I think especially for websites / web apps you'd want to start with a "mobile first" design, as is common in web design.
It's almost always easier to go from lower widths to large widths than the other way around for good responsive designs. This, and 1000px being an arbitrary number, doesn't give me confidence.
And, Hey, I promise, I wouldn't willingly... go *out of my way ON A late FRIDAY EVENING (EST)* -- as I low key peruse HN on some sh* mobile browser at a dinner event --
This is really awesome. It took me a second, but then I realized I'm inline tinkering with the product.
Don't let these "pitter, patter" HN minor critiques get in the way of delivering a novel product that is currently on the top of HN.
If it auto-went to dark mode, I would've probably been even more "whoa, i get it..". I know it makes no sense, but, default to dark-mode w/ a fallback header "light" option?
Also, for someone who has dark mode configured on their most visited websites, the light is a little overwhelming. Not at your fault by any means. It's just this may be your average visiting user on your app. I could be the only one, of course.
Please I’d rather have a bad mobile view than a blocker! I know it seems bad but you can always just tell the user to zoom out, a mobile view is coming!
Founder here. I hear you on not having a mobile version available.
I really value the attention to detail and a mobile version of this feels like almost like a different product. I need to start working on a mobile version! :)
I do hope AI will really allow folks to build products with better UX. The problem traditionally is that the UX gets "stuck" - gmail, google maps, they cannot really change because of user's expectations and the big orgs that run them as products. And building new things from scratch was fairly expensive. But now with AI (and modern UI tooling) the equation is at least partially changing.
Love the social feeling of UX. I screen through 100s of emails every day and if you are interested in some pointers for potential uses I would die for on our current corporate set-up...
Visual ques that are extracted from the email context
-Due in 3 days >> has a timer with a 3 day countdown
-Urgent- action immedaitely >> adds an urgency mark to the email
Emails that can get diarized, then brought back up automatically
-Follow up when client is back form their trip >> sorts email into folder, but
brings it back up when the date comes
Assign emails like tasks
-X action need to be done by another person but you need to provide oversight >> tag the team or person and get notified when actioned or not actioned
You've got a shot at building a new Sparrow here. This looks really nice. Unsure that every email message needs a tab on the top, too—I think you could almost just rely on the left sidebar and treating the emails there as tabs. That, and some j/k shortcuts to quickly flip through the emails in your inbox would be great.
This is great feedback. Not sure about the top tabs as well as I spend most of my time inside an IDE, I got used to those. Maybe I'll add a hide option in settings?
On Android, you need to request the desktop version, rotate the phone to landscape, and refresh; assuming you have a tall enough screen (in px).
If it was like Cursor with BYOK, custom instructions, and the ability to have it automatically draft replies when I open an email, and integration with popular suites like Google and Outlook (even if via MCP or CLI) and integration with whatever else I want to integrate it with, you'd have something special.
It could cater to the same type of people who love tinkering with their ide, emacs, vim, etc. I don't know if that's necessarily a market but it would be cool.
Thank you for helping folks with instructions, I really need to start exploring a mobile version! :)
> with whatever else I want to integrate it with, you'd have something special.
What integrations would I need to have to avoid a deal breaker for you?
> On Android, you need to request the desktop version, rotate the phone to landscape, and refresh; assuming you have a tall enough screen (in px).
That's what I thought should do the trick, too, but for me it worked neither in Firefox nor in Chrome. (I have a Pixel 10 Pro, so the resolution should definitely be high enough.)
> [email's] underlying experience didn’t evolve much
In stories of architecture, this is the beaten path that becomes the walkway.
> Is it worth continuing to explore this idea?
It has to be worth it to you.
If you open-source it, you get to articulate what's important and shift from doing to leading. That's a forcing function to state values that inspire people.
For me, UI is a frustrating 1:N problem, where 1 designer(s) make trade-off's for many users. You're bound to get some early accolades, but expanding surface area scales mainly to frustrating everyone in some manner.
I'd like a UI that settles per user or use-case: automatically pruning things I don't use and hoisting things I do, often adapting use-case driven patterns. (The eclipse IDE UI had workspaces suited to different activities, and Mylyn task-based UI which hide or highlighted resources in the workspace for a given task; and that task context could be shared, e.g., attached to a bug, so anyone working on the bug would see (only) the relevant files or methods.)
The key question is what's different now with AI. Email or DB forms are presenting data in ways you can arbitrarily explore.
But when co-working with others or AI, it's more about watching messages and command streams between users, agents, etc, with varying levels of detail. AI is more about queueing up and automating interactions with a given intent. So in this case I'd e.g., enforce a GTD workflow by making queues for simple or hard, with contingencies on approvals or work, spawning actions that reply, and some ways to correlate related streams. To scale you need completion functions, archives, task debt tracking, etc. so you're always starting with a clean slate but someone can always pick up where you left off.
The thing about email is that it has mostly outlived a bazillion contenders, because the data conventions are dead simple and it has relevance built in, where each message (should) start with next steps and provide necessary context (intent and context: sound familiar?). And they're queued in your inbox, giving you instant organization (urgent X important). Combine it with markdown...
Thank you so much for taking the time to provide such incredible feedback, I really appreciate it.
> I'd like a UI that settles per user or use-case: automatically pruning things I don't use and hoisting things I do, often adapting use-case driven patterns.
This is an incredible idea, thanks for sharing!
> AI is more about queueing up and automating interactions with a given intent.
This is what I'm aiming for!
I'm learning this the hard way. I really value the attention to detail and a mobile version of this feels like almost like a different product.
But yea, I need to start working on a mobile version asap!
I think a mobile version should be your sole focus.
Look at it that way, emails are in fact better managed with traditional, deterministic tools. Can move things around, labels, filter, batch delete fast.
On mobile the power tools are difficult to navigate. We are typically on the go, we want quick in and out. That's the main use case to splash an LLM over emails. In fact. It could be just the prompt, to find, read and for quick replies.
Call me a sceptic, I don't see AI replacing something as sensitive as the e-mail interface we know and trust, at least not soon.
Opportunity missed with me, not everyone browses HN on their desktop.
At least have some screenshots of your app so I am motivated to check later.
Here's a video, if folks just want to see how it looks: https://x.com/define_app/status/2030023425150910907?s=20
It's almost always easier to go from lower widths to large widths than the other way around for good responsive designs. This, and 1000px being an arbitrary number, doesn't give me confidence.
And, Hey, I promise, I wouldn't willingly... go *out of my way ON A late FRIDAY EVENING (EST)* -- as I low key peruse HN on some sh* mobile browser at a dinner event --
This is really awesome. It took me a second, but then I realized I'm inline tinkering with the product.
Don't let these "pitter, patter" HN minor critiques get in the way of delivering a novel product that is currently on the top of HN.
If it auto-went to dark mode, I would've probably been even more "whoa, i get it..". I know it makes no sense, but, default to dark-mode w/ a fallback header "light" option?
Also, for someone who has dark mode configured on their most visited websites, the light is a little overwhelming. Not at your fault by any means. It's just this may be your average visiting user on your app. I could be the only one, of course.
Excited to see where this goes.
That's a great feedback, I'll explore a dark mode option.
But yea, I need to start working on a mobile version!
I really value the attention to detail and a mobile version of this feels like almost like a different product. I need to start working on a mobile version! :)
Here's a video, if folks just want to see how it looks: https://x.com/define_app/status/2030023425150910907?s=20
I do hope AI will really allow folks to build products with better UX. The problem traditionally is that the UX gets "stuck" - gmail, google maps, they cannot really change because of user's expectations and the big orgs that run them as products. And building new things from scratch was fairly expensive. But now with AI (and modern UI tooling) the equation is at least partially changing.
Visual ques that are extracted from the email context -Due in 3 days >> has a timer with a 3 day countdown -Urgent- action immedaitely >> adds an urgency mark to the email
Emails that can get diarized, then brought back up automatically -Follow up when client is back form their trip >> sorts email into folder, but brings it back up when the date comes
Assign emails like tasks -X action need to be done by another person but you need to provide oversight >> tag the team or person and get notified when actioned or not actioned
Best of luck!
If it was like Cursor with BYOK, custom instructions, and the ability to have it automatically draft replies when I open an email, and integration with popular suites like Google and Outlook (even if via MCP or CLI) and integration with whatever else I want to integrate it with, you'd have something special.
It could cater to the same type of people who love tinkering with their ide, emacs, vim, etc. I don't know if that's necessarily a market but it would be cool.
> with whatever else I want to integrate it with, you'd have something special. What integrations would I need to have to avoid a deal breaker for you?
That's what I thought should do the trick, too, but for me it worked neither in Firefox nor in Chrome. (I have a Pixel 10 Pro, so the resolution should definitely be high enough.)
I haven't worked closely with web stuff in years, either, though- so I haven't looked into why. But that button just feels like it taunts me, now.
In stories of architecture, this is the beaten path that becomes the walkway.
> Is it worth continuing to explore this idea?
It has to be worth it to you.
If you open-source it, you get to articulate what's important and shift from doing to leading. That's a forcing function to state values that inspire people.
For me, UI is a frustrating 1:N problem, where 1 designer(s) make trade-off's for many users. You're bound to get some early accolades, but expanding surface area scales mainly to frustrating everyone in some manner.
I'd like a UI that settles per user or use-case: automatically pruning things I don't use and hoisting things I do, often adapting use-case driven patterns. (The eclipse IDE UI had workspaces suited to different activities, and Mylyn task-based UI which hide or highlighted resources in the workspace for a given task; and that task context could be shared, e.g., attached to a bug, so anyone working on the bug would see (only) the relevant files or methods.)
The key question is what's different now with AI. Email or DB forms are presenting data in ways you can arbitrarily explore.
But when co-working with others or AI, it's more about watching messages and command streams between users, agents, etc, with varying levels of detail. AI is more about queueing up and automating interactions with a given intent. So in this case I'd e.g., enforce a GTD workflow by making queues for simple or hard, with contingencies on approvals or work, spawning actions that reply, and some ways to correlate related streams. To scale you need completion functions, archives, task debt tracking, etc. so you're always starting with a clean slate but someone can always pick up where you left off.
The thing about email is that it has mostly outlived a bazillion contenders, because the data conventions are dead simple and it has relevance built in, where each message (should) start with next steps and provide necessary context (intent and context: sound familiar?). And they're queued in your inbox, giving you instant organization (urgent X important). Combine it with markdown...
> I'd like a UI that settles per user or use-case: automatically pruning things I don't use and hoisting things I do, often adapting use-case driven patterns. This is an incredible idea, thanks for sharing!
> AI is more about queueing up and automating interactions with a given intent. This is what I'm aiming for!
> (intent and context: sound familiar?) Oh yea :)
Also, I cannot open your website, does not seem to work without JavaScript.
Could reword this: demo only available on...
Better would be to make it mobile responsive before it trends on hackernews.
I'm learning this the hard way. I really value the attention to detail and a mobile version of this feels like almost like a different product. But yea, I need to start working on a mobile version asap!
Here's a video with the demo: https://x.com/define_app/status/2030023425150910907?s=20
Look at it that way, emails are in fact better managed with traditional, deterministic tools. Can move things around, labels, filter, batch delete fast.
On mobile the power tools are difficult to navigate. We are typically on the go, we want quick in and out. That's the main use case to splash an LLM over emails. In fact. It could be just the prompt, to find, read and for quick replies.
Call me a sceptic, I don't see AI replacing something as sensitive as the e-mail interface we know and trust, at least not soon.
This is super interesting.
Great work :)
I will say, I do wish there was a conversation list when looking at folders--having conversations listed on the sidebar can get a bit busy.