Addressing the "AI job apocalypse" myth

How to "ride the models" to stay relevant

Final predictions and advice for listeners

Lightning round: Books, TV, products, motto, underrated AI tool

Super-Agent Model

A single, general AI agent deployed for an entire company, often in platforms like Slack, that employees can delegate tasks to. It requires a dedicated human (e.g., a forward-deployed engineer) to maintain and ensure its effectiveness.

Codex/Cloud Co-work as OS

The idea that AI-powered coding environments like OpenAI's Codex or Anthropic's Cloud Co-work (and its desktop app, Co-work) will become the primary operating system for all knowledge work, including email, documents, and research, by integrating in-app browsers and direct computer access.

Human-Agent Collaboration Paradigm

A new way of working where humans and AI agents interact simultaneously on the same piece of work (e.g., a document or codebase), requiring software to be designed for mutual visibility, seamless back-and-forth, and robust logging/rollback capabilities.

Automation Paradox

The phenomenon where increased automation and AI capabilities, despite benchmarks showing greater AI autonomy, lead to humans having more work, often in roles of oversight, refinement, and managing the AI systems themselves.

Yesterday's Human Competence

A concept describing how AI models ingest vast amounts of existing data and make previously valuable human skills or outputs (e.g., writing, coding) cheap and commoditized, thus shifting the focus for humans to creating new, unique, and interesting applications.

Ride the Models

An approach to staying relevant in the AI era by actively and playfully engaging with new AI models as they emerge, applying them to one's work, and continuously exploring their capabilities to extend one's own powers and adapt to evolving technology.

Spaciousness and Strength

A mindset for dealing with difficult or uncertain situations, particularly regarding AI's impact, by approaching them from a calm, empowered, and objective perspective rather than fear, leading to more productive engagement.

?
How will the way we work change in the coming year?

Work will bifurcate: everyone will have at least one AI agent to offload tasks to, and most work will occur within AI-powered environments like Codex or Cloud Co-work, which act as the new operating system.

?
Will companies adopt personal AI agents or a single company-wide agent?

For now, the model is shifting towards a single "super-agent" for the entire company, as personal agents require too much maintenance and a dedicated human connection to be consistently useful.

?
How will the rise of AI agents impact SaaS companies?

SaaS is not dead; agents will increase the number of users for SaaS products. SaaS economics will shift as users bring their own AI tokens into apps, improving SaaS margins by reducing the vendor's token costs.

?
Are Command Line Interfaces (CLIs) still relevant in the AI era?

CLIs are largely over as a primary work surface; while they won't disappear, the trend is moving towards graphical user interfaces (GUIs) within AI-powered environments like Codex for most knowledge work.

?
Does automation reduce human work?

No, automation is a "lie" in the sense that every automated process still requires a human to oversee, maintain, and ensure its proper functioning, often creating new roles like the "forward-deployed engineer."

?
Which product/tech roles are least changed by AI?

While many roles are fundamentally transformed, sales (especially the in-person aspect) and potentially senior leadership (CEOs, investors) might appear less changed, though the latter risks falling behind without direct engagement.

?
Will we read more AI-generated writing, and will we like it?

Yes, we will read significantly more AI-generated writing in documents and emails, and we will like it, especially when it's well-directed and helps operationalize ideas, as the bar for human-written routine communication is often low.

?
Which product/tech roles will thrive most in the AI era?

Product Managers with strong product sense and full-stack designers who can build their ideas directly will be "superpower people," as AI liberates them from organizational overhead and allows them to focus on creativity and problem-solving.

?
Is the AI job apocalypse happening?

No, the AI job apocalypse is not a real threat. While companies may reorganize, AI primarily makes "yesterday's human competence" cheap, creating new demands for humans to apply models to novel situations and manage AI systems.

?
How can individuals stay relevant and successful in the AI future?

The key is to "ride the models" by consistently using new AI tools, being curious and playful in applying them to one's work, and continuously exploring what they can do, rather than ignoring them out of fear.

1. Embrace AI as Primary Work Surface

Integrate tools like Codex or Cloud Co-work into your daily workflow, using their in-app browsers to perform tasks like email management, document creation, and research, as these environments provide powerful agent access to your entire computer.

2. Design SaaS for Human-Agent Collaboration

If building SaaS, shift from designing solely for humans or agents to creating software where humans and AI agents collaborate seamlessly, ensuring both have visibility and can interact effectively.

3. Build Company Super-Agents

For organizations, establish a single “super-agent” (e.g., in Slack) to which employees can offload work, and assign a forward-deployed engineer to maintain and ensure its utility across the company.

4. Ride the AI Models

To stay relevant, consistently use the latest AI models for your work, experimenting with new capabilities and applying them to your tasks, even if it means trying things that didn’t work previously.

5. Master AI-Assisted Output Review

When using AI to generate content (e.g., documents, emails), ensure you understand and can stand behind every line, as the expectation for human oversight remains crucial despite AI’s capabilities.

6. Cultivate Curiosity and Playfulness with AI

Approach new AI models with a curious and playful mindset, continuously exploring how they can be applied to your job or personal projects, as this iterative discovery is key to finding valuable uses.

7. Develop Full-Stack Design Skills

Designers should leverage AI to build their ideas directly, making pull requests and creating unique, high-quality interactions that stand out from generic AI-generated “slop.”

8. Focus on Product Sense as a PM

Product Managers should lean into their strong product sense and user understanding, using AI to accelerate building and iteration rather than managing large teams, effectively liberating them to focus on core problem-solving.

9. Relate to Challenges with Strength

When facing difficult situations, particularly concerning the future of work and AI, strive to approach them from a position of “spaciousness and strength” rather than fear or anxiety, which can lead to more productive engagement.

I'm simultaneously extremely AI pilled and very bullish on humans.

Dan Shipper

Automation is a lie. Every agent needs a human.

Dan Shipper

What models do in general is they make yesterday's human competence cheap. And so it becomes commoditized. It's not valuable anymore. What humans do is we go in there and we're like, yeah, we have all this frozen human competence from yesterday. How do I use this to make something new and interesting?

Dan Shipper

I think the SaaSpocalypse is dumb. I would buy SaaS stocks right now.

Dan Shipper

CLIs are over. We speed ran the CLI era. It was nice while it lasted, but I think CLIs are over.

Dan Shipper

The only thing you need to do is ride the models.

Dan Shipper

The edge of AI is wherever AI meets like a real human doing something because the people in San Francisco, they're making it, but they don't actually know a lot about how to use it.

Dan Shipper

The most useful thing you can do is like find ways to use it well in your life and share it.

Dan Shipper

Senior Engineer Benchmark

Dan Shipper
  1. Give a new model a prompt: "This is vibe coded slop. If you wanted to rewrite it from first principles, how would you write it? Go do it."
  2. Score the model's output against human senior engineer rewrites (high 80s/low 90s out of 100).
  3. Observe if the model attempts to fix specific issues or fundamentally rewrites the codebase from first principles.

Quarterly Planning with Notion Agents

Dan Shipper
  1. Establish a top-level company strategy.
  2. Have every employee interact with a Notion agent, answering questions about past performance, goals, metrics, and how their plans align with company strategy.
  3. Receive AI-generated strategy reports or quarterly plans for each team.
  4. Review reports to identify inter-team collaboration needs and assess quality.
almost 30 people
Every company size Doubled in the last year from 15 people.
17 hours at 50% accuracy
Mythos preview model's autonomous task duration and accuracy Measures how long the newest models can do tasks autonomously.
10 days straight
Dan Shipper's inbox zero duration Achieved using Codex to manage emails.
30 out of a hundred
AI models' score on senior engineer benchmark (before GPT 5.5) Measures how well AI can rewrite 'vibe coded slop' from first principles.
high eighties, low nineties out of a hundred
Human senior engineer's score on benchmark Reference score for the senior engineer benchmark.
62
GPT 5.5's score on senior engineer benchmark A significant 30-point bump, showing progress towards senior engineer level.
50,000 Codex instances
Pete's use of Codex instances to sort OpenClaw pull requests Used to sort through thousands of daily pull requests.
a thousand
Number of OpenClaw pull requests Pete merges Merged daily after sorting with Codex.
10,000 pages in like a second
Agent's reading speed Allows for deep operationalization of ideas.