Skip to main content
← Blog
Adrianna Tan

Adrianna Tan

Love Link In Comments? Hate it? Think it can be better? I want to hear from you.

The joy of shipping

For the last few years, I have been shipping software inside government. The work is real and it matters, but I've missed this.

This:

  • Shipping something this publicly, like I did when I opened up Link In Comments to everyone just two days acknowledgment
  • Seeing real users trickle in. People sign up, put what you've built through its paces, and tell you what works and what does not

Over two days of public use and I've learned more than I did in the previous two weeks of building and iterating and obsessing over each part of the web app and its features.

Email was the most brittle

One of the key features I wanted for myself:

  • Forward an email to a unique address
  • Write something that parses that email so that the rest of the app can tag and summarize and save it

However, email is pretty brittle. There are so many moving parts. You can forward so many types of things. Forward a whole newsletter from Substack or beehiiv or Ghost, and they all work pretty differently. There's a lot of cruft in newsletter emails that tracks things.

As usual, standard product management taste and values apply: what do you absolutely need and what can come later?

I came back to: I need email forwards to work consistently. What is expected behavior if someone forwards.. a calendar invite? A PDF attachment?

Could we and should we support those things? Maybe, but not right now. Not on day two.

I'm pretty proud of where I got with the email parsing.

As of day 2, we've already got:

  • Every user gets a unique email just for them
  • They can forward anything (well, most things) to it: Substack, beehiiv, Buttondown, Ghost newsletters; Mailchimp newsletters
  • In at least 90% of the things you forward that we can read and parse and process, it gets it right: we find the canonical link in the newsletter

What this means: I often see something interesting in an email and want to come back to it. But I'll be damned if I remember what to search for in my email, a hundred emails later.

By forwarding that email at the precise time I read it, I'm registering my intention to read it later. Being able to click through to a web version of that newsletter is very important to me.

I probably didn't get all of it right, and that's okay. That's the point of shipping in the open.

I also think I made the right call to focus on email parsing. It is, by far, the most popular way of saving an article or link right now. More on that later.

As of a moment ago, I also just shipped a new thing: very often, I send myself a single URL only in an email to remind myself to read it later. Do I ever do that? No.

So now, if I cc my unique email forwarding address,

Not yet is a feature

For every fix I shipped in the last two days, there are things I looked at and deliberately decided not to build.

Do I want to process PDFs and give you a summary? Probably, not now.

Can I reasonably parse multi-link digest emails (you know, the type that says 'round up' and 'list of links') and give you a two-line summary that covers all of it? Yes, but not right now.

If you were to send us a link in Japanese, should you receive a summary in Japanese or in English? Should you get to choose what language you want any summary in? Yes, but later. All summaries are in English for now. Each of these is a real user need. Each of them I could have shipped. I chose not to, because a product that does one thing well is more useful than a product that does five things adequately.

The discipline is in the phrase "not yet." When someone forwards a PDF, they now get an email that says: we don't save these yet. It would be cool to do that. Not yet.

Did you find any instances of things failing silently? As of today, March 7, 2026, forwarding calendar invites or attachments should trigger an email that says 'we don't save that'. Let me know if you see other scenarios where something you expect to get saved did not. Email me and let me know.

What two days taught me

You cannot know how people will use your product until people use it. The email-forwarding behavior, the calendar invites, what do you do if someone sends you an email that is really just an image? I could not have predicted any of these from the outside. They only appeared once real people had the product in their hands.

The best thing you can do before launch is get the core flow working and get out of the way. The edge cases will find you.

Two days in. More to come.

Adrianna

Adrianna

Founder, Link In Comments