blink vs just asking chatgpt

the difference is where the context lives

you’ve got a chatbot that can write anything. why install a Mac app to write replies?

because the chatbot doesn’t know what you’re looking at.

with ChatGPT, the work starts before the answer does. you switch to the chat window, paste in the thread, explain what you’re trying to do, read the response, and paste it back. every turn, you re-supply the situation. the model has no idea what’s on your screen unless you tell it.

that tax is small once and adds up across a day of replies. blink removes it: the context is already on screen, so you stop re-explaining yourself. you summon it, it reads the whole thread on demand (not just the last few visible messages), drafts a reply in your voice, and drops it into the box you were already typing in. you decide what to send.

but ChatGPT can see my screen now

sort of. ChatGPT’s macOS app has a “Work with Apps” feature, and it does read what’s in your apps. but it doesn’t see your screen the way the marketing implies. OpenAI’s own product lead confirmed to TechCrunch that it reads text via the macOS Accessibility API and can’t interpret photos, video, or visual layout. that’s the same class of API blink uses, so the screen-reading itself isn’t a differentiator. the differences are around it.

it only works in a curated list of apps. per OpenAI’s help center, Work with Apps covers coding and note tools: VS Code/Cursor/Windsurf, Xcode, JetBrains IDEs, terminals/iTerm2, Apple Notes, Notion, TextEdit, Quip. it does not cover iMessage, Slack, Gmail, Discord, or arbitrary web text fields. blink works anywhere you type.

you have to point it at the right window. you attach ChatGPT to a specific app and it pulls surrounding context, reportedly the foreground window / last ~200 lines. it isn’t aware of whatever you happen to be looking at; you tell it where to look. blink reads whatever’s in front of you.

it doesn’t write the answer back. even with Work with Apps, ChatGPT doesn’t insert the response into your app. for general apps you still copy-paste; for IDEs it generates a diff you review and apply. blink’s job is the insert. the reply lands in the box.

the voice gap

the default ChatGPT voice is generic unless you prompt it hard. that’s fine for a draft you’ll rewrite. it’s wrong for a reply that’s supposed to sound like you sent it. blink drafts in your voice because the whole point is the message goes out as yours.

when to just use ChatGPT

the chatbot is genuinely better at open-ended exploration and back-and-forth. if you’re brainstorming, learning something new, or working through a problem with no fixed answer, that’s a chat conversation, not a one-shot reply. and ChatGPT is cross-platform; blink is macOS-only.

but if the answer is a function of what’s already on your screen, a reply, an unblock, a quick response that the thread basically dictates, blink collapses it into a keystroke and skips the re-paste loop. they’re not competing for the same job.