_006 Data as currency

The recent layoffs at Tailwind are a warning shot: when a company’s IP can be effectively commoditised by someone else through large-scale scraping, the value chain breaks, and the creator loses.

There is a legitimate concern that LLMs are profiteering from organisations and individuals who create and publish content without any reimbursement whatsoever. The open web becomes a free raw-material supply chain, and the businesses funding the creation of that material are left to absorb the cost.

The inevitable outcome is clear: creators are incentivised to hide what they produce. Whether it’s Tailwind documentation or a humble blog post, the message is the same - if you allow your information to be accessed for free, AI will use it to compete with you. In the worst case, it will replace you.

The obvious solution is a revenue-share model: LLM providers pay commission to the websites they scrape, in proportion to usage and value derived. But looking at how aggressively many of these companies are financed, and how intense the race is to meet what are blatantly impossible investor expectations, this seems unlikely.

Now add the compounding factor: the internet is increasingly being filled with AI-generated content. If a large and growing portion of new content is machine-produced, and models still carry non-trivial error rates of 17% (“hallucinations”), then we’re heading toward a credibility death spiral. Low-integrity content starts to dominate. Models train on the outputs of other models. Errors become self-reinforcing. The signal-to-noise ratio collapses.

This is the real risk: human-verified, high-integrity information stops being socialised because creators don’t want it harvested and commoditised, while AI systems increasingly ingest a web saturated with synthetic output.

We don’t just get more content - we get more confidently wrong content.

Meanwhile, business leaders champion the automation of anything and everything with a short-term focus on next quarter’s numbers. What’s not being addressed is the long-term implication of an unresolved conflict: freely available high-integrity information is the fuel, but the people who create it are being punished for making it public.

Market solutions will have to emerge - monetisation, licensing, technical protection, verification layers, provenance, and “human-certified” supply chains of information - or we drift into an information desert where AI feeds on itself, quality decays, and businesses (and single operators) get wiped out along the way.

To be clear: I believe LLMs are here to stay. My concern is the cavalier approach many AI giants are taking to the commercial, qualitative, creative, and social impact of what they’re building -all while racing to satisfy increasingly absurd investment commitments.

The optimist’s view is that this creates opportunity: new security and provenance-focused products, new business models, new markets etc.

But the harder question is whether this disruption is proportionate, fair, safe, ultimately worthwhile….