Anestan — Designer

My Trello boards and ChatGPT are connected now, and it quietly changed how I run my weeks. I can add a task, move a card, or ask what needs attention without opening Trello at all.

Setup was the official Trello connector, the one-click kind. You authorize a workspace, and from then on the assistant can read boards, lists, cards and checklists, and write back to them. No Zap to maintain, no API key to babysit.

The everyday uses are small and that is the point. A thought arrives mid-conversation and becomes a card in the right list before I lose it. A planning chat ends and I say turn this into cards, and the board matches what we just discussed. No retyping, no momentum lost to tab-switching.

The one I love most is the weekly review. On schedule, it looks over my boards and tells me what moved, what is due, and what has gone quiet. Cards I would have found three weeks late now surface on Monday morning. It is the closest thing I have to someone tidying my desk while I sleep.

It is not perfect. When two boards have similar names it sometimes grabs the wrong one, so I have learned to name the board in the request. A small habit, worth it. The board stays the source of truth. The assistant just does the fetching and filing faster than I ever did.

There is a sloth living at the top of this page now. It follows your cursor while you read, and if you poke it, it blinks back at you. Poke it four times in a row and it gets dizzy. Fair enough.

It comes from page-mascot, a small project by Kamran Ahmed. The whole thing is delightfully low-tech: a character is just two sprite sheets, nine gaze directions and nine reactions, and the component swaps background positions depending on where your pointer is. There is a dead zone in the middle so it stops staring once you get close, plus a little squash animation when you boop it.

Fifty-three characters are already drawn, in six styles from flat colour to pencil sketch to pixel art. Or you can send a photo and get turned into a character yourself. Setup is three steps: install the package, drop the two webp sheets into your images folder, and point the component at them.

My favourite detail is the agent skill that draws new characters. It does not just generate the sheets, it checks its own work: it measures whether the character jumps or drifts between cells, and draws again if it does. I found out why that check exists the hard way. My first set of sloth sheets framed the character slightly differently in each cell, so it shuffled sideways every time it changed direction. Replacing them with the upstream sheets fixed it on the spot.

This site is not React, so I ported the behaviour to a small script instead: pointer tracking, the same timings, the same squash. The sleepy face also moved into the search box, where it naps whenever a query finds nothing. That felt like the right job for it.

Our network teammate was off today, and printing kept failing. The first explanation was an IP conflict. I am a software developer, not a network specialist, but I still wanted to help people get back to work.

I traced the connection from my laptop and worked through the problem with AI as a troubleshooting partner. I compared my laptop’s printer settings with a PC that could print. The configurations matched, but I still could not connect. The missing piece was joining my laptop to the domain. Once I did that, it could reach the printer.

There was another problem with the printer management application: a print job could bring up a login prompt on a different computer, even though the setting was supposed to remember the login until sign-out. After trying to get to the bottom of it, I made a practical call and temporarily stopped the application service. Printing worked again, and people could get on with their day. The underlying issue still needs the attention of whoever owns that system.

That is where AI helped me most. It gave me a way to investigate beyond my usual software work, compare possible causes, and make progress without pretending I had become the network expert. It also helped me recognize when a temporary workaround was the sensible choice and when the deeper fix should be handed to the right person.

I would also like to look into open-source print-management options, paying close attention to active maintenance and security updates. The goal is not to swap systems on impulse, but to understand whether there is a simpler, better-supported fit.

Today I watched a video about steering intelligence, and it put words to a shift I have felt over the past few months.

A couple of months ago, AI in my code editor mostly helped complete the next block of code. Now I can work with it to build much more: explore an idea, shape an interface, connect it to data, and refine the experience. The change is not only that AI can generate more code. I can steer its intelligence with context, constraints, and judgment.

That shift creates more room for software developers. As a solo developer, I can take a good idea much further before needing a team. I can test whether a product should exist, build something real, and learn from using it instead of stopping at “this would take a team.”

Teams will still matter for deep expertise, scale, and long-term stewardship. But the starting point has changed. More developers can build a strong first version on their own. The craft is moving toward choosing the right problem, directing the tools, judging the output, and owning what ships.