I started experimenting with the center stand, prototyping on a 3D printer to determine the geometry. The plastic proves to be surprisingly strong under compression loads even with a low infill percentage. It might be possible to do without any metal in this part at all.
I explored options for lightweight mid-drive motors. The task is complicated by the fact that I will need to learn how to obtain power meter and cadence readings from the motor controller, and none of the models considered even hint at documentation for the communication protocols. The options include Bosch Performance Line SX, Bafang M820, and the increasingly popular DJI Avinox (although it is clearly overpowered and heavier than the other two options). The only one that can be purchased in a relatively normal way is the Bafang; the others are aimed at manufacturers and can only be bought second-hand. Nothing with unusual components is ever simple.
I added a post to Telegram directly from the feed. Versions of the post in different languages are sent to the corresponding localized versions of the channel. And the subscribe button has now also appeared in the feed. Now we need to create backlinks to the website from messages in the channel and support editing.
I quickly rigged up Starlink to my bike to test it. Surprisingly, even in such an inefficient location with tall buildings around, it delivers 200 Mbps for download and 20 Mbps for upload. Stability is a bit lacking - you can't play online shooters, but for everything else, it's a highly workable option.
Today I came across an example of tasks that LLMs struggle with. I was integrating authorization through Yandex and Telegram into grey.one - in the case of Telegram, after a few clarifying questions, I managed to solve it all, but the Yandex button never worked as I wanted it to, after a couple of hours of trying. The reason is simple - there aren't good examples in the documentation, and finding them through searches is difficult. This is where LLMs start to hallucinate and offer harmful and confusing answers.
The Starlink kit has arrived. I will be installing it on my bicycle. It seems there won't be any power issues, as it can be powered directly from the 24V already on board; I just need to rig up a cable a bit. I still need to come up with a mount that provides the dish with maximum sky visibility, which means it should be attached at some high point, probably behind, where the WiFi antennas are now.
One of the undeniable advantages of pet projects is the opportunity to engage in completely irrational but interesting tasks in the moment. Today, I optimized unnecessary requests on the main grey.one. Was it necessary? Of course not – there are only a few visitors here each day. Did it become noticeably faster? No, not really. But the sense of beauty is slightly satisfied.
I added roadmap functionality for projects. For example, the plans for Electro Flevo are here. By the way, it would be good to come up with a more descriptive name for it.
I added new blocks with code highlighting, a Miro diagram, and the ability to hide content under a spoiler. All following the post about Kafka. A wild overengineering for such a simple task, but the reason is, as always, "Because I can."
I wrote a short post about an intriguing technical problem with connecting to Kafka. During the process, a number of shortcomings in the post editor were discovered, but iterative improvements are everything.