The Real AI Gap in Cross-Border Ops Isn’t Model Quality — It’s Where the Cursor Lives
Every cross-border operator I know runs the same broken loop a dozen times a day: copy a supplier reply out of Alibaba, paste it into a chat window, ask for a cleaner English version, paste it back into email. Copy a German buyer message out of Amazon Seller Central, translate it, paste the reply back. Copy a policy paragraph out of a PDF, ask whether it exposes you to a return claim, paste the answer into Slack. The models are good enough. The workflow is the bottleneck. That’s why Edyt, launched this week by Ahmad Ahandani, is worth ten minutes of your attention even if you never install it — it’s a clean articulation of where AI tooling is heading for people who live inside fifteen tabs of seller dashboards.
What Edyt Actually Does (And What It Doesn’t)
The pitch is deliberately unglamorous: select text anywhere on Mac or Windows, hit a shortcut, pick rewrite, explain, translate, or one of your own saved prompts, and the result replaces the selection in place. If the text isn’t selectable — a PDF, a screenshot, a shared screen — you hold Option (or Alt) and drag over it, and the text is read on-device. The maker frames the whole thing around a single complaint: most AI tools come with a boundary — their own window, the browser only, one app, one OS — and the moment you step outside it you’re back to copy-paste.
For a cross-border seller, that boundary is not abstract. It’s the difference between a tool your ops team uses and a tool they abandon in week two. Your VA in Manila is working inside Seller Central, a supplier portal, a Google Sheet of tracking numbers, and a WhatsApp Web thread simultaneously. An AI that only lives in a browser extension covers maybe 40% of that surface. An AI that lives at the OS level covers all of it.
What you get on the free plan is roughly 130 quick edits a day, no credit card. The PRODUCTHUNT code takes 40% off the first year of Pro. Selected text is sent securely for processing and, per the maker, never stored; screen reading never leaves your machine. Linux and Ubuntu support are on the roadmap but not shipped.
The saved-prompts feature is the actual product
Buried under the toolbar bullet list is the thing that matters for teams: your saved prompts live on your account, and they follow you across your own machines. Ahandani is explicit that this is why he built it — he got tired of retyping the same instruction. For an operator, that’s a template library. Write the prompt once — “rewrite this supplier message in neutral business English, keep the SKU codes intact, don’t add pleasantries” — and it’s one keystroke away in every app, on every machine you sign into.
How It Stacks Up Against What You’re Probably Already Using
Let’s be honest about the competitive set, because Edyt is not entering an empty room.
Apple Intelligence Writing Tools is the most direct comparison, and a commenter on the launch thread asked exactly that. Ahandani’s answer is the right one: Writing Tools gives you a few fixed styles, whereas Edyt lets you write the instruction yourself, once, in plain words, and it stays yours. If you’ve ever tried to get Apple’s “Professional” tone to stop sounding like a LinkedIn motivational post, you understand the gap. Writing Tools is also Mac-only, which kills it for any team with Windows machines — which is most cross-border ops teams I’ve seen.
ChatGPT desktop and Claude desktop are the incumbent habit. They’re more capable on complex reasoning, they have memory and projects, and your team probably already pays for them. But they still require the copy-paste round trip unless you’re working inside a supported integration. The friction is small per instance and enormous per day.
Browser extensions like the various ChatGPT-for-Chrome tools cover web apps only. If your workflow is 100% Seller Central and Shopify admin, fine. The moment you touch a PDF spec sheet, a screenshot from a QC report, or a desktop ERP, they’re useless.
Raycast AI and similar launcher-based tools are the closest philosophical cousin on Mac — OS-level, shortcut-driven, prompt-library-driven. They’re excellent and Mac-centric. Edyt’s differentiator is the Windows parity plus the screen-reading drag gesture for non-selectable text, which is a genuinely novel bit of UX for anyone who’s spent time squinting at a scanned supplier invoice.
Why Amazon sellers should care more than Shopify ones
Here’s my read: the Shopify operator lives in a browser. Admin, Klaviyo, Meta Ads Manager, a helpdesk — it’s all web. A browser extension or a well-built ChatGPT sidebar covers 90% of their day, and the marginal value of an OS-level tool is real but modest.
The Amazon FBA seller lives in a fragmented desktop. Seller Central in a browser, but also: supplier emails in Outlook or Gmail desktop, a downloaded inventory report in Excel, a PDF of the FBA prep guide, screenshots from a freight forwarder’s WhatsApp, a desktop ERP or accounting package. The copy-paste tax compounds across every one of those surfaces. Same logic applies to Temu and SHEIN sellers juggling supplier portals that were clearly designed in 2014, and to eBay sellers running legacy desktop listing tools.
If your ops team’s day is more than half outside the browser, an OS-level text tool is worth more to you than another SaaS subscription inside the browser.
What Cross-Border Sellers Can Actually Borrow From This Launch
Three things, and none of them require you to install Edyt.
First, build a prompt library as team infrastructure, not personal habit. The single most valuable thing in this launch is the observation that people retype the same instruction dozens of times a day. Your team almost certainly has five or six recurring text tasks: rewriting supplier messages into clean English, translating buyer messages from DE/FR/JP/ES, drafting return-policy responses in a consistent tone, summarizing a long buyer complaint into a one-line ticket title, checking a listing description against a compliance checklist. Write each one down as a reusable prompt. Whether you execute them in Edyt, ChatGPT Projects, or a Claude Project, the artifact is the same and the compounding is real. Ahandani flagged that prompt sharing across teammates isn’t built yet — saved prompts are tied to your account and follow you across your own machines only. So for now, the library lives in a Notion doc and everyone pastes from it. That’s fine. It’s still a 10x improvement over retyping.
Second, treat “works on text you can’t select” as a procurement criterion. This is a small feature that reveals a big principle. Cross-border ops is full of locked text: scanned commercial invoices, screenshots from QC inspectors, PDFs of carrier rate cards, images of competitor listings. Any AI tool you buy for the team should have an answer for that surface. If it doesn’t, you’ll be paying for something that covers 60% of the job.
Third, watch the MCP question. Ahandani raises it himself and admits he’s going back and forth on it: should a prompt be able to call MCP servers, so that selecting a customer name pulls their last three tickets, or selecting a spec sentence checks it against your Linear issues? His hesitation is that people won’t set it up for a text tool. I think he’s half wrong. For a solo blogger, sure. For a cross-border brand running a support desk, a 3PL integration, and a supplier management system, the ability to select a buyer’s name and have a prompt surface their order history and last two tickets is not a gimmick — it’s the whole ballgame. The moment a text-selection tool can reach into your stack, it stops being a writing assistant and becomes an ops console. Watch this space.
Where the math breaks
Let me be the skeptic for a paragraph. The free tier at ~130 edits a day is genuinely generous for an individual, and that’s the right call for adoption. But the pricing model creates a predictable cliff for teams. A support team of five handling 400 buyer messages a day will blow through individual free tiers immediately and needs five Pro seats. At that point you’re comparing against a ChatGPT Team plan that also gives you document analysis, image generation, and a shared workspace. Edyt’s advantage — OS-level, no copy-paste — is real but narrow. Whether it justifies a separate line item depends entirely on how much of your team’s day is spent in non-browser, non-OpenAI surfaces. For a lean DTC brand, probably not yet. For a 20-person Amazon aggregator with an offshore support desk, probably yes.
There’s also the platform risk. A tool that hooks into OS-level text selection on both Mac and Windows is doing something the OS vendors would very much like to own themselves. Apple already ships Writing Tools; Microsoft is shipping Copilot deeper into Windows every quarter. Edyt’s moat is the prompt library and the cross-app consistency, not the selection mechanism. That’s a defensible moat, but it’s a software moat, not a platform one.
What I’d Watch / Test Next
Three concrete things you can do this week without spending a dollar.
Build your prompt library today. Open a doc, list the six text tasks your team repeats most, and write each one as a reusable instruction. Be specific about tone, format, and what must not change (SKU codes, order numbers, currency figures). This is the highest-ROI hour you’ll spend on AI tooling this quarter, and it’s portable across every tool you’ll ever evaluate.
Audit your team’s surface area. For three days, have one ops person log every app they paste text into or out of. If more than half the entries are outside the browser, an OS-level tool like Edyt belongs on your shortlist alongside whatever ChatGPT or Claude plan you already run. If it’s 90% browser, skip it and put the budget into a better browser extension or a well-configured helpdesk macro library.
Test the free tier against your real workflow, not a demo. Install it, run one full shift, and specifically try the screen-reading drag on a scanned supplier invoice or a screenshot from your 3PL. That’s the feature that either earns its place or doesn’t. The 40% off first-year Pro code is nice, but don’t let a discount decide your stack. Let the drag gesture decide.
The broader lesson from this launch isn’t about Edyt specifically. It’s that the next wave of useful AI tooling for cross-border operators won’t be smarter models — it’ll be tools that remove the copy-paste tax from the surfaces where work actually happens. Watch for that pattern in every launch you evaluate this year.
