We Need a Fight Club for Software

In 2021 I started a startup that commercially failed. Download majr and see what failure looks like. It’s gorgeous. That’s a story for another day. Today I want to talk about some freedoms that I believe in and that need to return if the software industry is going to unlock another era of growth and innovation.

Let’s begin with data freedom. Because good data is the fuel for artificial intelligence, we’re about to witness the largest data lockup in computing history. Large companies are already invoking security, privacy and in particular child safety as armature for why data can’t be made available to anyone but themselves. To be sure, those issues require total commitment. They are not, however, a justification for blocking access while simultaneously harvesting the same data for their own benefit. Watch yesterday’s WWDC and you’ll see why this is about to become a pressing issue. It took Craig Federighi 20 seconds to mention the children.

‘Customers owning their data’ is another popular rhetorical device used to obfuscate the issue of data freedom. In the context of freedom, ownership is a losing argument because ownership isn’t the question being asked. Nobody disputes that customers own the most valuable data. Another popular argument is that privacy is a fundamental human right and therefore requires fundamentalist protections. Nobody dares debates this because privacy is a fundamental right. Both of these issues focus on values, not choices.

Instead, I’d argue the question that matters is whether these companies will choose to allow customers to move, share, and derive new value from their own data. To do that well, we need a principled, good faith model for the interexchange of data by agentic systems. By the way, this is an old issue; e.g. 30 years ago we produced the Open Financial Exchange (OFX) to allow new products like Quicken to work with your private financial transactions so consumers could make budgets, pay bills and balance checkbooks. Pretty sensitive data! 30 years ago! Old issues made new again.

New products often require access to new data, some of it sensitive like financial transactions. Commensurate with the explosion of agentic software engineering, it should be the case that a software builder of any size can work with any customer data subject to legal customer consent. To get to that future, we need to demand new governance structures that put customers in the driver’s seat; controlling how they share and trust-source their data. One model I really like is a legal concept borrowed from open-source licensing called mere aggregation; aggregation allowed GPL licensed code to touch proprietary software while maintaining their separateness. I’ve heard it also described as a conveyance. For data, this might imply that data can be accessed for the purpose of derived generation, but origin data isn’t retained or stored at rest. This would allow customers to source data for novel scenarios without violating datacenter sovereignty. Think of Google’s gmail.readonly scope, but more like a functional spec to generate new data from what’s read, while the original e-mail stays protected. Another idea is a data harness that sits between data and AI systems to coordinate permissions that are maximally safe and maximally free.

The next freedom is for our designers. We have a generation of incredibly talented creatives who are so depraved they’re willing to celebrate an opacity slider. I wish they’d realize that by skinning an entire interface in glass, we’ve suffocated the ability to create anything much different than what must converge to conform to transparency and refractions. Unified means unified and now your UX has been unified. This is a mostly terrible outcome, notably for consumers. The same can be said of the web with programming, styling and build pipelines so complex, no one can afford interesting design. Today’s frameworks and large platform guidelines invert the golden rule: form follows function. Instead, function follows form, constricting the possibilities available to designers.

Consequently, end users have come to expect conformity; to never expect the unexpected. How boring, droll and unoriginal we expect our software design to be. There are many more reasons than Apple’s guidelines that drive this monoculture, but it represents the problem set well. I think back to Winamp, Myst and Kai’s Power Tools. We used to be more free. Now we pay for Figma while Claude /frontend-design corrals our work to the dulcet tones of Taylor Swift.

As I was writing this I came across: Apple says it may remove apps if they don’t attract users. Surely this is designed to combat AI cloneware, but it simultaneously sends the message, “Don’t bother. It’s solved.” In an age where the marginal cost of software creation has gone from a $200,000 salary to $20/month, the largest technology company in the world declares, ‘nothing to make here.’ Interesting. Where is the Laural Canyon of software going to emerge?

Lastly, we need to prison break the snatty ratty risk-takers. Do you remember when Microsoft and AOL played cat and mouse every single day with IM interop? When Sean Parker released Napster? When the Netscape and Internet Explorer release velocity was weekly, adding fully formed features every single ship cycle? Again, Myst is a great example; it was an insane act to release an intellectual, point and click CD-ROM puzzlebox when the world wanted isometric arcade games. Today, a product manager would tell the Rand brothres not to bother because the games category is already saturated. Myst became the best-selling game of the decade.

Outsized rewards often flow from outsized risks. But making such bets requires risk-bearing on the part of a human being. Today, bearing that load means tolerating systems that have become rectilinearly rigid with policy concerns, regulatory compliance and the ambient stench of litigation. Creative impulses are cost prohibitive (I wrote that sentence, not AI.) If no reasonable person would incur risk bearing, then reasonable people will leave. Kudos to Epic for taking a stand on exactly this kind of thing with the App Store.

One idea I take inspiration from is so-called Insiders and Developer Previews that many companies use to encourage adoption and testing of cutting-edge releases. These releases are filled with bugs, security issues and experiments so that said companies can continue developing great products. We should have the same kind of sandboxing for dicey software. Want to build something high-risk? Great! We have a place for you to do that in a way where everyone consents. We need a Fight Club for software.

Of course software is infinitely malleable and therein lies the challenge. No company or government can account for every permutation of what and how to build. So we need a spirit of adventure, too.

I used to be snatty ratty when I was young. But when you end up smothering creativity, you kill the artist inside all of us. My startup failed for many more reasons than what’s discussed here, but I remember moving to LA and having conversations with our customers, mostly emerging artists. More than one was afraid of saying or doing anything to upset Spotify. I laughed at the time.

Maybe we do want uniformity. It can feel comforting. I’d argue it robs us of a richer, more interesting and unexpected future.