Lettertrail

Inspiration

I am interesting in a lot of newsletters and they clog up my inbox.

When I search for important mails, it feels like searching for the needle in a haystack. So many newsletter mails are mixed in the search results.

It is also hard for me to keep track of my reading progress and especially resuming reading since I dont have that much time to read a newsletter from start to finish.

I wanted to have one spot where I can do my newsletter reading and clean up / free up my personal inbox.

What it does

The main screen is an inbox view where each newsletter is in a state "reading", "unread", "read". Per default der "reading" filter is active, so I can pickup my reading quick without having to search first.

In the reading screen newsletters are normalized and shown in a calm view with less distractions. Form there I can do following actions:

Mark the whole newsletter as read
Read through the newsletter itself and tick checkboxes for paragraphs I read or I am not interesting, so when I later resume reading, I know where to continue to read. Lettertrail also keeps track of your scroll position, so next time there also less scrolling
You can select text and just highlight it, add notes to highligths or add those to your "To Read" list. The usecase is: I skimm through a newsletter while not having that much time. I mark the interesting bits with "To Read", so I can create my own reading journey

There are dedicated views to show your "To Read" items and Notes, so you can keep an overview and directly jump to the newsletter scrolled to the saved paragraph.

Lettertrails provides an email address where I can forward existing newsletters, so they get ingested into the system.

In the future I want to directly sign up with this special email address, so that my inbox will be skipped entirely.

How we built it

I did a long 3h planing run with Matt Pocock's wayfinder skill. It really helped me to get into the details, focus, prototype and come up with a plan and implementation tickets, I can hand to an agent.

Then I used Kun Chen's Firstmate skill and supporting tools to autonomously build the project. It worked really well. Ran for approx. 24h and landed on prod.

I was very pleased by the results.

Then I did approx. 1 day of polish, adding further features and get it into shape. The autonomous outcame was great, but lacked the human touch /eye. In this phase I did much tighter building loops and more steering of the agent.

Challenges we ran into

Going out of tokens. I am on the 100$ plan and my weekly quota was full in the beginning. I used that and the 100$ credits given by the Build Week and this was not enough.

I had to use 3 of my resets to get to the finish line.

Accomplishments that we're proud of

Having the agent run automously approx 24 hours.

Having a quite good polished MVP which I can really use. It provides me real value to me and I am happy that I joined the build week

Letting go of the infrastructure side and let the agent do it. I created a safe space with hard limits, where the agent can work and if it messes up, risk is low.

What we learned

Letting the agent handle infrastructure works well nowadays. But there are still safety concerns I have to solve if I want to use it in a real project.