completed
University of Toronto
M.Sc. Applied Computing — computer systems & LLMs
AGENT SYSTEMSIN CODE.
I research and build agentic systems at the frontier of agent orchestration, harness evolution, and evals.
Building now · Ensemble ↗ — orchestrating agent teams
Jiongan Mu · 2026i'm an AI Engineer at Ploy. previously, i worked at ema.ai, where i researched and built the AutoBuilder/AutoPilot agent harness, focusing on agent orchestration, harness evolution, and evals.
i recently completed my Master of Science in Applied Computing at the University of Toronto, focusing on computer systems and large language models. previously i worked as a backend software developer at Kijiji, and as both a research and teaching assistant at U of T.
outside of work i sing, play piano, play guitar, and lose to my friends at video games. i also love making and eating food. my recipes are where the home-cooked ones live.




A look inside Grok Bot’s context windows, channels, and messaging tools—and how they support an assistant that works more like a teammate.
Read note ↗most systems define subagents through fixed profiles, dynamic prompts, or explicit graphs. a simpler, more powerful primitive is to let the main agent create them with ordinary Python code.
Jun '26a month of new agent papers, sorted by topic, with the high-level idea and the authors' own line of evidence. skim what catches your eye.
Apr '26how coding agents delegate to other agents, and the three paradigms emerging right now: preset profiles, runtime construction, and dependency graphs.
Apr '26running multiple coding agents in parallel without losing your mind in terminal tabs. local-first, open-source, provider-agnostic — and surprisingly, the product is just organization.
completed
M.Sc. Applied Computing — computer systems & LLMs
completed
Hons. B.Sc. Computer Science — GPA 3.94/4.0