Back to Videos

Workflow

YouTube

This video captures a speaker, Dan Shipper, Co-founder & CEO of Every, delivering a presentation at the AI Engineer Code Summit. Shipper enthusiastically shares his company's approach to building an "AI-native company," emphasizing that the playbook for this is still being invented. He highlights Every's success in running six business units and four software products with only 15 full-time employees, with 99% of their code written by AI agents. This unique setup allows individual developers to build and maintain complex production applications, a feat he claims would have been impossible even a year ago. Shipper outlines Every's "compounding engineering" process, a four-step loop of Plan, Delegate, Assess, and Codify. He explains that unlike traditional engineering where each new feature makes the next harder, compounding engineering aims to make each subsequent feature easier to build by codifying tacit knowledge into prompts for AI agents. This approach fosters parallel work, rapid prototyping of risky ideas, a demo-driven culture, and enables easier knowledge sharing among developers, allowing them to work with their preferred tech stacks and languages. Key takeaways include the 10x productivity difference with 100% AI adoption, the ability for a single engineer to manage complex products, the iterative nature of compounding engineering, and the emergence of non-obvious second-order effects like managers contributing code and seamless onboarding for new hires. Shipper concludes by inviting the audience to explore Every's offerings to stay at the cutting edge of AI.

AI-native companycompounding engineeringAI engineeringdeveloper productivityAI agents
View on Youtube

Introduction and Personal Anecdote

  • The speaker, Dan Shipper, Co-founder & CEO of Every, introduces himself as the last speaker of the day.
  • He expresses surprise at the audience size, jokingly referencing a tweet about AI FOMO in New York vs. San Francisco, affirming his love for New York.

The AI-Native Playbook is Being Invented

  • Shipper states that there is no existing playbook for building AI-native companies, or if there is, he doesn't have it.
  • He believes "What an AI-native company even is is being invented right now by all of us."
  • His talk will offer "dispatches from the frontier" based on Every's experiences.

10x Difference with 100% AI Engineers

  • He highlights a "10x difference between an org with 90% vs. 100% AI engineers."
  • This is because with a traditional engineering method, even 10% non-AI work can pull the entire organization back.

About Every

  • Every runs 6 business units and 4 software products with just 15 full-time employees.
  • They've grown MRR double digits every month for the last 6 months, with 7,000+ paying subscribers.
  • They've raised over $1M in total.
  • "99% of our code is written by AI agents."

Single-Developer Apps

  • Each of Every's apps is built and run by a single developer.
  • He showcases "Cora," an AI email management app that summarizes emails and provides an assistant, primarily built by one engineer.
  • He also shows "Monologue," a speech-to-text app, and "Spiral," a planning and writing app, both built by single engineers and complex.
  • This level of output by single developers "would obviously not have been possible to do even a year ago."

Benefits of AI-Native Engineering

  • Parallel Work: Developers can work on multiple features and bugs in parallel.
  • Rapid Prototyping: Risky ideas can be rapidly prototyped, allowing for more experiments.
  • Demo Culture: Companies move toward a demo culture instead of a written or oral culture, enabling weirder things that one can only get if they can "feel it."

Compounding Engineering Process

  • AI has caused Every to create an entirely new set of engineering primitives and processes.
  • This process is called "compounding engineering."
  • Traditional engineering: each feature makes the next feature harder to build.
  • Compounding engineering: each feature makes the next feature easier to build.
  • It's a four-step loop:
    1. Plan: Detailed planning when working with AI agents is crucial.
    2. Delegate: Tell the AI agent to execute the plan.
    3. Assess: Evaluate whether the agent's work is good (tests, trial, agent code review, human code review).
    4. Codify: This is the "money step" where learned tacit knowledge is compounded back into explicit prompts for the organization to use.

Powerful Second-Order Effects

  • Tacit Code-Sharing: Developers can learn from each other's code with no time cost, as AI can easily understand and translate across different tech stacks/languages.
  • New Hires Productivity: New hires are productive on their first day because codified knowledge allows AI agents to set up environments and guide them effectively. This also enables hiring expert freelancers to "drop in" for specific tasks.
  • Flexible Tech Stack: Teams can work with the stack and language they like best; AI translates between them, removing the need for standardization.
  • Managers Commit Code: Managers, if technical, can commit code (even the CEO) because AI allows for working with "fractured attention," making it possible to contribute to the codebase without deep immersion.

Summary of Key Points

  1. 10x difference when you hit 100% AI adoption.
  2. A single engineer should be able to build and maintain a complex, production product.
  3. Compounding engineering makes each feature easier to build.
  4. There are many non-obvious second-order effects once you adopt it.
  5. Many people in SF don't know this yet. (A humorous jab at Silicon Valley).

Every - The Edge of AI

  • Every provides ideas, apps, and training to help people stay at the edge of AI.
  • They offer a daily AI newsletter, reviews of new models and products, a bundle of AI apps, and consulting for large companies.

Timestamps

00:00
IntroductionSpeaker Dan Shipper introduces himself and the topic of AI-native companies.
00:30
AI-Native Playbook is Being InventedShipper explains that the methods for AI-native companies are still being developed, and he will share insights from Every.
02:09
10x Productivity with 100% AI EngineersHighlights the significant productivity difference when all engineers adopt AI.
03:07
About EveryDetails Every's operations: 6 business units, 4 software products, 15 employees, 99% AI-written code.
04:04
Single-Developer ApplicationsShowcases complex apps (Cora, Monologue, Spiral) built and maintained by single developers using AI.
05:57
Benefits of AI-Native EngineeringDiscusses parallel work, rapid prototyping, and a demo-driven culture.
08:11
Compounding Engineering ProcessIntroduces Every's 4-step loop: Plan, Delegate, Assess, Codify, designed to make future features easier to build.
11:00
Powerful Second-Order EffectsExplores benefits like tacit code-sharing, day-one productivity for new hires, flexible tech stacks, and managers committing code.
16:20
Summary of Key LearningsRecaps the main takeaways from the presentation.
17:03
Call to ActionEncourages audience to learn more about Every and stay at the edge of AI.

Target Audience

Software engineers, engineering managers, CTOs, startup founders, AI researchers, and anyone interested in the future of software development and organizational productivity with AI.

Use Cases

  • -Adopting AI tools for enhanced developer productivity.
  • -Restructuring engineering teams for AI-native workflows.
  • -Rapid prototyping and experimentation with AI-generated code.
  • -Leveraging AI for knowledge sharing and onboarding in technical organizations.
  • -Enabling non-coding roles (e.g., managers) to contribute to technical projects.

Key Topics

AI-driven software developmentOrganizational transformation with AIFuture of engineering workflows