Just published a long technical writeup on Skiagrafia — a local-only, batch-capable macOS pipeline that converts image folders into layered SVG + alpha-matte TIFF assets. Everything runs on-device on Apple Silicon. No cloud, no API calls during inference.
tsevis.com/every-pixel-is-a-tesserae
github.com/tsevis/skiagrafia

Meet Hipparchus, a free and open-source desktop app for creating clean vector maps from OpenStreetMap data.
The idea is simple: maps should be useful creative material. Hipparchus lets you fetch a location, choose the map layers you want, preview the result, and export a layered SVG where roads, buildings, parks, water, coastlines, railways and labels remain separate and editable.
https://github.com/tsevis/Hipparchus

I just released Skiagrafia, a desktop app built for batch image processing.
The goal is simple: give people a way to generate masks or vector silhouettes for entire folders of images, not one file at a time. It uses local, open-source VL, CV, and language models to drive segmentation, refinement, and structured output generation.
Repo: https://github.com/tsevis/skiagrafia

For more than two years now, I’ve been experimenting with building software tools with AI assistants.
Many people call this “vibe coding”. I’m not very fond of the term. Coding itself is not really the objective. It’s simply one of the languages we use to communicate with machines.
What matters is the system we build and the usefulness it brings.
Some thoughts about this in a short essay:
https://tsevis.com/vibecoding-or-ai-assistant-system-building

MacWhisperer: Private, local Whisper for Mac. Turn media into editable text. https://github.com/tsevis/MacWhisperer

I've subscribed to the Qwen Coding Plan by @alibaba.cloud. The initial setup was a bit more complicated than I expected, but the offer of 8 models for $10/month was a no-brainer. I've been a long-time Qwen Code user, though I've occasionally used MiniMax, Kimi, and GLM models. Let's see how this will help my workflow.

Qwen3 Coder dropped!
• 0.5B to 72B parameters
• Open source (Apache 2.0 license)
• Trained for code generation, completion, and understanding
• Supports multiple programming languages
• Chat-style coding assistant
https://github.com/QwenLM/Qwen3-Coder

I created Elon Musk's mosaic portrait for TIME Magazine when I believed in the vision, electric cars, spaceflight, solving the "unsolvable."
But something shifted. The "ego/brains >> 1 problem" he once criticized became his own.
As artists, we don't just place shapes—we place meaning. Did my portraits help build a myth?
A question that hurts but needs facing.
https://tsevis.com/elon-musk-ego-and-the-mosaic-of-misjudgment
#Art #ArtisticEthics #CreativeResponsibility

Excited about the "Diffusion Effect" in GenAI! It's like painting with broad strokes first, then adding details to perfect it.
This approach makes text generation an iterative process, refining outputs step by step. It's like a conversation where each response gets better, enhancing speed and accuracy.
https://chat.inceptionlabs.ai
#AI #DiffusionEffect

I had the honor and the opportunity to create the main visual for the Good Doctor's farewell season.
Here is the whole project with some behind the scenes informations and bonus artworks.
A big thank you to all ABC Network people.
https://www.behance.net/gallery/211818857/The-Good-Doctor-Farewell-Season-Artwork