Solutions Architect building in public ▪ Designing systems that have to survive production ▪ Architecture, cloud, AI agents ▪ Proud dad of two daughters
Day job: I design systems where being wrong is expensive (fintech, events ticketing, agentic AI).
Evenings: I build one alone, with AI writing most of the code.
The second one is the fastest way I've found to learn which of my architecture habits were load-bearing.
⚡️❗️УНІКАЛЬНІ АРХІВИ : записи програм з українського телебачення Торонто кінця 80-х років ХХ століття.
▪️ Українська 🇺🇦 діаспора в Канаді 🇨🇦 у ті роки вважала, що українська культура і мова на їх Батьківщині безповоротно втрачені і саме вони - емігранти - повинні зберегти це все і повернути в Україну, коли вона стане вільною.
Відео на YouTube — https://t.co/UK4AfqAyJN
@dev_maims What makes me worry is that such person can try 50 bad ideas without asking for budget, headcount or permission and then get competitive advantage in a brute force mode ))
@vimtor I am a bit terrified by this trend because in this case AI owns the whole interface and your app risks becoming an interchangeable backend for it.
Meanwhile in China, researchers wrote 75 pages on how AI could improve itself. There are five stages. In the last one, AI starts changing the way it makes itself smarter.
Looks like humanity got assigned the documentation for the singularity.
https://t.co/vIXsZrMJ09
At work, I create Agent Skills for others, but they stay under NDA.
So, I published two from my personal setup:
• "trim-instructions" shortens model instructions without changing behavior
• "tidy-docs" multi-pass clean-up of Markdown for people
https://t.co/IU7KjSqKHm
@maxedapps I do this at work, but those skills are under NDA. So I published two from my personal setup:
• trim-instructions: shortens model instructions without changing behavior
• tidy-docs: multi-pass clean-up of human-facing Markdown
https://t.co/IU7KjSqKHm
@maxedapps I do this at work, but those skills are under NDA. So I published two from my personal setup:
• trim-instructions: shortens model instructions without changing behavior
• tidy-docs: multi-pass clean-up of human-facing Markdown
https://t.co/IU7KjSqKHm
After 3 #Puliares interviews, I had exactly 0 Interest signals.
I could blame the sample, then open the backlog and build another feature.
But another feature won’t tell me why nobody was interested. So next step is simple: interview a different segment or test a new channel.