
I think every company should open source their code.
Audio Summary
AI Summary
The speaker expresses a growing distrust in closed-source software, believing the future lies in open-source solutions that allow users to make changes and customize products to their needs. While acknowledging the increased risk profile of open-sourcing applications—such as competitors cloning work, customers self-hosting, and security vulnerabilities—the speaker argues that not embracing open source could be detrimental for businesses.
The core argument for open source is framed through an analysis of successful, large-scale companies like AWS, Salesforce, and Retool. These companies often have thousands of features, but any given customer typically uses only a small fraction of them. A significant portion of these used features are common across most customers, while the remaining features are highly niche, used by less than 1% of the customer base. This creates a barrier for competitors, as even if they build 99% of a customer's needed features, the absence of one obscure, bespoke feature can prevent customer migration. Historically, the solution for these large companies was to hire more engineers and build every conceivable feature, locking customers into their ecosystem.
However, this dynamic is changing due to the rise of AI and a shift towards a "building block economy." The speaker highlights Vercel as an example. Vercel, despite offering far fewer features than AWS, excels in specific use cases like web application deployment. It allows users to "plug in" missing functionalities with their own code, rather than relying on a rigid plugin system. This modular approach, where users write code to integrate with other services (e.g., Cloudflare for firewalls, Superbase for databases), enables customization without Vercel needing to build everything.
The speaker's own project, T3 Code, an open-source alternative to AI coding GUIs, further illustrates this trend. T3 Code has seen a remarkable 10% of its weekly users fork the project to make customizations. This high volume of user-driven modification, exemplified by a user who built multiple terminals and a queuing system into their fork, demonstrates a desire for personalization that traditional closed-source software struggles to accommodate. The speaker notes that AI significantly lowers the cost of customization, as even non-developers can now prompt features into existence within a codebase. This eliminates the need for complex, often problematic, plugin architectures.
Mitchell, the creator of HashiCorp and Ghosty, reinforces this "building block economy" concept. He observes that the most effective way to achieve massive software adoption is no longer through high-quality mainline apps, but through building blocks that encourage others to build "quantity over quality." Ghosty, a popular terminal, gained a million daily users in 18 months, but its underlying library, Lib Ghosty, achieved multiple million daily users in just 2 months. This indicates that AI agents, and increasingly businesses, prefer modular components that can be assembled and customized. AI models, when unsteered, tend to recommend open, modular, and easily installable tools (like npm packages) over closed, monolithic solutions.
The advantages of this building block approach for businesses are numerous:
* **Lower Quality Bar for Niche Features:** Mainline apps must weigh every feature against long-term vision and maintenance for millions of users. Factory artifacts (customized forks) for a small user base don't have these constraints, allowing for faster and looser development of niche features.
* **Increased Awareness:** Niche communities can create specialized versions of tools, increasing awareness of the underlying building blocks.
* **Lower Maintenance Burden:** Maintainers can more easily decline feature requests, knowing users can fork and implement them independently.
* **Outsourced R&D:** Maintainers can observe user-created forks, cherry-picking the best ideas to integrate into the mainline product.
The commercialization aspect remains a complex challenge. Mitchell admits he doesn't have a concrete answer, noting that agents tend to favor open and free software over closed and commercial alternatives. However, the speaker proposes solutions for commercialization within an open-source framework, focusing on "stickiness." Instead of locking users in with features, businesses can offer core services (e.g., backend infrastructure, managed databases) while allowing users to fully own and extend the client-side codebase. This could involve "self-forking" software, where users customize an in-app fork, relying on the company for backend services.
The speaker introduces the concept of a "patch.md" file, which would encode the intent of user customizations. When an update occurs, an agent would attempt to resolve merge conflicts based on this patch.md, allowing users to maintain their personalized features while staying updated with the mainline project. This vision imagines a future where software is "self-forking, self-customizing, and self-healing," enabling normal users to modify their applications without extensive development knowledge.
In conclusion, the speaker asserts that open-sourcing a business offers significant advantages: customers contribute research, the need to support niche features is reduced, AI agents prefer open-source tools, and a strong community can form. This community, in turn, drives agent adoption and customer acquisition. The speaker is betting his portfolio on this shift towards an open, malleable, and community-driven software future, urging others to embrace it to ensure an open future for software.