๐Helper lines in Datascale.
Visual guides that snap nodes into perfect alignment as you drag them around!
The math behind detecting alignment zones and handling multiple snap points turned into a rabbit hole.
Loveย these small details that make a difference. ๐คฉ
I stopped writing PR descriptions the hard way:
Now I just dump my git diff into a file.
Feed it to @Cursor .
Get a perfect description in seconds!
git diff HEAD~1 >> dif.txt
๐ฅ
Dynamic 3D network built with Three.js and JavaScript. Click/tap sends energy pulses animated via custom GLSL shaders, brightening nodes/connections as they expand. Includes themes & density controls. Source code/CodePen link in the comment below. ๐
This model can output tokens in parallel! ๐ฅ
Unlike traditional models that output one token at a time.
This Diffusion-based LLM performs 10x faster than other models.
by @InceptionAILabs
2 months of my work, distilled into this blog
What Iโve learned aboutย tree- & DAG-based modeling problems.
We went from:
- Adjacency lists (simple but limited to parent-child)
- Closure tables (fast access, but hard for flexible relationships)
- When does Neo4j make sense?