Survivable Fragmentation

How independent media stays alive while the big networks consolidate, and where the gaps are that anyone who can build things could fill.

When the reporter you trust has been pushed out of a briefing room, off a network or out of a budget line, you end up following that person from one small platform to the next, a newsletter here, a livestream there, a social account that may or may not survive the year. That scattering is where independent media lives now, and most of the tools holding it together were built by companies with their own investors, their own politics and their own reasons to change the terms on short notice. Most of what keeps an independent outlet alive is plumbing. The plumbing is where the fight is.

What changed in September 2026

Three things happened within a few weeks of each other, and together they touch every way broadcast news reaches the public.

On September 21 the White House launched "Trump TV: The Essentials Station," a round-the-clock video channel on the White House website and its YouTube account, carrying material the administration selects and produces about itself. The launch followed the administration barring CNN, MS NOW and Politico, and three outlets have since sued over access to the White House. Critics called it state-run television, and the White House said the press had reported the administration's accomplishments inaccurately or not at all.

Paramount, run by David Ellison, is buying Warner Bros. Discovery for $31 a share in cash, with his father Larry Ellison backstopping roughly $47 billion in equity and royal funds from Saudi Arabia, Qatar and Abu Dhabi committing about $24 billion for a stake. A coalition of twelve state attorneys general led by California's Rob Bonta sued to block the deal and settled instead, and the merger is expected to close by early October. When it does, CBS, CNN, HBO and two of the largest film and television libraries in the world will answer to one owner.

Public broadcasting took the third hit. Congress rescinded funding for the Corporation for Public Broadcasting in 2025, and by early 2026 the corporation had begun dissolution proceedings. The Federal Communications Commission, chaired by Brendan Carr, who wrote the agency's chapter of Project 2025, spent 2025 reopening previously closed news-distortion complaints against broadcasters, and Project 2025 goes further, proposing that public stations lose their noncommercial educational status.

A state channel, a consolidated private network and a defunded public one arrived in the same season. None of that requires reading anyone's intentions, because the record is public, and Poynter keeps a running count of which Project 2025 media proposals have been carried out, which is the fairest way to separate what was proposed from what was done before you decide how alarmed to be.

The goal is a cheap exit

A platform that cannot be corrupted does not exist, and chasing one wastes years. Every service an independent outlet depends on sits on a stack of other services, and each of those is a place where a lawsuit, a regulator, an acquisition or a change of heart can land. Bluesky is a company with investors. Substack raised about $100 million in 2025 from venture firms rather than readers. A Mastodon server is often one volunteer, and it goes dark when that person burns out or receives a threatening letter.

What you can build toward is a cheap exit. The test is whether a journalist, on the day a platform turns against them, can leave with their audience, their archive and their income intact in a single afternoon. A platform that knows its writers can leave that easily has to keep earning them, and a platform that holds the only copy of the subscriber list owns the writer whatever its politics happen to be.

Judge every tool by those four and most of the choices that follow make themselves.

Where the pressure lands

Decentralizing one layer while the others stay centralized buys very little, and that is the most common mistake people make when they set out to protect a publication. Think of an outlet as seven layers, each held by someone who can say no.

Payments are the hardest layer to route around. Stripe, PayPal and the card networks behind them are not banks, no comprehensive federal consumer protection law governs how they treat the businesses they serve, and their terms of service let them close an account at any time for any reason. A payment ban is total, because it cuts off revenue across every channel at once, and outlets from every corner of the political map have lost their processors before.

App stores come next. Apple and Google between them decide what reaches a phone, and there is no meaningful appeal when either says no.

Domain registrars and DNS providers can be routed around with preparation, and sites have been dropped at this layer and come back. Hosting and content delivery are portable if you planned for a move and painful if you did not. Identity, meaning the handle or address readers follow, is locked to the platform almost everywhere, with Bluesky's protocol the notable exception.

Discovery is the soft layer, and it decides more than any of the others. A platform does not have to ban anyone to silence them, because it only has to stop recommending them.

Legal exposure is not technical at all. A defamation suit with no merit can still bankrupt a one-person outlet through legal fees, and the protection you have depends on whether your state has a strong anti-SLAPP law, the kind that lets a judge dismiss such a suit early and make the plaintiff pay your costs.

Of all seven, payments get the least attention from the people who build tools for independent media, because the work is regulated and dull. That neglect is exactly why the need is greatest there.

Going live on several platforms at once

When Lev Parnas began streaming on Kick alongside his Substack this year, pitching it as a way to reach a younger audience that would never find him otherwise, the arrangement looked new. The technology underneath has been routine for about ten years, and anyone with a laptop and a decent connection can do the same thing tonight.

Almost every live platform accepts video the same way, through a protocol called RTMP. The platform gives you a server address and a secret stream key, you paste both into broadcasting software such as the free OBS Studio, and the software sends your video to that address. Kick hands out a single global ingest address, a hostname that belongs to Amazon's IVS streaming service, with no region to choose. YouTube, Twitch and Facebook work the same way. Substack added RTMP input as a beta for selected creators, with one catch that trips people up, which is that it issues a fresh key for every stream, so the key from last week will not work tonight.

Going live on several platforms at once, which streamers call multistreaming or simulcasting, means encoding your video once and sending a copy to every address. There are two ways to do the copying.

Copying from your own computer

A free OBS plugin forks the encoded stream and pushes a separate copy to each platform, with Aitum Multistream the common recommendation in 2026 and the older obs-multi-rtmp still working. Nothing costs money and nothing depends on a third party, but every copy leaves through your own internet connection, and upload speed is where home setups fail. Four platforms at six megabits per second each is twenty-four megabits of sustained upload, and the usual guidance is to have half again as much headroom, so you need roughly thirty-six megabits that hold steady for the whole broadcast. Local multistreaming is generally recommended only above fifty. Cable connections often upload a small fraction of what they download, so measure yours on a wired connection with any VPN switched off before you plan around it.

Copying through a relay

The alternative sends one stream to a relay service, which copies it to every platform from its own servers. Restream is the best known, and its free tier covers two platforms with its branding laid over your video, while paid plans lift both limits. You trade money, a dependency and a third party sitting in the middle of your broadcast for an upload that only has to carry one copy. For anything with an audience that matters, the relay is the safer choice, because a home connection that sags halfway through a broadcast is the most common live-streaming disaster and the one you cannot fix while it is happening.

A few rules differ by platform. Kick publishes no restriction on simulcasting, although its Partner program can reduce payouts for streamers who multistream. Twitch used to forbid simulcasting and has relaxed that rule. Instagram Live and TikTok are built for vertical phone video and fight a horizontal studio setup, so most multistreamers leave them out.

You can rehearse all of this with nobody watching. YouTube lets you mark a stream unlisted or private, every platform lets you delete the recording afterwards, and a first test on accounts with no followers reaches no one. Ten minutes of your desktop sent to three platforms at once will teach you more about which one lags and which one drops frames than any guide can.

The channel nobody can take away

Owncast is free, open-source software that turns a server you rent for a few dollars a month into your own live channel, with its own chat, and it can notify followers on Mastodon when you go live. Add it as one more destination and you have a stage no platform can revoke. Owncast also teaches the hard limit of owning your own video, because it sends every viewer a separate copy straight from your server, and ten viewers at three megabits each is thirty megabits leaving your machine. Self-hosted live video does not scale on any connection an individual can afford, which is why the large platforms still matter and why PeerTube, the federated alternative to YouTube, struggles with the cost of bandwidth.

Choosing a home for the writing

The newsletter platform you choose decides two things that matter more than its feature list, namely where it sends you to find readers and what it costs you to leave.

Substack costs nothing until you charge readers, then takes ten percent of subscription revenue on top of card fees, and setup takes minutes. What the ten percent buys is distribution, because Substack routes writers into its own app, its Notes feed and a recommendation network that sends new readers your way, and that network is the reason writers stay somewhere they distrust. The platform has lost writers over its refusal to remove extremist newsletters, a fight that reopened in February 2026 when a Guardian investigation reported Substack earning money from far-right and antisemitic publications, and some of its biggest names, The Ankler among them, have moved to rivals over the fee and the loss of control.

Ghost takes nothing from subscription revenue. The software is open source and owned by a non-profit foundation, you can run it yourself on a rented server or pay Ghost from roughly fifteen to eighteen dollars a month to host it, and your publication lives on your own domain with your own design. Since version 6.0 a Ghost publication joins the fediverse through ActivityPub, so people on Mastodon, Threads, Flipboard and a growing list of other services can follow it without signing up for anything. No other option here scores as well on both capability and resistance to capture. The price is distribution, because Ghost gives you ownership and takes away the discovery network, and growth is entirely your own work.

Beehiiv also takes nothing from subscription revenue, although charging readers requires its paid plan at about forty-nine dollars a month, and its free tier covers 2,500 subscribers. Its strength is growth machinery, with referral programs, paid recommendations between newsletters, a built-in ad network and discovery tools that Ghost does not offer. That machinery is also the lock-in, and Beehiiv is a venture-funded company.

Buttondown is the smallest and simplest, free up to 100 subscribers and inexpensive after that, and it is an email tool rather than a platform, with nothing to discover and very little to leave. ActivityPub support is on its way there too.

The trade across all four is the same. Substack sells you readers and charges you independence for them, Ghost sells you independence and leaves the readers to you, and the others sit in between. If you already have an audience, Ghost is the strongest home. If you have none, Substack's network is hard to replace, and the sensible course is to use it while exporting your list regularly so the exit stays cheap.

Social and video

For social media, the two open options fail in opposite directions. Bluesky is built on the AT Protocol, which ties your identity to your account rather than to a server, so you can use a domain you own as your handle and move your entire following from one host to another without starting over. That is the best portability in social media by a wide margin, with a large catch explained below. Mastodon and the wider fediverse run on ActivityPub, where thousands of independently run servers talk to each other and no company sits at the center, but your identity belongs to the server you joined, and when a volunteer-run server shuts down, moving elsewhere loses things and is sometimes impossible.

For video, YouTube remains the archive and the discovery layer whatever you think of Google, which is worth remembering given that the White House chose YouTube to host its own channel. Kick skews younger and imposes no simulcast restriction. Twitch has the largest live audience. Owncast and PeerTube are the self-hosted alternatives, excellent as insurance and hard to scale as a main stage.

What your subscriber list is, and who holds it

If you write on Substack, your list is yours to take. From the Subscribers page, the menu above the list of all subscribers offers an export that downloads a spreadsheet of every reader's email address, their status as active, paused or cancelled, the date they subscribed and whether they pay. Posts export separately. That export is what let writers leave Substack when they wanted to, and whatever else is true of the platform, it works.

What Substack does not show you is who left. The dashboard reports your current list and has no view of unsubscribes, so churn stays invisible unless you export on a schedule and compare the files, or connect Substack to a separate email service that records each change as it happens. The number that would tell a writer whether something they published drove readers away is the one number the dashboard leaves out.

If you read on Substack and the volume of email has become unmanageable, you do not have to unsubscribe to stop it. In the Substack app, open your profile, go to Settings, then Notifications, then Newsletter delivery, and choose Prefer push. Every subscription stays in place, the email stops, and you read in the app instead. Individual publications can be adjusted from their Manage subscription page, though readers report the per-publication settings are confusing, and the global switch is the reliable one.

That fix treats a symptom. Anyone who supports a few hundred independent writers ends up with thousands of emails every few days and reads almost none of them, which is the reader's side of fragmentation, and nobody has built a good answer to it yet. Substack's answer is to read everything inside its app, which pulls independent media back into one venture-funded company and recreates the problem it was escaping.

Decentralization theater

Bluesky's portability is real, and it buys less than it appears to, for reasons that come down to how the network is assembled.

The AT Protocol splits a social network into three parts. A personal data server holds your posts, follows and identity, and anyone can run one, since Bluesky publishes a ready-made container for it, although early-access guidance caps a self-hosted server at about ten accounts, 1,500 events an hour and 10,000 a day. A relay gathers every event from every personal data server into one continuous stream. An AppView reads that stream and builds what people see, the timelines, threads, notifications and search results.

The protocol allows as many relays and AppViews as anyone cares to run. As of early 2026, Bluesky the company operated the only AppView covering the whole network and one very large relay that nearly everyone used.

That arrangement means you can run your own server, own your handle as a domain and hold every post you have written, and still reach no one except through Bluesky's infrastructure. Whoever runs the relay decides what enters the network's field of view. Whoever runs the AppView decides what gets shown, ranked or quietly pushed down, and if that AppView fails or turns hostile, every app built on it breaks at once, including the one showing your self-hosted posts. An assessment published in October 2025 called this decentralization theater, an architecture that permits decentralization running as a centralized service with an escape hatch nobody has used.

The defense is fair as far as it goes. Every component can be replaced, and an exit that exists but has not been used is different from no exit at all. Blacksky, an independent operator running its own AT Protocol infrastructure and moderation for Black users, shows the hatch can be opened. Running an AppView for the whole network takes serious money and engineering, though, and that cost is where the centralization hides.

The measure to watch is not whether the protocol is open, because it is. Watch how many independent relays and AppViews run in production and whether any of them pays for itself, a count nobody publishes. Set beside Mastodon, the pattern is plain enough. Mastodon has hundreds of independent servers and weak portability, Bluesky has excellent portability and essentially one operator, and neither has solved the problem, having failed in opposite directions.

What the co-ops teach

If you want a newsroom that cannot be bought, the strongest defense found so far is not a technology. Ownership by the people who work there leaves nobody to buy it from and nobody above the staff to pressure.

Most writing about media cooperatives blurs two very different structures, and separating them explains nearly every success and failure. In a worker-owned cooperative the journalists own the outlet, usually somewhere between five and thirty people, all of whom work there every day and vote on its decisions. In a community-owned cooperative readers buy shares and become owners, sometimes hundreds or thousands of them, none of whom work there. The successes are almost all worker-owned. The collapses are almost all community-owned.

The ones that worked

Defector, started by former Deadspin staff, brought in about $3.2 million from more than 40,000 paying subscribers in its first year and about $4.6 million in the year to August 2024, and it publishes its finances every year. 404 Media, founded by four reporters from Motherboard and modeled on Defector, was reported profitable by February 2024. Hell Gate, a worker-owned New York local outlet, doubled its subscription revenue in its second year behind a hard paywall and a redesign that made subscribing easier. Each launched with an audience its founders brought from a previous job, charges readers directly, took no venture money and has no board above the staff.

Defector's figures give you a number worth working out for yourself. At roughly $80 per paying subscriber per year, divide the revenue by the staff count in its published report and you get the number of paying readers it takes to support one journalist's salary, which is the basic arithmetic of independent media and a number almost nobody quotes.

The ones that failed

The Devil Strip in Akron is the failure worth studying closely. The alt-weekly converted to a community-owned cooperative in 2019 and was celebrated for it, then suspended operations in October 2021 without warning. Its owners had not been told it was in trouble, let alone consulted about what to do. The nine staff learned on a Friday that there was no money for payroll past that week, most of the board had already resigned, and the staff were laid off. Forensic accountants later put the debt at about $186,000, close to $80,000 more than the board had acknowledged, run up both before and after the conversion, and the investigators could not get bookkeeping or tax records from before 2020. The bylaws had made governance close to impossible, requiring a simple majority of dues-paying members for a quorum, which meant assembling 356 people to make a business decision.

The Chicago News Cooperative, which produced Chicago coverage partly for the New York Times, shut down in February 2012 when the Times declined to keep funding it. A cooperative that depends on one patron is a client. NewsTilt, a startup-backed platform for independent journalists launched in 2010, was gone within months, and it stands as the model case of technologists building a platform for journalists that journalists did not need.

The patterns

Governance gets worse as owners multiply, unless the owners also do the work. Bylaws get written to sound democratic and turn out to be mathematically unworkable, so every rule deserves testing against whether five tired people could carry it out on an ordinary Tuesday. Financial secrecy is the killer, because in every collapse the owners learned the real numbers only when it was over, while every success publishes them voluntarily. Dependence on one funder passes for independence until the funder leaves. Debt carried into a conversion lands on owners who never saw it. A cooperative structure is also a way of owning an audience, not a way of finding one, and that misunderstanding sits behind most failed starts.

Starting one, or joining one

If you set out to start one, begin worker-owned and small, under fifteen owners who all work there, and give readers membership rather than ownership, with subscriptions, perks and a voice but no vote, so you keep their loyalty without making the outlet ungovernable. Bring an audience or wait until you have one.

For the legal form, a small group usually chooses between an LLC run cooperatively, which is flexible and taxed as a pass-through, and a cooperative corporation, which is more formal and carries statutory protections in the states that offer them. Massachusetts, New York, Vermont, Connecticut, Maine and Ohio are among the states with laws written specifically for worker cooperatives.

Bylaws should never be drafted from a blank page. The Sustainable Economies Law Center publishes free legal templates, the ICA Group publishes model bylaws for worker cooperatives with one or two classes of shares, the Democracy at Work Institute offers a toolkit for writing participatory bylaws, and the University of Wisconsin Center for Cooperatives publishes a guide to bylaws and operating agreements. Have a lawyer review whatever you adapt.

If founding is premature, joining is usually the better start. Existing cooperatives need help with the work surrounding the journalism, and several small outlets sharing that help is an arrangement almost nobody has set up.

Fragmentation is the defense

A common complaint holds that the left is fractured while the right has a deeply funded set of platforms moving in lockstep. The fracture is real, and the lockstep is not the advantage it looks like.

Consider what consolidation bought. When one company owns CBS, CNN and HBO, a single change of ownership changes the editorial posture of all of them at once, and that change closes next month. A captured hub captures everything downstream of it. A centralized message cannot adapt to local conditions and cannot route around its own mistakes. A network of independent outlets has the opposite profile, with no single target to buy, no single legal chokepoint and no single editorial line to capture. Consolidated media is built for offense. Distributed media is built to survive, and when the threat is capture rather than competition, survival is the property that wins.

Fragmentation feels like weakness because it imposes costs that consolidation pays once, centrally, and a distributed network pays over and over. Readers cannot find the outlets. Readers who do subscribe drown in them. Every outlet rebuilds the same hosting, email and archive from scratch. One outlet sued alone loses, where twenty sharing legal resources would not.

Every one of those is an infrastructure problem, not an ideological one, and none of them requires anyone to agree with anyone else. Independent media does not need to unify. It needs the pieces to work together.

Federation is that idea expressed as technology, with many independent nodes, no central authority, shared protocols, and a layer on top that gives readers a coherent view without handing anyone control. Consolidation gets coherence by eliminating independence. Aggregation gets coherence while keeping it, and nobody has built the aggregation layer.

Bad-faith accounts and coordinated campaigns inside progressive spaces are a documented problem too, and researchers have working methods for spotting them through account age, posting patterns and coordination across platforms. Treat every disagreement as infiltration, though, and a movement will tear itself apart faster than any opponent could. The useful skill is telling the two apart with evidence.

What nobody has built yet

Several newsletter platforms, two social protocols, a good self-hosted streaming server and free cooperative legal templates already exist, and building another of any of them helps nobody. The gaps sit elsewhere.

A reader's aggregation layer

Nothing takes in Ghost sites, Substacks, Mastodon and Bluesky accounts, YouTube channels and podcasts and turns them into one prioritized daily read. Every one of those sources already publishes a feed, since each Substack has one at its address followed by /feed and YouTube channels have them too, so the raw material is sitting there. RSS readers list items one after another, while this would need to group five outlets covering the same story into one entry with five links and rank by importance rather than by time. Such a tool would host nothing, publish nothing and moderate nothing, every item would link back to the outlet that wrote it, and if it disappeared, every reader would still have every subscription.

Exit-cost auditing

No tool answers the question a writer should ask before choosing a platform, which is what leaving would cost. Every part of the answer can be checked from outside without access to anyone's account, including who registered the domain and where its DNS is hosted, which company handles its email and whether that is the same company as its publishing platform, whether it publishes a feed that makes its writing exportable, which payment processor its checkout uses, and whether the Wayback Machine holds copies of its pages. An outlet with no archived copies is one lawsuit away from disappearing.

Continuous financial transparency for co-ops

Every collapse described above was a failure of financial openness, and the successful co-ops publish their numbers by hand, as a yearly writing project. A dashboard that reports revenue, costs and runway to every worker-owner automatically is ordinary software. Nobody has built it, because it sits between the cooperative movement and software development, and few people stand in both.

Shared infrastructure for small outlets

Ten independent publications each handling hosting, email deliverability, backups and security badly, none of them able to afford a systems administrator, could easily afford one together. That arrangement is a services cooperative, not a platform.

Archiving that survives a lawsuit

ArchiveBox and the Wayback Machine exist, but nothing mirrors a small outlet's work automatically to several places in several jurisdictions as part of publishing, so that a court order in one place cannot erase it.

A public count of independent infrastructure

A small, inexpensive tracker of how many AT Protocol relays and AppViews are running, and who pays for them, would give the whole field the one measure it lacks.

Email that reaches the inbox

A newsletter list is the only channel where the publisher holds the relationship with the reader, and keeping mail out of spam folders depends on authentication records called SPF, DKIM and DMARC and on a sending reputation that takes months to build and can be lost in a week. Few small outlets understand any of it, and helping them costs very little.

Where a builder fits

If you can write software but you are neither a famous broadcaster nor a donor, you are not locked out of this fight. The personalities and the money are necessary, and neither one is the bottleneck. Independent media is splitting apart faster than anyone is building the infrastructure that would make the split survivable, and most independent outlets fail on operations, not on journalism or politics.

Resist building a new platform. That graveyard is enormous, because platforms fail on trust and network effects rather than technology, and one person cannot supply either. Build what makes leaving cheap and what makes a fragmented ecosystem as usable as a consolidated one, and start with a problem you have yourself, something useful to one person on its first day, that fails gracefully and needs nobody's permission.

Then find one working journalist, small and local rather than famous, since people with audiences already have help, and offer them one specific thing. A general offer of tech help gets ignored every time. An offer to set up automatic archiving of everything they publish to the Wayback Machine and to a mirror on a domain they control, free, taking an hour and asking nothing of them, gets accepted, and one working relationship will correct more wrong assumptions in a month than a hundred hours of reading.

The parts of this that will keep independent media alive are the dull ones, payments, email deliverability, archiving and worker ownership. The exciting version of this project fails. The dull version works.

Sources

• Variety on the Trump TV launch
• The Hill on Trump TV and the access lawsuits
• CNN on the Paramount and Warner Bros. Discovery settlement
• Poynter on which Project 2025 media proposals were carried out
• Brookings on the FCC under Brendan Carr
• FIRE on payment processors and free speech
• Substack on live video via RTMP
• Streams Charts on multistreaming rules in 2026
• Upstream on OBS multistreaming and relay options
• Owncast
• Revenue Memo on who owns Substack
• Techbuzz on writers leaving Substack
• TechCrunch on Ghost joining the fediverse
• Substack on exporting your email list
• Bluesky's personal data server
• Bluesky on federation for self-hosters
• Rethinking Bluesky's decentralization, January 2026
• How decentralized is Bluesky really?
• Blacksky, Bluesky and the ATmosphere
• Columbia Journalism Review on 404 Media and worker ownership
• Nieman Lab on the Devil Strip's closure
• Ideastream on the Devil Strip's final accounting
• Nieman Lab on working inside a journalism co-op
• Chicago News Cooperative
• Sustainable Economies Law Center legal templates
• ICA Group model bylaws for a worker cooperative
• University of Wisconsin Center for Cooperatives on bylaws and operating agreements