apple intelligence on mac, and what siri ai still cannot recall
apple shipped the assistant people had been waiting three years for, and it answered a different question than the one this category is about.
at wwdc on june 8, 2026, apple introduced siri ai, built on the next generation of apple intelligence, with three stated pillars: personal context understanding, broad world knowledge, and onscreen awareness. on the mac it lives in spotlight, in right click menus and in a dedicated siri app.
it is good. it is also present tense, and that distinction decides whether you need anything else.
what apple actually shipped
siri ai can see what is on your screen and reason about it, right now.
apple's own announcement describes an assistant that can answer questions about the content on a user's screen, search across apps using personal context, and go out to the web. the broader apple intelligence release puts visual intelligence and onscreen awareness across the operating systems.
three capabilities, and it is worth being precise about the scope of each.
| pillar | what it covers | tense |
|---|---|---|
| onscreen awareness | the content on your display right now | present |
| personal context | your messages, mail, photos, calendar and similar apple app data | present, plus what apple's apps stored |
| world knowledge | the web | present |
all three are about answering now. none of them is an archive. independent write ups of apple intelligence on mac land in the same place once they get past the feature list.
what onscreen awareness is, and is not
it is a live read of your display, not a recording of it.
you ask about the thing in front of you and siri answers without you describing it. that removes a real friction, and on a mac it is genuinely useful in spotlight and a right click menu, where it meets you in the flow of work.
when you move on, the screen is gone. there is no index, no ocr text sitting in a database, nothing to query in november about what was on that dashboard in july. this is the same shape as chatgpt's screen sharing, which we took apart in can chatgpt see my screen: the model can look, and looking leaves no trace on your side.
so onscreen awareness answers "what is this." it does not answer "what was that."
the scope of personal context
personal context is deep inside apple's apps and absent outside them.
that is not a criticism, it is the architecture. siri ai can search your messages, your mail, your photos and your calendar because apple owns those stores and can index them safely. it is very good at "find the email where she sent the address."
now count what crosses your mac screen in a working day that lives in none of those apps:
- a browser. documentation, dashboards, a spreadsheet in a web app, a pull request.
- a terminal. the error at 11pm that you fixed and cannot now remember.
- a video call. the slide someone shared for 40 seconds.
- slack, notion, linear, figma. where most decisions actually get recorded.
- a pdf you skimmed and closed. never saved, never filed.
for most knowledge workers that is the overwhelming majority of the day. personal context does not reach any of it, because none of it is apple app data.
why this gap is not an oversight
apple built the assistant the data allowed it to build.
per the anthropic economic index for may 2026, the number one work task in sampled ai conversations is searching electronic sources for information at 4.95%, with reference searching second at 3.74%. retrieval is the dominant use. now look at what remains tiny.
| request topic | share of conversations |
|---|---|
| content creation and copywriting | 22.72% |
| education and learning | 13.23% |
| software development | 11.51% |
| research and intelligence | 10.94% |
| document processing and extraction | 4.32% |
| knowledge retrieval and enterprise search | 3.61% |
| personal ai assistant | 2.86% |
| conversation and meeting intelligence | 0.26% |
the personal assistant category everyone is building toward is 2.86% of conversations. searching your own knowledge is 3.61%. those shares are low because the underlying archive does not exist, on any platform, by default.
every major vendor has now attacked this from the side they control. microsoft built recall into windows, rebuilt after its 2024 reception with biometric gating and on device encryption. openai shipped a background memory system on june 4, 2026. anthropic reworked claude's memory into editable entries on july 10, 2026. google rebranded gemini's context layer as personal intelligence in january 2026. apple shipped personal context and onscreen awareness in june 2026.
five different answers, one shared limit: each covers the data its owner already holds. we compared what each actually retains in ai memory is everywhere in 2026.
apple intelligence privacy, fairly stated
apple's default is stronger than most, and the limit is scope rather than security.
a great deal runs on device, and what does not is routed through private cloud compute, which is a more serious privacy architecture than the industry norm. if your concern is who can read what you hand the assistant, apple has done more work here than its competitors.
the thing to understand is what that protection covers. it protects the data apple's apps hold and the queries you make. it says nothing about the browser tab, the terminal or the call, because those were never in scope. a strong guarantee over a narrow surface is not the same as coverage.
that is the same distinction we insist on for our own claims, which is why the security page scopes them to storage rather than promising that nothing ever leaves the machine.
do you still need a mac ai assistant that remembers?
only if your recurring question is "what was that," rather than "what is this."
be honest about which one you actually ask. if you mostly need help with the thing in front of you, or want to find an email, siri ai is the better tool, it is free, and it is already installed. do not add software for that.
the case for anything else is narrower and specific. you keep losing things that were on screen and are now unfindable. the number from the call. the vendor you ruled out. the stack overflow answer from july, now past the roughly 90 day browser history window. that pattern is what remynd exists for: it captures the focused window on a mac, runs ocr locally so what you looked at becomes searchable text, and keeps the archive on your machine.
since august 2026 it also exposes that archive to claude code and codex over a local connector, so an agent can query your history the way it queries a file, covered in how to give an ai agent context. capture, ocr and storage stay on your mac by default and you can exclude specific apps or sites entirely, though sign in and the cloud agent do reach the network.
the honest summary
siri ai closed the "help me with this" gap on the mac and left the "what was that" gap exactly where it was.
that is a reasonable place for apple to stop. an operating system vendor building a searchable archive of everything on your display is a much heavier decision than building a better assistant, as microsoft discovered in may 2024 and spent 2 years answering for, which we assessed in is microsoft recall safe.
the third party that tried it on the mac is also instructive. rewind built exactly this archive, and in december 2025 meta acquired the company and disabled all capture on december 19, 2025, with 14 days of notice.
so the two tools are not really competing. one is an assistant with excellent manners and a narrow view of your day. the other is a record of the day itself. more on why the record has to sit on your side in what happens to your data when an ai app shuts down and private ai on your mac.
download remynd for mac if the question you keep asking is what was that.