I’ve known Brad Feld for close to 30 years. To say I would not be the person I am today had I not met Brad Feld would be an understatement. The lessons I learned directly from him and a man named Len Fassler and frankly from just being around them could fill several books and have powered me for a lifetime.
Thirty years is long enough to see multiple iterations of a person. Long enough to know what is real and what is not. What is at the core of a person. I feel like I know what is at the core of Brad. What you see is what you get, he doesn’t hold back, transparency and honesty are his hallmarks, Oh and smart too
Back in the dot-com days, Brad was everywhere. It felt like he sat on every board that mattered. This was when Brad was managing partner over at SoftBank Capital, the Keiretsu days, when capital was flowing and everyone thought the party would never end. I was just getting my first company off the ground.
From the first time I met him, and yes, the first time he told me to move to Boulder, something stuck. It wasn’t just the intelligence. There have always been plenty of smart people in tech. It was the consistency. The integrity. The sense that the person you met in a boardroom was the same person you saw outside of it. It was also the passion. Brad was powered by it because he believed in what he was doing.
I wound up selling that company into what became Interliant, with Brad and his group involved. I have said it before and I will say it again. I would not be here without some of the things I saw and learned from him.
Over the years, I also saw something else. After a gazillion board meetings, 100s of startups succeed, fail or somewhere in between, Brad seemed to tire of the repetition in tech. The same cycles. The same noise. Innovation that felt more incremental than meaningful. The last few times we spoke, the energy was still there, but it felt more measured.
So when I saw him light up again around AI, around vibecoding, around what might actually be possible, it stood out. Not just because I was happy for him (and Amy). Because it reminded me that this moment we are in might actually matter. That it was time to change the world again.
Which brings me to his post, Opt-Out Is Not Consent.
This is classic Brad. Clear. Direct. Willing to say something plainly when it needs to be said.
And he is right.
Microsoft Knows Better
Let’s talk about Microsoft and GitHub.
By now you might have heard what the beef is, if not read Brad’s blog. Copilot will use your interaction data to train models by default. If you do not want that, you need to go find the setting and turn it off.
Buried in settings somewhere, not easy to find. Quiet defaults. A familiar pattern.
Brad framed it simply. Opt-out is not consent. Opt-out should be default and if not default, at least easy.
It is worth stating just as clearly as Brad said. Microsoft is a $3 trillion company. They can afford to ask. They can afford to make this opt-in. They can afford to make opting out obvious and accessible.
They chose not to.
This is not a technical limitation. Microsoft already built the version where consent is explicit. It exists today in their enterprise offerings. If you are a large customer with a contract, your data is protected.
If you are an individual developer or a small company, even a paying one, the experience is different.
That distinction is not accidental.
This Is Not Just Microsoft
It would be easy to focus on Microsoft and stop there. They always seem to make a good whipping boy. That would miss the larger point.
We have seen similar approaches from Salesforce and from a growing number of vendors across the AI and developer tooling ecosystem.
Default collection. Optional controls. Limited visibility.
This is not an isolated product decision. It is a pattern. And patterns like this tend to reflect underlying economics.
The Foundation of the LLM Economy
The current wave of AI did not emerge in a vacuum. It was built on vast amounts of data. In many cases, data that was not originally created with this use in mind. Content, code, and behavior that became inputs into model training.
Let me be blunt. Today’s LLM frontier models were built using data that the owners of the data did not necessarily give permission to be used for this purpose.
A few years ago, it was possible to argue that the implications were not fully understood. The technology was moving quickly. The boundaries were not clear.
That is no longer the case.
There have been legal challenges. There has been public scrutiny. There has been enough discussion across the industry that the issues are well understood.
Using data without clear disclosure, without meaningful consent, and without any form of value exchange is not a gray area anymore. It is a choice.
And it is one that deserves more scrutiny than it is getting.
Not Just Enterprise Rights
One of the more telling aspects of this situation is the difference in how users are treated.
Enterprise customers receive contractual protections. Individuals are expected to manage their own exposure through settings and preferences.
We see this across the industry.
Large organizations negotiate terms. Everyone else inherits defaults.
There are only five hundred companies in the Fortune 500. There are millions, tens of millions of developers, small businesses, and individual practitioners who contribute to and depend on these platforms.
They are not second class citizens in this ecosystem. Their work, their behavior, and their usage patterns are part of what makes these systems valuable.
They should not have to work harder to protect that contribution.
A More Sustainable Approach
None of this requires a fundamental reinvention.
Make participation explicit. Explain what is being collected and how it will be used. Make the controls visible and easy to understand.
If the data is valuable enough to improve the models, then it is valuable enough to acknowledge. Offering incentives or benefits in exchange for participation is not unreasonable. It is consistent with how value exchange works in every other part of the technology industry.
This is not about slowing innovation. It is about aligning it with practices that users can understand and trust.
Shimmy’s Take
I am glad Brad wrote what he wrote. I am glad he did not soften it.
And I am especially glad to see that energy again. The light is burning bright again out in Colorado. The same clarity and conviction that stood out to me 30 years ago is here and it says to me as I said earlier, hey, it’s time to change the world again. I have been feeling, doing, writing and speaking about it for a while now. But this feels like validation to me.
This moment calls for it. But if we are going to change the world, lets do it the right way. Let’s empower and embrace, not use, abuse and disrespect.
This is not just about Microsoft or GitHub. It is about the direction this industry is taking as it builds the next generation of systems.
We have enough experience now to understand the tradeoffs. We have seen enough to know where the lines should be.
The question is whether we are willing to draw them.
