I went looking for a draft list this afternoon. Substack's dashboard told me I had thirty-two unpublished drafts, sixteen of them touched in April. I asked my AI assistant to compare that draft list against my published list so I could see whether anything important was sitting unsent.
What followed was an hour of friction I did not budget for. None of it was anyone's fault. All of it was the same problem in different costumes.
The Substack MCP I use returned ten drafts and refused to return more, regardless of the limit I asked for. I dropped to the underlying API and got back twelve hundred and fifty entries — most of them duplicates of the same ID, because the pagination parameter was being silently ignored. I tried three other endpoints. One returned a 403. One returned a 400 with an unhelpful error. One returned the same first page over and over. After half an hour I stepped back and looked at what I was actually doing, which was reverse-engineering an admin UI that the dashboard already shows for free.
The dashboard had my answer. I had been ignoring it because the dashboard wasn't programmable.
This is the integration tax. We pay it constantly, and most of us have stopped noticing.
The shape of the tax
Every creator I know of in 2026 is the integration layer between five to fifteen platforms. A typical week looks something like this:
You write in some editor. You publish to Substack. You cross-post to your own site. You send a LinkedIn version. You send an X version. You maybe touch Threads and Bluesky. You consider Medium. You think about whether to send the email or not. You upload a cover image somewhere. You try to remember whether you already posted on a topic. You check what the algorithm seems to be rewarding this week. You glance at four sets of analytics that count things differently. You wonder whether your Substack subscription emails are even reaching the inbox tab anymore. You schedule something for Tuesday morning and forget which app you scheduled it in.
Each individual step is small. Each individual platform is reasonable. Each individual UI is, in isolation, fine.
The problem is that nobody is responsible for the seam. Substack doesn't know you also publish to Medium. LinkedIn doesn't know you sent the same post on X. Your X scheduler doesn't know your LinkedIn scheduler. Your Substack email metrics don't talk to your LinkedIn impressions. Your draft on Substack does not know that the same idea exists half-written in a Notion page from three weeks ago.
You hold all of that in your head. The platforms hold none of it.
This works for one platform. It works, with effort, for two. It starts to crack at three. By the time you're a real creator with five active distribution channels, you are spending more time integrating than writing — and you don't notice it because the integration tax is paid in tiny units, fifteen seconds here, two minutes there, half an hour spent reverse-engineering a draft list. The tax is invisible because it doesn't appear on any single bill.
What the tax pays for
It pays for the tools to not have to talk to each other.
That's the cleanest way to say it. Each platform is incentivized to be a closed loop. Substack benefits when creators write for Substack, with Substack's audience, in Substack's email format, optimized for Substack's discovery surfaces. LinkedIn benefits when creators stay native — LinkedIn rewards posts that don't link out. X has its own version of the same story.
So when you cross-post — when you behave like a creator with multiple platforms instead of a Substack-creator or a LinkedIn-creator or an X-creator — every platform quietly fights you. Substack will warn that links to your own site might hurt deliverability. LinkedIn will quietly down-rank posts that point to other domains. X will collapse your formatting and clip your length. Each platform is technically allowing the cross-post, but none of them is making it easy.
The integration tax exists because there is no commercial entity whose job is to remove it. Every platform is paid to keep you in. No platform is paid to help you operate across many of them gracefully.
What I do today
I have improvised. So has anyone reading this who is more than one platform deep.
I have an MCP server that posts to X. Another that posts to LinkedIn. Another that creates Substack drafts. A protocol document my AI assistant reads at every session that codifies pacing rules across all of them. A publication log that records every piece across every surface, manually maintained because no platform's analytics talk to any other platform's analytics. A "next session" handoff document so the next time I sit down, the queued amplifications fire in the right order with the right cooldowns.
It works. It also took months to wire up, exists nowhere outside my personal toolchain, and would be unrepeatable for anyone who didn't independently build the same scaffold.
When I try to look across the platforms — like just now, when I asked for a draft inventory — the seams show. The MCP returns ten because it caps at ten. The underlying API returns twelve hundred and fifty because it doesn't paginate the way I expected. The dashboard shows thirty-two because that's the count that actually matters. Three different numbers from the same publishing platform, all technically correct, all answering subtly different questions, none of them easily reconciled by a script.
I am the reconciliation layer. So is every creator with more than one platform. So is every small business with more than one tool.
This is the missing layer. A real one. The one that holds your intent across your tools and pays the integration tax on your behalf.
Why nobody has built it
Three reasons, in increasing order of how hard they are to fix.
The easy one is that platforms make it hard. APIs are inconsistent. Some are paid. Some require browser-fingerprinted sessions. Some change weekly. X's posting API costs a hundred dollars a month minimum. Substack's admin endpoints aren't documented. LinkedIn requires OAuth flows that most users will never complete. Building one integration is hard. Building all of them and keeping them working is genuinely difficult.
The medium one is that the integration layer doesn't fit any existing product category. It's not a CMS — you don't write in it. It's not a scheduler — schedulers are platform-specific. It's not an analytics tool — analytics tools report after the fact, not steer in the moment. It's not a CRM. It's not a marketing automation platform either, though it borrows ideas from there. It's something new: a single mental model of your published presence, with the actions that maintain it expressed once instead of five times.
The hard one is that AI just made it possible. Until very recently, you needed a software product with a user interface that mediated every action. Now you can have an AI assistant that holds your intent, knows your tools, calls the right APIs, and maintains the protocol. The product surface gets a lot smaller because most of the user experience is just talking to your assistant. That's a different kind of product. We don't have a lot of templates yet for what it should look like.
What it would actually do
Imagine an assistant that reliably handled the following on your behalf, without you having to think about any of it:
You finish writing. You hit save. The piece deploys to your canonical site within a minute. A draft is created on Substack with the correct subtitle and tags. A LinkedIn version is composed in your voice with the correct hashtags. An X long-post version is composed with the platform-specific gotchas already worked around. All of these wait in queue.
You decide which pieces are flagships and which are secondary. The assistant honors your pacing doctrine — email on flagships only, LinkedIn no more than one per seventy-two hours, X no more than one a day — and tells you when something would violate it. You override when you want; otherwise the assistant just runs.
When something publishes, the assistant updates your publication log. Tracks the URL. Notes which surfaces went live, which are still queued, which are blocked by cooldowns. Shows you a dashboard that is actually accurate across platforms instead of a per-platform fragment.
When you ask it "what did I publish last month" or "is anything from the November cluster still unsent," it answers in one breath. It doesn't make you reverse-engineer four different APIs to find out.
This is not exotic. Almost every piece of it is wireable today with the APIs that already exist. The reason it doesn't exist as a product is partly because the AI piece needed to mature first, and partly because nobody has done the long, unglamorous work of just sitting down and building it.
The honest part
I do not have this product. I have improvised pieces of it for myself, and writing this article makes me realize how thoroughly I have been treating my personal scaffold as a substitute for the missing product. That's a mistake.
A personal scaffold is fragile, opinionated, and not portable. It doesn't help anyone except me. It doesn't get better unless I improve it manually. It doesn't share knowledge across users. It can't aggregate the platform-specific gotchas that other creators have already discovered. It's a workaround, not a solution.
The Third Layer — the layer that holds your intent across your tools — should be a real product. Not something each creator has to assemble from API keys and prompts and Markdown documents. Something you install. Something that learns the platforms once and benefits every user from that knowledge. Something that turns the integration tax from a per-creator cost into a one-time engineering cost paid by the layer's authors and amortized across everyone who uses it.
That product is what I'm going to start building, and I'm announcing it here because writing the announcement is what makes me commit to it. I have improvised long enough. The improvisation has shown me the shape of the thing that should exist. The next step is to make it.
What you can do, if you're nodding
If you're reading this and recognizing the same friction in your own work, three things might help in the meantime.
First, name the problem. Stop accepting the integration tax as just "what publishing online is like." It isn't. It's a specific failure of the existing tooling to take responsibility for the seam between platforms. Naming it as a tax means you can decide whether to keep paying it.
Second, automate the parts you can. If you have the technical capacity, an AI assistant with an MCP server per platform gets you most of the way to a personal scaffold. It won't be the product. It will be an improvisation. But it will save you hours a week and teach you what the real product needs to do. The pacing protocol I shared in the previous article is one specific piece of the scaffold, focused on cadence; you can borrow it as a starting point.
Third, push on the platforms. Tell Substack what's missing. Tell LinkedIn that their throttle penalizes you for serving multiple audiences. Tell X that the long-post format needs to coexist with cross-posting tools. Vendors hear this. Some of them will eventually act on it, especially if enough creators name the same friction at the same time.
But none of those three is the real solution. The real solution is the Third Layer, built as a product. And until that product exists, the rest of us are paying the tax.
I have been paying it for a long time. I think I'm done.