Hello World
The first post on Ben Ebsworth's blog: a living, personal-wiki space for exploring technologies, capturing learnings, and refining ideas over time.
This is my first post on my new fake blog! How exciting!
This site was originally built on the Gatsby base blog template here, with styling, tuning and components taken from the incredibly impressive blog done by Dan Abramov called overreacted.io. The site has since been rebuilt in Next.js (App Router).
The goals of this medium is to explore ideas in an open format to hopefully act as a mechanism for clean note taking but also encouraging more structured exploration of technologies. I would like to play with different technology and write posts here capturing the learnings for my own future reference. I'd like to continually update/clean blog posts overtime, further refining ideas, almost like a personal wiki.
Try it in the lab
All effects →Black Hole Ray Tracer
cosmologyA general-relativistic ray tracer: per-pixel null geodesics around a Schwarzschild black hole, with a physically-shaded accretion disk, Doppler beaming, and live EHT-scale readouts.
general relativityray tracinggeodesicsGravitational Lensing
cosmologyA foreground mass bends background light into twin arcs and a full Einstein ring.
general relativitylensingeinstein ringSpacetime Curvature
cosmologyA mass curving an embedding sheet, with geodesics and bending light rays.
general relativitygeodesicsgravity
More from the blog
Two Number Sum
Three solutions to Two Number Sum, from O(n^2) brute force to an O(n) hash set to an O(1)-space two-pointer scan — and the memory-vs-compute trade-off behind each.
Twelve free models just walked into our benchmark — three of them beat the frontier
We wired OpenRouter's free tier into our 7-task LLM harness, registered 13 models with full metadata, and ran a fair 5-iteration sweep across all of them. Ling 3.0 Tiny, Laguna XS 2.1 and Gemma 4 26B posted averages above 98 on a board that Kimi K3 leads at 90.5 — and the entire run cost us nothing.
The delta rule: linear attention for a million-token context
Full attention pays an n² bill that a 1M-token context can't afford. Linear attention swaps the bill for a memory you write to — and the delta rule is what makes that memory smart. Kimi calls K3's KDA a 'hybrid linear attention mechanism'; this is the family it belongs to, from the kernel trick to gated delta updates.