Simon Willison Releases Bluesky Reply Bot Checker Tool

What happened

Tech practitioner Simon Willison delivered the closing keynote at the North America edition of the WeAreDevelopers World Congress held in San Jose, releasing the Kākāpō Party tool to map key LLM developments since 2026 on a timeline. He coined terms related to AI burnout and used his self-built benchmark of generating a pelican riding a bicycle in SVG to test model capabilities. Separately, leveraging the open Bluesky API, he rapidly developed a Bluesky auto-reply bot detection tool through vibe coding with Opus 5.5, scraping signals across multiple account types to identify spam bots.

Key facts

Speaker

Simon Willison

Tools Released

Kākāpō Party, Bluesky reply bot detector

New Industry Concepts Proposed

AI burnout terms Deep Blue, AI mania

Conference Topic Scale

277 sessions in total, 40 focused on agent security/sandboxing

Self-built Evaluation Benchmark

SVG capability test of generating a pelican riding a bicycle

Key Model Milestones

Claude Opus 4.5 and GPT-5.1 released in November 2025; Opus 5.5 used to develop the detection tool

Background

When Claude Opus 4.5 and GPT-5.1 were released in November 2025, the accompanying coding agents frequently made errors and were difficult to use in daily work; the Bluesky platform, similar to Twitter, hosts large volumes of meaningless auto-reply bots. With LLM iterations accelerating, practical deployment of coding agents and agent security are core industry exploration directions.

Why it matters

For the industry, the mature deployment of coding agents will raise software production efficiency and drive investment in agent security, while lightweight small tools can also supplement platform governance blind spots. Developers can leverage open APIs plus large models to quickly ship practical projects and expand application boundaries. General users can both enjoy more efficient AI-assisted development products and gain a cleaner community experience.