There is a line from Francis Bacon that has survived four centuries of misquotation: knowledge is power. What Bacon actually wrote, in 1597, was closer to "knowledge itself is power" — a distinction that matters. He wasn’t describing something you possess - he was describing the intrinsic power in the act of knowing. The internet, at its idealistic peak, seemed like it might finally deliver on that promise. It did not, and the failure is instructive.

01 / THE PROMISE VS. THE REALITY

The Web Was a Utopian Idea.

The Market Had Other Plans.

Tim Berners-Lee submitted his proposal for the World Wide Web to CERN in March 1989 with a note from his supervisor that read, famously, “vague but exciting.” The proposal was neither a product roadmap nor an investment thesis. It was a philosophical argument: that knowledge, distributed freely across institutions and borders, would compound in ways that no single institution could contain or predict. The hyperlink was radical not because it was technically clever, but because it was structurally generous. Every document could point outward. Every insight could extend the commons.

For a stretch in the late 1990s and early 2000s, that promise felt genuinely operational. You could find a research paper from a university in Helsinki, a local news story from rural Montana, and a technical manual from a Japanese electronics company, all in the same afternoon, at no cost beyond your internet connection. The friction between a question and its answer was collapsing faster than anyone predicted.

Then the business model arrived. Digital advertising seemed, briefly, like an elegant solution: give the content away, monetize the attention. Publishers flooded the web. Audiences grew. Google became the de facto custodian of human knowledge. And for a moment, the utopian vision and the market seemed compatible.

They were not. The advertising model optimized ruthlessly for one thing: engagement. Not understanding, not accuracy, not depth. Engagement. Over twenty years, the gravitational pull of the system bent steadily toward outrage, sensationalism, and novelty. The publishers who refused to compete on those terms faced a hard choice: retreat behind paywalls or disappear. Most chose the wall. The Reuters Institute’s 2024 Digital News Report found that roughly 76% of major English-language publishers now operate some form of subscription gate or metered paywall.¹ The commons was being parceled into private estates.

Subscriptions were not irrational. When the inherent challenges of a purely advertising based model became clear, the paywall was the logical solution. Publishers who built them survived. The New York Times has eleven million subscribers. The problem is not that subscriptions exist. The problem is that they have no complement for the high-value moment when a non-subscriber needs one specific thing: the CFO who needs one market analysis before a board meeting, the journalist who needs one archived article to verify a fact, the researcher who needs one dataset to complete a paper. That intent is commercially significant. It is the highest-value moment in the reader relationship. It’s why search has always been such a powerful tool for connecting readers with information.  But the current infrastructure has no way to serve it.

The web did not fail because knowledge became less valuable. It failed because we built an economic system that made distributing cheap attention more profitable than distributing hard-won insight.

What is left is a landscape that would baffle Berners-Lee. Not because information is scarce, but because accessing the right information at the right moment now requires navigating a labyrinth of subscriptions, account walls, and friction-by-design. The web got bigger. It did not get more open. And the next phase of the internet - one where research functions are increasingly driven by AI and autonomous agents - is about to make that tension impossible to ignore.

02 / CONSUMER UN-FRIENDLY AI

The All-or-Nothing Store.

Where the Only Options Are Commitment or Crumbs.

Imagine walking into a grocery store where the only way to buy a banana is to sign an annual contract for the entire produce section. You do not want the mangoes. You are not sure about the star fruit. You just want the banana. The cashier shrugs. Annual subscription or the door. This is, with only minor exaggeration, how premium digital information works today.

The subscription model was built for loyalty. It assumes that the highest-value reader is the one who wants everything from a single source. For many publishers, that assumption has held. For the vast majority of readers — and for the billions of people who will never subscribe to any given outlet — it leaves the most commercially significant moment in the reader relationship unmonetized. The publisher loses a revenue event. The reader leaves frustrated. The information stays locked.

This was already a friction problem for humans. For AI, it is a structural failure. AI agents, which are now among the most active consumers of digital IP, do not navigate paywalls. They hit them and stop, route around them, or, in a pattern that has generated lawsuits from the New York Times to Getty Images, find ways to ingest information without a dollar returning to the source.²

The gap between what AI agents consume and what publishers earn from that consumption has no infrastructure solution today.

+300% 

YoY growth in AI bot traffic.  AI Bots now represent ~4–6% of all HTML traffic (and rising) 

Cloudflare Radar Year In Review, 2025 [3]

$0

Revenue returned to independent journalists, specialist archives, and regional outlets without the scale to negotiate direct licensing agreements, when their content enters AI pipelines

Authors Guild commentary on AI licensing, 2024 [4]

The result is a system producing two bad outcomes simultaneously. Either quality information gets walled off from AI retrieval pipelines, making every AI output worse and making those tools less valuable to the users paying for them. Or it gets scraped wholesale, with long-tail publishers receiving neither compensation nor attribution. The major players have begun negotiating licensing agreements. The independent journalist covering water rights in New Mexico has not, and probably cannot.  If we continue down this road, I think this will be the single biggest failure of the Internet era.  The intention was to have knowledge distributed, and enable the distributed discovery of knowledge to flourish.  If we concentrate that discovery process in the hands of a few large institutions, we’re all going to be worse off.  

Subscription fatigue compounds the problem. Deloitte’s 2024 Digital Media Trends survey found that the average American household pays for more than four streaming or digital media services at once, and over 60% of subscribers feel they are paying for access they rarely use.⁵ Annual churn rates in mid-tier digital news hover between 35 and 40%, with most departing subscribers citing cost and infrequent use.⁶ People will pay for information. They will not pay, indefinitely, for the possibility that they might need it.

03 / A BETTER SYSTEM IS EASY TO IMAGINE

One Article. One Banana. One Pair of Socks.

The Transactional Web We Already Know How to Build.

The iTunes Store solved this problem for music in 2003. Before it, you either bought the album or you pirated the track. Apple’s insight was that the real unit of value was not the album but the song. Ninety-nine cents, frictionless, immediate. In its first week, the iTunes Store sold one million songs. The music industry, which had spent years suing teenagers for piracy, suddenly had a legal transaction architecture that consumers actually preferred.⁷ The lesson was not that people do not want to pay. It is that people will not pay for more than they need.

iTunes also showed what happens when one company builds the rails: Apple captured the majority of the value. Musicians saw marginal improvement. The infrastructure determined who won, and the infrastructure was proprietary. That outcome is not an argument against unit-pricing. It is an argument for building the rails differently this time, on open standards that publishers can connect to and leave without data lock-in.

The same logic applies to information. A reader who would pay two dollars for one article about a breaking story they urgently need will not sign up for a twenty-dollar-per-month subscription on the same impulse. The moment of genuine need goes unmonetized. Everyone loses.

Micropayments have a history worth acknowledging honestly. Blendle, Flattr, and Scroll all ran experiments on this model in the past decade and ran into the same constraint: the friction of individual payment decisions exceeded the value of individual transactions. Nick Szabo identified this in 1997 as the “mental transaction cost” problem. The objection is real. The question is whether the infrastructure has changed enough to overcome it.  And whether this new consumer - agents and AI discovery tools - that never gets tired of the user defined job it’s been given, offers enough of a shift in the demand equation to allow this system to flourish at scale.   

We do not need to invent a new idea here. We need to build the infrastructure for an old one. Pay for what you use. Value flows to where insight is created.

This model also addresses something that gets lost in the platform-versus-publisher debate: the fate of small, specialist IP owners. The current system concentrates economic power with whoever can afford to market a subscription at scale. A freelance investigative journalist covering municipal finance in Tulsa, a documentary filmmaker with deep access to a niche community, a fitness coach with genuinely exceptional programming — none of them can realistically compete on the subscription model’s terms. In a transactional system, their work gets paid for each time it is actually used. Quality, not marketing budget, determines revenue.

Ted Nelson was sketching micropayment architectures for the web in the 1960s, long before the web existed, under the project name Xanadu.⁸ What is new is that the infrastructure to actually execute this now has most of its components in place: fast API-accessible payment rails, AI systems capable of expressing precise information needs at the query level, and seamless licensing systems which allow publishers to price individual assets at scale. The integration layer connecting these components is exactly what LedeWire is building.

And those rails change the incentive structure at the foundation. In an advertising economy, the incentive is clicks. In a subscription economy, it is retention. In a usage economy, it is quality: producing the specific insight that someone, somewhere, needed enough to pay for in the moment they needed it. That is not just a better business model. It is a better epistemic environment. The best information should win. Right now, the loudest does.

04 / INTRODUCING LEDEWIRE

The Transaction Layer the Open Internet Always Needed.

Fulfilling the promise of the open web.

LedeWire is not a subscription replacement. Subscriptions work for publishers who can achieve the scale to sustain them, and for the independent writer or data owner that offers value to their core audience. LedeWire serves a different moment: the billions of high-value discrete interactions that subscriptions structurally cannot reach. The subscription and the transaction are not competitors. They are different instruments for different points in the reader relationship. Publishers who already run subscription businesses are among the most natural partners for what LedeWire is building.

The underlying plumbing problem is specific: publishers have no way to receive compensation when their content is retrieved by AI systems at the query level, and human readers have no way to pay for individual articles without committing to a subscription they may not be able to justify. LedeWire is the transaction layer that addresses both.

It is a shared wallet infrastructure that allows both human readers and AI agents to discover, license, and pay for information at the query level, with revenue routing automatically back to IP owners. Publishers receive the majority of each transaction. LedeWire’s infrastructure fee is deducted at settlement, on the same model as payment infrastructure that preceded us. Publishers connect through existing protocols, including the Tollbit publisher network and the x402 standard, an emerging machine-to-machine payment protocol currently in active development.⁹ 

8,800 to 1

A few years ago Google crawled 2 web pages for every 1 human user it referred.  Recently that ratio is up to 8,800 to 1 for some AI companies

Cloudflare and Business Insider

$3–5T

Projected agentic commerce market by 2030, in need of the infrastructure rails LedeWire is building

Derived from McKinsey Global Institute generative AI projections, 2023 [10]

The categories where this unlocks value extend beyond journalism. Documentary filmmakers who want to license specific research or archival access. Independent financial analysts whose models are more accurate than anything a wire service produces but whose audience is narrow and specialist. Alternative dataset owners who have unique insights into the health of an industry or business. In every case, the existing system either excludes them or underpays them. A transactional layer changes that. The independent reporter covering water rights in New Mexico gets paid every time a policy researcher pulls her archive. The documentary editor fact-checking a historical claim pays for exactly the primary source they needed, and the archive that holds it gets compensated for the first time.

Berners-Lee’s original insight was about generous architecture: build something that points outward, that allows every node in the network to connect to every other, and the system will grow in ways no single institution can predict or contain. The paywall inverted that logic. LedeWire is an attempt to recover it — not by making information free, but by making it accessible. Paid, licensed, compensated, and open to anyone willing to pay for what they actually need.

The rails are being laid right now. If you are publishing content that deserves to be paid for, or building AI systems that need to access valuable work, this is the moment to be involved.

Get in Touch

Publishers, AI builders, and media investors: the window to shape this is open.

hello@ledewire.com

SOURCES & NOTES

1. Reuters Institute for the Study of Journalism, Digital News Report 2024. University of Oxford. reutersinstitute.politics.ox.ac.uk
2. The New York Times Co. v. Microsoft Corp. et al., U.S. District Court, S.D.N.Y., filed December 27, 2023. See also Getty Images (US), Inc. v. Stability AI, Ltd., D. Del., filed February 2023.
3. Cloudflare, 2025 Year in Review. https://blog.cloudflare.com/radar-2025-year-in-review/. Note: this figure covers all automated web requests, including training-time crawlers. Inference-time retrieval represents a smaller, commercially distinct, and faster-growing category.
4. Based on standard AI crawler behavior absent licensing agreements. See Authors Guild et al. commentary on AI content licensing, 2024.
5. Deloitte, Digital Media Trends Survey, 18th Edition, 2024. deloitte.com
6. Press Gazette / FT Strategies, Digital News Subscription Benchmarking Report, 2023–2024. Figures reflect mid-tier digital news publishers; premium tier churn is materially lower.
7. Apple Inc., iTunes Store launch press release, April 28, 2003. apple.com/newsroom
8. Ted Nelson, Literary Machines, 1981. Project Xanadu, originally conceived 1960. See also Wolf, Gary. “The Curse of Xanadu.” Wired, June 1995. wired.com/1995/06/xanadu
9. LedeWire traction data, Q4 2024. Tollbit network publisher figures as of active commercial pilot, 2025. x402 is an emerging machine-to-machine payment standard currently in active development.
10. McKinsey Global Institute, The Economic Potential of Generative AI, June 2023. Agentic commerce projection derived from McKinsey’s generative AI economic impact estimates, applied to the commerce-enabling subset of agentic activity.