The Subscription Model Maxed Out

Carney's case is arithmetic before it is grievance. A living on reader support requires roughly a thousand people at $8 a month. The number of newsletters a reader can actually hold in a week is closer to five, and the wallet is already committed to streaming, video, and one or two legacy publications. Writers therefore do not compete for a growing pool; they compete for cancellations.

“On a good day my brain can only keep up with about 5 newsletters. Meanwhile my wallet is skeptical of the value of even paying for that number.”

01 · The constraint is cognitive, not moral

The instinct in these conversations is to reach for character: readers have stopped valuing writing, platforms have stopped valuing writers. The cheaper explanation is a ceiling. Subscription revenue per reader is bounded by attention hours, and attention hours are bounded by biology. No amount of quality raises that ceiling, which is why Carney's honest admission — that his work has improved while his subscriptions have not — is the most diagnostic line in the post.

Established: per-unit subscription pricing scales with the number of products a household can consume, and that number is small and roughly fixed. Working claim: beyond a threshold, adding writers to a subscription platform redistributes revenue rather than increasing it.

02 · The feed pivot inverted the incentive

The second half of the argument is the part this project has been describing from the inside. When a platform built to sell considered work adds a short-form feed, live video, and machine-generated clips, it does not merely add a channel — it changes what the reader is trained to do on arrival. Free, fast, and agreeable content sits in front of slow, paywalled, resistant content, and the interface teaches the habit that pays the platform.

That is the same mechanism as The Interface Changed How We Learn, transposed into publishing: fluency without friction. The paywall then compounds it. It puts the work with the most resistance behind the highest barrier, so the essays that would build a reader's tolerance for difficulty are precisely the ones fewest readers meet.

No villain is required, and I decline the one on offer. Valuation grows with engaged users; author income grows with paid conversion. Once those two metrics diverge, a rational operator optimises the one investors price. That is an incentive shape, the same argument made in The Human-Centered Alibi and The Profit Case for Humane AI.

03 · Bundling is the structural fix, and it has a cost

Carney's proposed remedy is the right shape: let twenty adjacent writers pool into one $15 subscription, which is what a newspaper was — a bundle that let unprofitable rigour ride on profitable circulation. Bundles raise total revenue because they price against a household's single decision rather than twenty of them.

The cost is editorial. A shared pool needs a split rule, and any split rule is a reward function. Weight it by clicks and the bundle reproduces the feed it was meant to escape; weight it by editorial judgement and someone must be empowered to refuse. This is the reward-function problem from The Reward Function of Meaning arriving as a payments schedule: whatever the bundle measures is what the bundle will eventually publish.

04 · Sponsorship, stated plainly

Carney's own resolution is sponsorship with declared exclusions — no supplements, no crypto, no gambling, no wellness. That is the correct structure for the disclosure: a published refusal list is auditable in a way that a stated commitment to integrity is not. It is the same test this project applies to corporate humane-AI language in the Due Diligence Field Guide: a constraint counts only when refusing it costs the author money.

05 · Where this site stands

The relevant note here is not strategic advantage but a different exposure. This archive is not paywalled and does not depend on conversion, which removes the incentive to soften an argument for retention — and removes the revenue that would let anyone work on it full time. That is a trade, not a solution, and it only works because the work is done as citizen-scientist service rather than as a product.

What follows practically is smaller than a business model. Publish the rigorous work where it can be read, keep the machine-readable copy open — this site's works index and MCP endpoint exist for exactly that — and treat platform reach as weather rather than as a verdict on the writing.

Falsifiers

• If platform-wide paid revenue per writer is shown to have risen through the feed expansion, the cannibalisation claim in section 02 fails and the decline is writer-specific rather than architectural.
• If curated multi-writer bundles are tried at scale and total household spend does not exceed the sum of standalone subscriptions they replace, section 03 is wrong about bundling economics.
• If readers demonstrably sustain more than a handful of paid text subscriptions once discovery improves, the attention ceiling in section 01 is a pricing problem, not a biological one.
• If sponsor-funded independent work shows no measurable editorial drift relative to reader-funded work over several years, the disclosure argument in section 04 is precautionary rather than necessary.

Return to the essay index — or continue to The Interface Changed How We Learn.