# Dispatch from the Future: Building an AI-Native Company

Dan Shipper, Every | AI Engineer CODE 2025 | 17:58

Source: https://www.youtube.com/watch?v=MGzymaYBiss
Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks.
Page: https://aietalks.com/talks/dispatch-from-the-future-building-an-ai-native-company
Published: 2025-12-18
Tags: agents, coding-agents, context-engineering, team-adoption, workflows

## TL;DR
- Every sees a 10x difference between an organization where 90% of engineers use AI and one where 100% do.
- AI agents let one developer build and maintain complex production applications, while managers can contribute code in short blocks of time.
- Compounding engineering turns plans, delegation, assessment, and lessons from development into prompts and artifacts that make later work easier.

## Summary
Dan Shipper describes Every as a small company testing what an AI-native organization can do. The company runs four software products with 15 people, and Shipper says 99% of its code is written by AI agents. Each app is primarily built by one developer. He connects this structure to agentic coding workflows, where developers delegate tasks, run several agents in parallel, prototype risky ideas cheaply, and spend less time inside a traditional code editor. This changes how the company decides what to build. A working demo can replace some of the memos and decks previously needed to win approval. Shipper calls the broader process compounding engineering. Teams plan, delegate, assess, and then codify what they learned into prompts, commands, and agent instructions. Those artifacts help new hires, freelancers, and developers working across products. He is also frank that this approach lets managers, including the CEO, commit production code with fractured attention.

## Key ideas
### Full AI adoption changes how the organization works
[02:16](https://www.youtube.com/watch?v=MGzymaYBiss&t=136s)
Shipper says there is a "10x difference" between an organization where 90% of engineers use AI and one where 100% do. The remaining 10% forces the company to keep working around traditional engineering methods. That limits what the organization can do when everyone else has stopped typing code into an editor and is delegating work to agents. At Every, he says this change has transformed what a small company can build. He presents the company as a lab for this model rather than as a finished example with a settled playbook.

### Every runs several substantial products with a small team
[03:00](https://www.youtube.com/watch?v=MGzymaYBiss&t=180s)
Every has six business units and four software products, built by 15 people. Shipper says the products have grown monthly recurring revenue by double digits for the last six months, with more than 7,000 paying subscribers and more than 100,000 free subscribers. The company has raised about a million dollars in total. He says 99% of the code is written by AI agents, using tools such as Claude Code, Codex, and Droid. Each application is primarily built by one developer. Kora manages email, Monologue converts speech to text, and Spiral is another large application built by one engineer.

### Agentic coding makes parallel work practical
[05:28](https://www.youtube.com/watch?v=MGzymaYBiss&t=328s)
Shipper traces Every's change to Claude Code and similar terminal-based tools that move work away from the traditional code editor. Developers delegate tasks to agents and can work on several features or bugs in parallel. He rejects the idea that using several agent panes is automatically unproductive. Engineers at Every use four panes at once and get useful work done. This parallel workflow helps one developer build and operate a production application. Because code is cheap to produce, developers can also ask an agent to research a risky refactor or prototype an idea before committing to the full work.

### Working demos replace some internal persuasion
[07:26](https://www.youtube.com/watch?v=MGzymaYBiss&t=446s)
Every is moving toward what Shipper calls a "demo culture." Previously, someone proposing a product or feature might need to write a memo, prepare a deck, or persuade colleagues to spend time on the idea. With agentic coding, a developer can make a rough version in a couple of hours and show people what the idea feels like. Shipper says this makes it possible to pursue stranger ideas, because people can respond to an actual working version instead of an explanation. The change affects how the company chooses and discusses work, as well as how quickly it can test an idea.

### Compounding engineering turns experience into reusable instructions
[08:50](https://www.youtube.com/watch?v=MGzymaYBiss&t=530s)
Shipper defines compounding engineering as a process where each feature makes the next feature easier to build. Its loop has four steps: plan, delegate, assess, and codify. Teams make a detailed plan, ask an agent to carry it out, inspect the result through tests, experiments, and code review, then turn what they learned into prompts and instructions. These can live in files such as CLAUDE.md, subagents, or slash commands. The goal is to turn tacit knowledge about bugs, planning, and good implementation into an explicit library that the whole organization can use.

### Agents make knowledge sharing work across products and stacks
[11:08](https://www.youtube.com/watch?v=MGzymaYBiss&t=668s)
At Every, agents can inspect how another developer built a feature in a different repository and reproduce the underlying process in another language, framework, or technology stack. Shipper gives shared features and OAuth as examples. This avoids some of the cost of extracting a general library or explaining the implementation by hand. He also says new hires can be productive on their first day because agent configuration files contain instructions for setting up an environment and creating a good pull request. Expert freelancers can drop into a project for a focused task without paying the usual setup cost.

### Managers can contribute code in short, interrupted sessions
[15:09](https://www.youtube.com/watch?v=MGzymaYBiss&t=909s)
Shipper says developers at Every sometimes submit pull requests to other products when they find a bug or a small quality-of-life issue. He expects customers may eventually be able to use agents to fix bugs and submit pull requests too, though he presents that as speculative. AI also lets managers work with fractured attention. Instead of needing a three-hour block to solve a problem, a manager can ask an agent to investigate a bug during a meeting, return later to review the plan or root cause, and submit a pull request. Shipper says he has committed production code despite running the company and managing four products.

## Notable quotes
- "I actually don't have one, unfortunately. And that's because I think the playbook is actually being invented right now." (01:14)
- "There is definitely a huge there's a 10x difference between an org where 90% of the engineers are using AI versus an org where 100% of the engineers are using AI." (02:16)
- "In traditional engineering, each feature makes the next feature harder to build. In compounding engineering, your goal is to make sure that each feature makes the next feature easier to build." (08:50)
- "AI allows engineers to work with fractured attention." (15:43)

## Tools & references mentioned
- Every
- Kora
- Monologue
- Spiral
- Claude Code
- Codex
- Droid
- SuperWhisper
- Whisper Flow

## Who should watch
- You are deciding whether your engineering team should standardize on agent-based development and want an example from a small company already using it across products.
- Your team spends a lot of time writing proposals before testing ideas, and you want to understand how working demos can change that process.
- You manage engineers or products and want to see what AI-assisted development looks like when work happens in short, interrupted sessions.

## Related talks

- [From AI-Assisted to AI-Native: Building a Frontier Development Team](https://aietalks.com/talks/from-ai-assisted-to-ai-native-building-a-frontier-development-team) (Clare Liguori, AWS, 20:57)
- [The Agent Native Company](https://aietalks.com/talks/the-agent-native-company) (Rick Blalock, Agentuity, 20:58)
- [Agents for Everything Else](https://aietalks.com/talks/agents-for-everything-else) (swyx, AI Engineer, 14:10)
- [The 4 Patterns of AI Native Development](https://aietalks.com/talks/the-4-patterns-of-ai-native-development) (Patrick Debois, 14:11)
- [Building an Autonomous Engineering Org](https://aietalks.com/talks/building-an-autonomous-engineering-org) (Angie Jones, Agentic AI Foundation, 17:36)