• The Tailwind Paradox: Framework Up, Business Down
• $30 vs. $100
• Moving Up the Value Chain
• The Winners and Losers Framework
  • Struggling Business Models
  • Resilient Business Models
• The Adapt-or-Pivot Reality
• What Tailwind Could Become
• What This Means for OSS Maintainers
• The Services Business Is Thriving
• The Accountability Premium
• The Bottom Line

This is Part 2 of a two-part series. Part 1 covered why AI is accelerating open source adoption, not killing it. This post examines which business models thrive and which struggle in the AI era. Subscribe to our mailing list if you haven’t already.

In Part 1, I argued that AI isn’t killing open source - it’s amplifying adoption at unprecedented rates. Akka.NET’s 35% year-over-year download growth in 2025, combined with industry-wide metrics showing 70-87% growth across major package registries, proves that LLMs are recommending and adopting libraries faster than humans ever could.

But there’s a second story buried in the Tailwind CSS saga that deserves its own analysis: AI isn’t killing open source - it’s disrupting specific types of open source businesses.

Adam Wathan captured this perfectly in his GitHub comment:

“Tailwind is growing faster than it ever has and is bigger than it ever has been, and our revenue is down close to 80%.”

Read that again. The framework is thriving. The business is struggling. This isn’t a contradiction - it’s the clearest illustration of where AI is creating winners and losers in the open source economy.

Let me break down what’s actually happening, using both Tailwind’s situation and our own experience at Petabridge as a counter-example.

The Tailwind Paradox: Framework Up, Business Down

Here are the exact numbers - framework adoption vs. business health:

Metric	Tailwind Framework	Tailwind Labs Business
2024 Downloads	481,129,145	Revenue: Baseline
2025 Downloads	1,089,875,862	Revenue: -80%
YoY Change	+126.5%	Engineering team: -75%

Let that sink in. The framework more than doubled its downloads in 2025 - the same year the business saw massive revenue decline and 75% of the engineering team was let go.

Adam also shared:

• Docs traffic: Down ~40% from early 2023
• Framework adoption: Higher than ever

This is the paradox that confused so many commentators. How can something be more popular than ever while the business behind it struggles?

The answer is simple when you understand what Tailwind actually sells:

Paid Products:

• Tailwind Plus (formerly Tailwind UI) - $299 personal / $979 teams - 500+ premium components, site templates, and the Catalyst UI kit
• Refactoring UI - $99-$149 - Design book and video tutorials teaching developers to design (20,000+ copies sold)

Marketing Channel:

• Documentation - Free, comprehensive docs that served as the primary funnel driving developers to discover the paid products

Every single one of these revenue streams was built on top of learning curve friction. People paid because understanding Tailwind had a cost, and paying for pre-built examples and education was cheaper than figuring it out yourself.

AI eliminated that friction overnight.

$30 vs. $100

Here’s the economic reality that’s reshaping entire categories of open source businesses. The math is simple:

Before AI:

• Learning Tailwind: 10-20 hours
• Building a component from scratch: 2-4 hours per component (best case)
• Buying Tailwind UI: $299 one-time, instant access to hundreds of components
• Decision: Buy the components, save time

After AI:

• Learning Tailwind: Ask Cursor, get working code in seconds
• Building a component from scratch: Describe it, get it generated
• Buying Tailwind UI: $299 for… what exactly?
• Decision: Why pay for what AI gives me instantly?

The component library business model just evaporated. Not because Tailwind became less useful - because the education and examples built on top of it became commoditized.

Moving Up the Value Chain

Here’s my read on Tailwind’s situation: individual developers with AI subscriptions don’t need component libraries. They don’t need tutorials. They don’t even need documentation - they just ask the AI. That market is gone.

But enterprises? Enterprises need something AI can’t provide: someone to be responsible when things break. That’s where the value chain moves - away from selling to individual creators, toward enterprise customers who pay for support, SLAs, and accountability.

The Winners and Losers Framework

Let me lay out the business models that AI is disrupting versus the ones it’s strengthening:

Struggling Business Models

1. Component Libraries and Templates

• WordPress themes
• UI component libraries
• Pre-built widgets and plugins
• CSS frameworks with paid add-ons

All of these share a common trait: low substitution cost, commodity output. If AI can generate something equivalent in seconds, why pay for it?

2. Education-as-a-Business

• Premium tutorials
• Video courses on library usage
• Documentation paywalls
• “How to use X” content

To be clear: education content isn’t going away. People still want to learn, and video content in particular still resonates - humans enjoy watching other humans explain things. But the bar is much higher than it used to be. Written tutorials increasingly get consumed by LLMs before humans ever see them, and “how to configure X” content is exactly what AI generates on demand. Education-as-a-business isn’t dead, but it’s a harder hill to climb than it was five years ago.

3. Learning-Curve-as-Moat

• Any business whose value proposition was “we already know how to use this complex thing”
• Consulting built purely on library familiarity
• Implementation services for well-documented frameworks

If the framework is in the training data, the LLM knows it too.

Resilient Business Models

1. Support and Accountability Services

This is where Petabridge lives. Our support subscription business had its best year ever in 2025 - 19% net subscriber growth and near-perfect retention for most of the year. Not despite AI adoption - arguably because of it.

Why? Because support subscriptions don’t sell information. They sell accountability and availability. When something breaks at 2 AM on a production system, you don’t want an LLM to hallucinate solutions. You want a human expert who knows your system and will stay on the call until it’s fixed.

AI doesn’t replace that function. It might help us diagnose faster, but the customer is paying for someone to be responsible - not for information.

2. Infrastructure and Managed Services

AI generates code. It doesn’t run code. It doesn’t monitor code. It doesn’t wake up when your cluster splits.

The entire observability, hosting, and DevOps ecosystem is more valuable in an AI world, not less. More code being generated means more code that needs:

• Monitoring
• Alerting
• Performance analysis
• Infrastructure management
• Deployment pipelines

Our observability product Phobos saw similar retention improvements throughout 2025. AI-generated Akka.NET applications still need to be monitored.

3. Complex Frameworks Requiring Deep Understanding

Akka.NET is a distributed actor system with cluster sharding, persistence, streams, and split-brain resolution. Can AI help you use it? Absolutely - that’s part of why our downloads are up 35%.

Can AI replace understanding why your cluster isn’t forming correctly? Why your shard entities are being passivated? Why your persistence journal is showing gaps?

Technically, it can try. But having reviewed plenty of AI-generated assessments of production issues from our customers - often using frontier models - the quality ranges from “half right” to “wildly insane.” LLMs are not AGI. They’re text prediction engines - remarkably effective ones, but not intelligent. When the pattern matching fails, it fails spectacularly. And that’s the business.

NServiceBus provides another powerful proof point. This is a 15+ year old commercial enterprise messaging framework - exactly the kind of “critical infrastructure” that should be disrupted by AI according to the doomsayers:

Period	Total Downloads	New Downloads	Growth
2024 (Jan→Dec)	35.0M → 47.8M	+12.8M	+36.6%
2025 (Jan→Dec)	47.8M → 64.6M	+16.7M	+35.0%

NServiceBus maintained 35%+ annual growth through peak AI adoption. Enterprise customers aren’t asking Claude to generate custom message bus implementations - they’re adopting proven frameworks at an accelerating rate.

The frameworks that are “challengeable” by AI are the simple ones - the ones where generating custom code is comparable to using the library. Complex frameworks with years of edge cases, production war stories, and subtle configuration interactions? Those become more valuable because AI makes them more accessible while the expertise to truly understand them remains scarce.

The Adapt-or-Pivot Reality

Here’s the uncomfortable truth for open source maintainers: you can’t put the AI genie back in the bottle. I’ve been writing about OSS sustainability for years, and one thing is clear - the economics keep shifting, and maintainers have to shift with them.

Adam explored this when he closed the llms.txt PR that would have made Tailwind’s documentation more LLM-accessible:

“The docs are the only way people find out about our commercial products, and without customers we can’t afford to maintain the framework.”

His concern: if LLMs can answer Tailwind questions without sending users to the docs, fewer people discover the paid products exist. I understand the impulse, but this is a losing strategy. You can make your documentation slightly less AI-friendly, but you can’t stop Claude from learning Tailwind from the millions of examples in its training data. You’re just making your framework slightly less competitive.

The sustainable path isn’t protecting the old moat. It’s building a new one. Some projects have responded to economic pressure by changing their licenses entirely. And here’s an interesting wrinkle: relicensing might actually be more viable in an AI world. If LLMs are predisposed to recommend popular solutions regardless of license, the free lunch ends for companies freeloading on OSS - the LLM becomes a flywheel driving traffic to your now-commercially-licensed product rather than a competitor.

And if some indie hacker is upset about spending $100/year on a license for their side project that probably won’t make money? They were never the target audience anyway. This is about moving up the value chain - selling to businesses and enterprises to whom licensing costs are trivial. That said, licensing changes are still a blunt instrument. The smarter play is often evolving the business model while keeping the project open.

What Tailwind Could Become

If I were advising Tailwind, I’d say: become Vercel.

Vercel solved this problem years ago. Next.js is free and open source. Anyone can deploy it anywhere. But Vercel sells:

• Hosting infrastructure
• Performance optimization
• Enterprise features
• Support and SLAs

They moved up the value chain from “library” to “platform.” The library drives adoption. The platform drives revenue.

Similarly, Tailwind could pivot toward:

• AI-powered design tools (like v0, but for Tailwind)
• Enterprise design systems management
• Hosted component previews and collaboration
• Support contracts for enterprise customers

The framework remains free. The business builds on infrastructure and services that AI can’t replace.

What This Means for OSS Maintainers

If you’re running an open source project with commercial aspirations, audit your business model against this framework:

Ask yourself:

1. Am I selling information that AI can now provide for free?
2. Is my value proposition based on a learning curve that AI has eliminated?
3. Am I selling components that AI can generate equivalently?
4. Or am I selling accountability, infrastructure, or expertise that AI amplifies rather than replaces?

If your answers lean toward the first three, you’re standing on melting ice. AI is going to disrupt these models - it’s time to start adapting.

The transformation framework:

Before AI	After AI
Sell components and templates	Sell AI-powered generation tools
Sell documentation and tutorials	Sell support and consulting
Sell library familiarity	Sell system architecture expertise
Sell one-time education	Sell ongoing accountability

The Services Business Is Thriving

At Petabridge, our services-based business model - support subscriptions, training, consulting - is stronger than ever. Support subscriptions grew 19%, Phobos retention improved dramatically, and framework adoption is up 36% year-over-year.

These aren’t contradictory trends. They’re complementary. More people using Akka.NET means more potential support customers. AI making Akka.NET easier to adopt means more people getting started who will eventually need expert help.

The key insight: we don’t sell information. We sell being there when it matters.

The Accountability Premium

There’s a concept I want to introduce: the accountability premium.

When you’re an individual developer using Cursor to generate Tailwind components, you don’t need accountability. If the component is wrong, you fix it. If it’s ugly, you regenerate it. The cost of failure is measured in minutes.

When you’re running a production distributed system processing millions of transactions, the cost of failure is measured in revenue, reputation, and potentially careers. In that environment, you don’t want to ask an AI for help. You want to call someone who is contractually obligated to help you fix it.

That accountability - the SLA, the phone number, the human who will stay on the call until it’s resolved - is what enterprises pay for. And AI doesn’t provide it.

This is exactly why the path forward for companies like Tailwind is enterprise, not individual creators. Enterprise customers aren’t buying information or components. They’re buying accountability, and that market is growing.

The Bottom Line

AI is transforming open source business models, but not uniformly:

Struggling:

• Component libraries and templates
• Documentation and education businesses
• Learning-curve-as-moat strategies
• Anything where AI can generate equivalent output

Thriving:

• Support and accountability services
• Infrastructure and managed platforms
• Complex frameworks that require deep expertise
• Anything where AI amplifies but can’t replace human judgment

Tailwind CSS isn’t struggling. Tailwind’s component business is struggling. The solution isn’t to fight AI adoption - it’s to pivot toward business models that AI strengthens rather than undermines.

At Petabridge, we’ve watched our framework adoption surge 35% while our support subscriptions grew 19%. That’s not despite AI - it’s because AI makes complex frameworks more accessible, which creates more demand for expert support when things get complicated.

The future of open source business isn’t bleak. It’s just different. The projects and companies that recognize this shift early and adapt their business models accordingly will thrive. The ones clinging to education-based revenue and component sales will need to evolve.

The market is separating. Choose your side wisely.

---

This is Part 2 of a two-part series. Read Part 1: AI Won’t Kill Open Source - It Will Amplify It for the data showing AI is accelerating OSS adoption, not replacing it. Subscribe to our mailing list for more posts like this.

What’s your open source business model? Is it positioned for an AI world? Find me on Twitter/X - I’d love to hear how others are thinking about this transition.

If you liked this post, you can share it with your followers or follow us on Twitter!