Week 1 - change of plans.

I'm building Paperjoy, a service where a robot writes real-pen handwritten postcards and mails them anywhere in Poland. The plan was full automation, but this week I gave up on that, for now.

Bad news first.

I wasn't able to make the paper feeder reliable in time.

I tried multiple designs and their iterations, using both friction and vacuum. In the end I was always running into paper separation issues.

There is no point in automating this process if I'll have to babysit the thing all the time. On the other hand, I have to launch. I have reasonable savings, but with no income source this might not be true soon.

New constraints. Speed > perfection.

I'll launch with a couple of small plotters that I will load manually for now. In the meantime, I'll keep tinkering with the feeder and hopefully will get it reliable soon enough; otherwise I'll order a commercial plotter (ouch) if the demand is too high for me to keep loading the paper.

I have decided to rebuild my plotter again. I went with a very simple CoreXY design inspired by DrawBot 1.1 by Henry Arnold using all the spare parts from the main plotter I've been working on for months. It's funny because this was the first plotter I ever built just to test this idea, so I came full circle :).

To draw a picture here.. I can build 4 plotters like this from the materials I used for the old design.

It's easier to assemble and maintain than Henry's design (I want to be able to spin up a new unit in under an hour). It's still very cheap, but removes some cost-cutting and uses more off-the-shelf components. All this for the cost of some sagging.. which I'm fine with given it's an A6/C6 format.

Don't prototype for the looks.

I was spending way too much time designing things no one cares about. Like the electronics box. Wtf.

I now have a simple PLA wall with mounting holes for boards and wire management.

Neat. Simple. Easy to build, easy to fix.

Meta Ads

I ran a campaign with a CTA to order a free sample and it went moderately well.

Got some leads, with an average cost per lead of ~50PLN (13USD), which on one hand is on the higher end, but on the other it is a B2B ad for a product that isn't available yet - so I'm somewhat optimistic. Lead quality could be better. I'd say 50% are really good, but when it comes to the other half I probably shouldn't even ship samples to them.

We'll see how many of those leads will convert, hoping for 10-20%.

Optimization and SEO

Since there is no real competition in Poland, I want to go hard into SEO from the beginning. I spent a lot of time this week optimizing the page to 95-100 Lighthouse scores, as well as bringing the page to a 99 is-agentic score. I also:

• revamped this personal page to get the first backlink :D
• published the first "keyworded" page and have another one in the making.

First post on socials

I finally stopped working in stealth mode and posted about Paperjoy across all of my socials.

X post went viral with over 30,400 views which is a really good outcome for an account with less than 2000 followers that hasn't been active for a long time. Comments were rather positive which is nice, but 0 leads:) Some haters rushing in to enlighten me that I didn't invent a pen plotter. Duh.

LinkedIn got fewer views but still got some clicks, and of all places I got the most leads from Facebook (?).

Documentation, DX

Published the initial version of Paperjoy docs! It still has some rough edges that I need to work on, but I decided that they are good enough to finally publish them. The initial draft was written way back by GPT-5.6, and now they've been hugely rewritten with Opus 5.5 (which even took most of the screenshots from the app - what a time to be alive).

I asked some of my engineering friends to skim through them and the reaction was positive!

Next week.

The plan for next week, in no particular order:

• Drafts. Sending postcards through AI agents is cool until one of them mails 200 cards with a typo. I want agents to create drafts that a human approves before anything hits the plotter. Safer for everyone.
• Ship samples to leads. Starting with the good half :)
• Build a second machine. One plotter is a single point of failure, and I don't want to launch with one. It's also a good test of my "new unit in under an hour" goal.
• Ship another SEO page.
• Rebuild onboarding. It needs to be dead simple before strangers start using it.
• Start preparing launch assets. Ads, directory listings, all the boring stuff that needs to be ready before launch day.

Lots to do. See you next week.