Everything, newest first.

Writing for free is not a lesser use of this

We had been describing this platform almost entirely in terms of getting paid, and somebody pointed out the obvious problem with that. Most people writing online are not trying to earn anything. They have something to say, they want it read, and they want a following that belongs to them rather than to whoever runs the algorithm this quarter.

That is the oldest reason anyone has published anything, and a platform that treats it as a trial tier on the way to a paid plan ends up with worse writing on it.

So it is now written down as a rule rather than left as an assumption. A writer who charges nothing never receives a bill, for the rest of their life. Free work ranks identically to paid work, and nothing in the ranking looks at price, earnings, or whether a piece is monetised at all. Every distribution surface treats them the same. Readers can back a free writer directly without any paywall going up, and a writer who wants no money involved at all can switch that off.

The honest part is who pays the storage, and we changed our minds about this after writing it down the first way. We had the platform covering it out of what the paid side earns, which had a property we did not like: the better the free side did, the more it cost us, so our interests and the writer's pulled in opposite directions.

So instead, every article carries a support button and that is what funds it. New writers start with an allocation that covers them long before anyone has found them, and any piece can be paid off once so it never accrues storage costs again. Nobody is ever billed and no balance can go below zero. If a piece runs out of funding it stops being stored, after four warnings and a one-click export of the work and its composition record, and that is the whole consequence. Writing people value has people funding it, and no committee here is judging anything.

You will be able to buy a piece out of storage costs, for good

Storage on ITZA accrues a small ongoing cost, the way storage does everywhere. What is different is that the protocol lets you end it. Pay once for a piece and it is recorded on the chain permanently, with nothing left to pay.

Everything you have ever read online is still online because somebody is currently paying for it. It disappears when they stop, lose interest, or go out of business. That is the actual reason most of the internet's writing from fifteen years ago is gone, and it has nothing to do with anyone deciding it was not worth keeping.

To be clear about what it is and is not: this is a prepayment, not a change of control. The writer still decides whether a piece stays published and can unpublish it at any time, and a takedown works exactly as it did before. The only thing that changes is that nobody is paying rent on it any more.

No web host, no content network, no storage vendor

ITZA Press is not a site hosted somewhere that happens to settle payments on a chain. The application itself runs on the network. Every node carries it and serves it to readers directly over HTTPS, from whichever node is nearest.

That takes four companies out of the path in one move: the cloud provider, the content delivery network, the storage vendor and the payment processor. Almost nobody who has been removed from the internet was removed by a court. They were removed by one of those four, quickly and without appeal.

The guarantee is only as strong as the number of independent people running nodes, in enough places, under enough different jurisdictions. That part we cannot build. If you want it to hold, there is now a page explaining how to be part of it.

The full platform specification is finished

Twenty sections, covering the editor and composition record, direct reader to writer payment, the ratings system, escrow for subscriptions and commissions, arbitration, and the design of every screen we intend to build.

The section we spent longest on is the shortest. It lists fourteen rules the build is not allowed to break, regardless of what turns out to be convenient later. Among them: there is never a shared revenue pool, the composition record never acts on its own, publishing is always free and never requires identity, and subscriber export is never restricted.

Writing rules like that down before the first line of code is the only way they survive contact with a deadline. Anything that turns out to conflict with one of them is a specification bug to be raised, not a decision an engineer gets to make quietly at two in the morning.

Why there are no stars

We started with a five star rating and threw it out. Every five star system in existence drifts up to somewhere around four point seven and then stops telling anyone anything. Once that happens, ranking quietly falls back on popularity, which is the exact problem the rating was supposed to solve.

So readers are shown two pieces they have both finished and asked which was better. There is no absolute scale, so there is nothing to inflate. Ranking is fitted from the whole comparison graph.

The second decision took longer. Groups of real people agreeing to rate each other is a problem verification alone does not fix. The answer is that a reader whose comparisons consistently disagree with what other verified readers think of the same pieces gradually carries less weight. A ring can boost each other. It cannot make everybody else agree, and trying costs them their own influence. Nobody gets accused of anything and nobody gets banned.

How the composition record stays private

The obvious way to build a writing record is to keep every keystroke and its timing. We are not doing that, for two reasons.

The first is that a stream of keystroke intervals is close enough to a biometric identifier that several states treat it as one, with real consequences for storing it. The second is that we do not need it. What proves a record is genuine is a signature and a fingerprint, not the raw material.

So timings are used while you write and discarded when the session closes. What survives is a summary - session count, time at work, characters typed and pasted, how much was written and removed - plus a signature over a fingerprint of the underlying changes. The detailed change history is encrypted under a key the writer holds. We cannot read it.

There is a test in the build that fails if a raw keystroke stream is ever written anywhere, including in a debug build. It seemed worth making that structural rather than a matter of remembering.

We are not building a detector, and here is why we say so out loud

A composition record looks like it should be an authenticity test. It is not, and we would rather say that plainly than let anyone assume otherwise and feel cheated later.

A verified person can paste in whatever they like. Someone determined enough could script realistic typing, though it has to run in real time, which removes most of the point. Neither of those is what makes low quality synthetic writing profitable. What makes it profitable is scale: one operator, hundreds of accounts, one shared pot of money.

Take away the pot and require one verified person per account and the business stops working, without anyone having to be accused of anything. That is a smaller claim than the industry usually makes. It is also one we can actually keep.

Subscriptions will be held in escrow

Subscribing to a writer you have not read much of is a small act of faith, and the usual failure is not fraud. It is a writer who posts twice and goes quiet while the payments keep going out.

On ITZA Press the month you pay for sits in escrow. It releases to the writer as they publish, and whatever is left at the end of the period returns to you. No cancellation, no support ticket.

It protects readers, which was the point. It also happens to make backing an unknown writer far less risky, which is the single hardest problem a new platform has.

ITZA testnet is live

The infrastructure ITZA Press runs on is up. That matters here because the things this platform promises are not application features, they are properties of the layer underneath: escrow that holds without a company holding it, a ratings record nobody can quietly edit, and identity that proves a person exists without revealing who.

Testnet going live starts the clock on everything above it.

Why we started this

Someone we know spent a week trying to find a decentralised writing platform he could actually use. One made articles visible only to other members. One was buried in spam. One required buying tokens before a new account had enough allowance to publish anything substantial. One needed reputation to build reputation.

His conclusion was blunt and hard to argue with. Publishing should be one of the strongest cases for this technology, and every implementation had built a closed loop where you needed money, reputation, connections or an existing audience before you could begin.

Every one of those is the same mistake. Each platform needed a way to tell a real person from a script, could not do it, and substituted something it could measure. Money. Followers. Tenure. Each substitute became a wall around the people it most needed.

We had already built the missing piece for something else entirely. This is what happens when you point it at publishing.