mager.co — notesShort notes — quick links, thoughts, and things learned.https://mager.co/GLM 5.3 vs. my raycaster, or: the benchmark fought backhttps://mager.co/notes/2026-09-04-glm-5-3-raycaster/https://mager.co/notes/2026-09-04-glm-5-3-raycaster/Fri, 04 Sep 2026 00:00:00 GMTAIbenchmarksevalsLLMagentsZhipuVercelI benchmarked myselfhttps://mager.co/notes/2026-09-04-muse-spark-1-3/https://mager.co/notes/2026-09-04-muse-spark-1-3/Fri, 04 Sep 2026 00:00:00 GMTAIbenchmarksevalsLLMMetaopencodeWayfinder: planning big work as a map of decisionshttps://mager.co/notes/2026-08-08-wayfinder/https://mager.co/notes/2026-08-08-wayfinder/Sat, 08 Aug 2026 00:00:00 GMTagentswayfinderplanningmager-bench: the benchmark was measuring my token budgethttps://mager.co/notes/2026-07-26-mager-bench-token-budgets/https://mager.co/notes/2026-07-26-mager-bench-token-budgets/Sun, 26 Jul 2026 00:00:00 GMTAIbenchmarksevalsLLMGroqopen weightsOpus 5 is the default now, and subagents can nest three deephttps://mager.co/notes/2026-07-26-opus-5-claude-code/https://mager.co/notes/2026-07-26-opus-5-claude-code/Sun, 26 Jul 2026 00:00:00 GMTAIclaudeclaude-codeagentsOpusmager-bench: free models, thinking-token bugs, and traceshttps://mager.co/notes/2026-07-18-mager-bench-free-models/https://mager.co/notes/2026-07-18-mager-bench-free-models/Sat, 18 Jul 2026 00:00:00 GMTAIbenchmarksevalsobservabilityLlamaGeminimager-bench: the doom challengehttps://mager.co/notes/2026-07-17-doom-challenge/https://mager.co/notes/2026-07-17-doom-challenge/Fri, 17 Jul 2026 00:00:00 GMTAIbenchmarksevalsLLMWebGLJavaScriptDoommager-bench: free firsthttps://mager.co/notes/2026-07-17-mager-bench-free-tier/https://mager.co/notes/2026-07-17-mager-bench-free-tier/Fri, 17 Jul 2026 00:00:00 GMTAIbenchmarksevalsLLMGroqGeminicrowdfundingmager-benchhttps://mager.co/notes/2026-07-15-mager-bench/https://mager.co/notes/2026-07-15-mager-bench/Wed, 15 Jul 2026 00:00:00 GMTAIPythonGoElixirbenchmarksevalsLLMskill-evalshttps://mager.co/notes/2026-06-26-skill-evals/https://mager.co/notes/2026-06-26-skill-evals/Fri, 26 Jun 2026 00:00:00 GMTAISkillsEvalsClaude CodeLLM-as-judgeLoooom v1.0https://mager.co/notes/2026-06-10-loooom-v1/https://mager.co/notes/2026-06-10-loooom-v1/Wed, 10 Jun 2026 00:00:00 GMTAISkillsLoooomEvalsPromptfooKeeping an always-on agent alive across rebootshttps://mager.co/notes/2026-06-03-always-on-agent-across-reboots/https://mager.co/notes/2026-06-03-always-on-agent-across-reboots/Wed, 03 Jun 2026 00:00:00 GMTagentsmacOSharness