Fetching from the wire…
Public story · 2026-08-17 · high
The August 16 tool update skips servers entirely, so platforms that block animated SVG will still take the file.
Why now: This lands now because sharing surfaces still won't render animated SVG natively, forcing workarounds like this one.
Willison shipped browser-side video export for animated SVG on August 16, per simonwillison.net. Plenty of platforms won't render an animated SVG in a feed or a chat window, but they'll play an MP4 without complaint. LLMs make animated SVGs easy to generate, which raises the odds someone wants to share one somewhere it won't render.
The tool detects whether an SVG has SMIL or CSS animation, then guesses the loop duration. It renders the frames and hands them to ffmpeg.wasm, which compiles an MP4 entirely in the browser. No server, no upload step. The commit is verifiable in simonw/tools as 73e0327, per simonwillison.net.
This is a patch on a platform problem, not a fix to the format itself. The SVG format works fine. The tool exists because sharing surfaces refuse to render it. As more people generate throwaway animations with LLMs, expect more conversion shims like this one. A client-side export tool ships in a day. A platform format change does not.
The source doesn't say how long the ffmpeg.wasm compile takes for longer animations, or whether the loop-duration guess ever misfires on irregular timing.
Each link below shares sources, entities, or timing with this story.
Simon Willison doesn't hand out superlatives. So when he writes that Z.ai's GLM-5.2 is "probably the most powerful text-only open weights LLM," that's worth stopping for. His June 17 evaluation walks through a 753B-parameter Mixture-of-Experts model with 40B active params, a 1...
No server. Conversations persist in the browser and export as JSON. He built it to run identical prompts against Qwen 3.8 27B served by LM Studio with its --cors flag, on an M5 MacBook Pro and an NVIDIA DGX Spark, and tested it against OpenRouter too (simonwillison.net). Two d...
Simon Willison found it in the OpenRouter price list. 1.6T MoE, ~49B active, 1M context, up to 384K output, at $0.435/M input (cache miss) and $0.87/M output. The agentic-coding deltas versus the preview are the story: DeepSWE 12.8 → 62.7, CyberGym 52.7 → 83.3, Terminal Bench...
Satya Nadella said companies routing everything through a single proprietary lab may not survive. His argument: you hand that lab your most sensitive business context, and the lab can turn it against you as a competitor. His prescription is an orchestration layer — keep the ha...
Willison launched datasette-apps (0.1a2) on June 18, hosting self-contained HTML+JS apps in a sandboxed iframe that run SQL against your data, read-only by default. He frames it as "Claude Artifacts reimagined for Datasette," artifacts backed by a JSON API to a relational data...
His conclusion is DuckDB matches or beats SQLite's safety for untrusted queries, but only with enable_external_access=false, lock_configuration=true, and a watchdog thread, since DuckDB lacks SQLite's opcode-based query timeouts. He ships a safe_duckdb.py helper and a Datasett...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.