Goldfish - Press Option. It knows your work and replies like you
by•
Most AI tools make you explain the context before they can help. Goldfish already has it. It privately remembers what you’ve been working on across your Mac, then helps you write better from any app. Press Option in a text field to draft replies, summarize threads, rewrite sentences, or recall important details from your recent work without copying, pasting, or re-explaining the whole backstory.
Replies
AI has memory like a goldfish is such a good way to put it. The real tax of every AI tool right now isn't the thinking, it's the re-explaining. How does Goldfish decide what's actually relevant context vs. just noise from everything open on the Mac? Feels like that filtering is the hard part.
@andrew_paul11 We think about it less as “throw everything into memory” and more as ranking a few layers separately.
The current surface wins first: the field you’re typing in, the thread around it, the app/window you’re in. Then recent activity is used for handoffs, like if you just switched from a doc or Slack thread into an email reply. Longer-term memory is mostly for patterns around people, projects, tone, and recurring work.
If Goldfish is unsure, it should stay general or ask, not confidently pull in some random tab from earlier. That filtering is definitely the hard part. More memory is easy, choosing the right memory is the product.
How does Goldfish handle forgetting old context that's no longer relevant like a finished project or an old draft? Is that automatic or does the user manage it?
@carlos_leonardo1 it’s mostly automatic. Goldfish prioritizes recent context for “what am I doing right now?” and uses longer-term memory more for patterns around projects, people, and tone.
Old stuff should naturally decay in relevance when you move on, but you can also control it manually: pause capture, exclude specific apps/domains, or delete stored history. The goal is that a finished project doesn’t keep leaking into unrelated drafts just because it once mattered.
Congratulations!!!
@madalina_barbu
Thank you Madalina!!
Hey Joel! It's so cool man. Guessing how it takes my own voice, vibe, jargon and so on. Anyway it's super interesting and I'm sure you gonna rock it
@german_merlo1 Thanks Germán! Really glad it picked up on the voice/vibe part, that’s exactly the feeling we’re trying to nail. Appreciate the support! 🐠
@joel_edholm hello, great app tbh. but i have 1 ques, is there a way to disable the orange line in the top of the screen and if i want to talk to goldfish i press on it directly? thanks in advance
@joel_edholm @itzw0lf You can turn on auto-hide in settings. Then just press Goldfish when you want it to appear on your desktop.
@itzw0lf hope you found it! What have been your main use cases?
Congrats on a great idea and pretty good execution. But a few things still feel rough:
1. It is 2026, and settings plus non-standard UX are a big distraction, especially when working across multiple screens.
2. Goldfish often seems to generate before fully reading the current context, so it loses the thread. It would be more useful if the icon showed actual Goldfish activity/status instead of just a fish flying around the screen.
@igor_gnot1 Hi Igor! :)
For 1, what parts felt rough specifically, the settings flow, multi-screen behavior, or something else?
For 2 you can usually steer it with a clearer prompt today, but would you prefer instead of "goldfishing.." for it to animate the actual goldfish? :-)
@igor_gnot1 thanks Igor, fair push. appreciate the direct feedback 🐠
I had to try Goldfish right here on Product Hunt while writing this comment. This is exactly where the product makes sense: it picks up the context without making you explain everything first.
Sounds really great - with lots of information every day it became really easy to loose important things. But I really wonder, how much space it takes on the computer, and would be my data safe?
Onboarding smooth af. I’ve only used it for a week or so and I’ve already found myself wishing I had it on my phone too. Very well done.
the "knows the internet but not the work" line is the actual gap. it's why every AI tool feels useful for 3 minutes and useless for the 4th. they're operating on context that's everyone's and nobody's at the same time.
real question: how do you handle the boundary between "knows my work" and "leaks my work"? the moment AI starts ingesting my drafts, my comp negotiations, my "i don't actually agree with this strategy" doc, the security threat model gets way more interesting.
if there's a clean answer to that i'm sold.