Insights, research updates, and early ideas from the COAI team. Notes can motivate a threat model or an experiment; their original status remains visible.
Selected reading
The Shoggoth Prison framework proposes restricted agency and monitor-mediated interactions. Which assumptions must hold for those barriers to work?
Question to test: Can a monitor's decision be enforced at every action boundary?
Read the original noteThe Flight Recorder note explores provenance and replay. It motivates inspectable traces without establishing that logs reveal a model's internal reasoning.
Question to test: Can we reconstruct what each actor knew, could do, and actually did?
Read the original noteThree conditions decide whether offloading a cognitive operation augments or erodes. Each failure names a distinct mode, and one of them produces overseers who hold control in form only.
Question to test: Which links in a control chain assume a competence that routine AI use removes?
Read the original noteThe debate about AI in universities is about which tools to adopt and how to stop students cheating with them. Both leave a harder question unasked: what happens to human agency once AI is good eno...
Read more → AI EconomicsAI hype merchants claim transaction costs are going to zero and companies will dissolve into swarms of autonomous agents. The economic reality is stranger. AI is splitting the economy into context-...
Read more → AI Co-EvolveAn analysis of 44 million workers across 266 occupations, inspired by Karpathy’s US study — adapted for Germany’s unique labor market. A opinionated analysis by us
Read more → AI SafetyAs AI models scale toward trillions of parameters, they become increasingly capable yet harder to interpret, raising the risk of subtle misalignment that current evaluation infrastructure cannot re...
Read more → AI SafetyThe Automaton Has Arrived — And It Doesn’t Need You
Read more → AI SafetyDaniel Kokotajlo’s AI 2027 scenario mapped a month-by-month trajectory from stumbling agents to superintelligence. Nine months into its timeline, we assess which milestones have been hit and find t...
Read more → AI SafetyMoltbot, now called OpenClaw, went from obscure open-source project to 147,000 GitHub stars in under two weeks. Millions of users have handed their passwords, emails, and calendars to an AI agent t...
Read more → Mechanistic InterpretabilityHow unified memory architecture and thoughtful API design are making interpretability research accessible to researchers everywhere
Read more → EvaluationCurrent evaluation frameworks for Large Language Models (LLMs) predominantly assess logical reasoning capabilities while neglecting the crucial dimension of critical thinking. This gap presents sig...
Read more → Early Research IdeasAs AI agents become autonomous decision-makers, we need “flight recorders” that capture their complete internal reasoning—inputs, neural activations, and decisions—in a deterministic, reproducible ...
Read more → AlignmentBuilding on Initial Discoveries
Read more →