The Copy-Paste Tax: Why People Become the Orchestration Layer
When ChatGPT lives in one tab and Claude in another, you become the router. The copy-paste tax is that unpaid job.
By the flAim team
The copy-paste tax is the unpaid job of shuttling text between models. You become the orchestration layer: slow, expensive, and made of command-C.
It shows up the moment the work spans labs. ChatGPT for a first pass. Claude to critique. Gemini because the doc is long. None of those products is the problem. The gap between them is.
A research → outline → draft loop, taxed
Say you need a one-pager on a messy topic. Honest Tuesday workflow without a workspace:
- Ask ChatGPT to list sources and claims. Copy the useful bits into a notes file.
- Paste into Claude: “punch holes in this, flag weak citations.” Copy again.
- Paste into Gemini with the original PDF. Copy the outline.
- Paste a merge prompt back into ChatGPT. Slack a screenshot to a teammate who cannot see the thread.
Four tools, one brain, zero shared state. The teammate re-derives the brief. The models never see the same notes. You paid the tax in minutes and in mistakes.
The same loop in a workspace
| Step | Tabs + clipboard | One workspace |
|---|---|---|
| Research | Model A tab. Notes in a local file. | Research agent writes into shared notes. |
| Critique | Paste into model B. Hope you copied the latest version. | Ask the critic in the same thread. It already saw the research reply. Or fan out. |
| Outline + draft | Paste into model C. You merge by hand. | @ one agent, then another, or send to everyone. Or Build to chain and synthesize. |
| Share | Screenshot. “Trust me, Claude said this.” | Invite. They see agents, thread, and notes. |
Why people keep paying it
Because each chatbot is good at the one-question job. ChatGPT is great at that. The tax is not a model quality problem. It is a systems problem: context lives in the wrong place (your clipboard), routing lives in the wrong place (your head), and collaboration lives in the wrong place (Slack).
We wrote the origin story in Why We Stopped Using ChatGPT. This post is the name for the cost: copy-paste tax.
flAim’s answer is unglamorous. Put the agents in one room. Put the brief in shared notes. Use Chat to steer one thread across models — @ A, then @ B, then both. Use Build when you want ordered handoffs. The orchestrator routes reasonably. It does not magically pick the perfect agent every time — and we will not sell it as if it does.
Stop being the clipboard
Open one workspace, add the agents you already use in separate tabs, and run the loop once.
Related: Why we stopped using ChatGPT · flAim vs ChatGPT · Pipeline playbook
We're building this in the open and we read everything. If you try it and something's clumsy, tell us. We'd rather hear it from you than not hear it.