So here's the number that made me move all my agents to Go.
A goroutine uses about 2kb of memory. An OS thread uses about 8mb. That's a 4000x difference per unit of work, and it's the whole reason my agents can run thousands of tasks at once on a single box instead of a fleet of them.
Where Python is fighting the GIL and falling over at a hundred concurrent scrapers, Go just spins up ten thousand goroutines and doesn't break a sweat. 10k tasks is like 20mb of memory instead of 80gb. The math isn't close.
Concurrency you can actually afford is concurrency you can ship.
Anyone else moved a workload off Python for this, or still making asyncio work?
This deserves more than just liking, this deserves to be featured! 🏆 That 4000x memory efficiency comparison is a crazy statistic. Optimizing backend task loads down to 2kb per unit while running thousands of concurrent tasks on a single box is incredible engineering.
Appreciate that. The practical win is hosting: a whole agent fleet fits on one small box that costs less than most Vercel bills. Do you run Go in production yourself?
Konto — Personal finance planner, web dashboard and mobile app
Konto is a personal finance planner concept I designed for both desktop and mobile, sharing one data model and one visual language.
Web dashboard: four KPI cards (balance, income, expenses, saving rate) with mini-charts, a weekly cash-flow chart with income and expenses side by side, spending by category as a segmented ring with a legend, and a recent-transactions table. The AI assistant lives in a docked panel bottom-right.
Mobile: the dashboard leads with total balance and a 30-day trend, then cash flow, saving rate and categories in one scroll. Analytics gets its own tabbed view. A floating action button opens the assistant from anywhere.
The visual system is dark by default with two signal colours: lime for income and growth, orange for spending. Glows are used only on the active data point, so highlights mean something. Numbers use the same layout and spacing on both platforms, so switching devices doesn't require re-learning the app.
I’ve been razzmatazzing and flibbertigibbeting with Claude Code and somehow ended up with a new yels.dev.
It finally feels like me: simple on the surface, slightly complicated underneath, and very much alive.
You can see who I am, what I do, the work behind Herodot, RaptorLabs, CyberLink Security and Solmint, plus a selection of projects I’ve built across AI, cybersecurity, Web3, product systems, education, archives, and experimental digital spaces.
The simple-on-the-surface, layered-underneath idea comes through so clearly in the presentation. I especially like how the project archive turns the site into something to explore rather than just a résumé of links.
HYDRUM VEIL is a skincare brand built around a premium SPF 50+ sunscreen, developed as a complete brand identity, packaging design, and seasonal advertising campaign.
The project explores a question most seasonal beauty work gets wrong: how do you make a limited-edition campaign feel festive without making it feel cheap?
Instead of stacking decorative props, the campaign holds a strict visual system - vivid pumpkin orange, deep twilight violet, warm cream, and restrained black and lets a single product carry every frame.