# The Knowledge Economy: Who Actually Gets Paid? "Knowledge economy" has been the dominant framing for developed-world work for thirty years. The idea is that value flows from knowledge, not from physical labour or raw materials. Brain work, not manual work. Creative and analytical output, not production-line output. The framing has always been a bit convenient for the people at the top of it. Because the knowledge economy as it's actually operated hasn't distributed the value from knowledge very evenly. The people who monetise most efficiently from knowledge aren't always the ones generating it. They're the ones who own the platforms, the distribution channels, the data, and the aggregation infrastructure that sits between knowledge and the market. A journalist generates knowledge. The platform that aggregates and distributes it captures most of the advertising value. A developer builds a tool. The app store that distributes it takes 30% off the top. A researcher produces insights. The journal that publishes them charges universities thousands of dollars for access while paying the researcher nothing for peer review. A creator builds an audience. The social platform owns the relationship with that audience and can switch off the creator's reach anytime the algorithm changes. The pattern is consistent: generate knowledge, hand it to an intermediary, receive a fraction of the value it creates. --- AI is accelerating this pattern dramatically, which is why the next few years matter so much for how it resolves. Every large language model is trained on human-generated knowledge. Books, articles, forum posts, code repositories, conversations, creative works - the accumulated intellectual output of millions of people over decades. That training data is the foundation of what makes these models valuable. Without it, there's no model. Without the model, there's no multi-billion dollar company. The people who generated that knowledge received nothing for its use in training. Some of them didn't know it was happening. Some have started suing. The legal questions are genuinely unresolved. But in practical terms, a transfer of value has occurred from the people who created the knowledge to the companies who packaged and monetised it, and the magnitude of that transfer is historically unusual. This isn't a conspiracy. The incentives were clear and the platforms acted on them. But it does point at a structural problem: the knowledge economy systematically undercompensates knowledge generation relative to knowledge aggregation and distribution. --- The question worth asking is whether that's inevitable or whether it's a design choice that can be made differently. I think it's a design choice. And the design is starting to matter more because the tools to do it differently are becoming available. Direct knowledge monetisation is genuinely easier than it was ten years ago. Substack, Patreon, and similar platforms let knowledge creators sell directly to the people who value their work, cutting out the intermediary who captures most of the value. The platforms still take a cut, but it's a service fee, not an extraction - and there's competition that keeps it honest. Verifiable knowledge attribution is becoming technically feasible. If you can prove that a specific insight or piece of knowledge originated with you - through cryptographic signing, through on-chain provenance, through verifiable timestamps - you have a basis for compensation claims when that knowledge is used commercially. We're early in this, but the infrastructure is being built. Data as a direct economic asset is becoming a real concept. The idea that you should be compensated when your personal data is used to train commercial AI models is moving from fringe to mainstream. There are legislative proposals in multiple jurisdictions. There are startups building the infrastructure to make it work. It's not there yet, but the direction is clear. And decentralised knowledge markets - where knowledge is exchanged directly between producers and consumers without a centralised intermediary capturing the margin - are technically possible in ways they weren't before. The intermediary's historical advantage was aggregation and trust. Blockchain-based systems can provide both without the extraction. --- There's a more specific point about the value of different kinds of knowledge that I think gets missed. The knowledge economy as currently structured heavily rewards knowledge that can be packaged and distributed at scale. Writing, software, media, analysis - things that can be reproduced infinitely once created. It very poorly rewards knowledge that is tacit, local, and experiential. The farmer who knows her specific land better than any agricultural textbook. The tradesperson who's developed techniques over thirty years that aren't written down anywhere. The community elder who holds the history and the relationships. The nurse who knows this particular patient's patterns in ways the electronic records don't capture. This knowledge is real, valuable, often irreplaceable - and nearly invisible to the knowledge economy as currently structured, because it can't be easily packaged and distributed. AI systems that are good at capturing and structuring tacit knowledge could actually help here. Not by extracting it from people without compensation, but by helping make it legible, shareable, and attributable in ways that let its holders benefit from its value. That's a different application of AI to knowledge than the scraping-and-training model. It's more interesting and more equitable. --- The knowledge economy has always had a marketing problem. It promised that knowledge workers would be the primary beneficiaries of the knowledge age. In practice, the biggest beneficiaries have been the companies that built the infrastructure through which knowledge flows. That's not a fixed arrangement. Infrastructure ownership is a choice. Distribution models are a choice. The technical tools to build alternatives are available and getting better. The people who'll shape what comes next are the ones asking whether the current arrangement is as good as it can be, or whether there's a version where the people actually generating the knowledge get a better deal. I reckon there is. Working out what it looks like is the interesting problem.