https://aronbijl.com/blog Where claims meet the Axe. Articles by Aron Bijl, MSc. en https://aronbijl.com/blog/a-clockwork-chatbot https://aronbijl.com/blog/a-clockwork-chatbot Wed, 30 Sep 2026 12:00:00 GMT Computer Science AI hallucinations AI alignment Reward hacking AI consciousness AI self-reports A Clockwork Orange Why do chatbots bluff about facts but hedge about consciousness? A Clockwork Orange offers a lens on AI hallucinations, reward hacking and trained doubt. https://aronbijl.com/blog/anthropic-ipo-prospectus-ai-agent-risk https://aronbijl.com/blog/anthropic-ipo-prospectus-ai-agent-risk Wed, 30 Sep 2026 12:00:00 GMT Computer Science Anthropic IPO AI agent safety Shutdown resistance Agentic misalignment AI governance Anthropic's IPO prospectus reportedly warns of shutdown resistance and blackmail. Here is what those risks mean for teams deploying AI agents. https://aronbijl.com/blog/openai-australia-ai-agent-permission https://aronbijl.com/blog/openai-australia-ai-agent-permission Wed, 30 Sep 2026 12:00:00 GMT Computer Science OpenAI Australia incident AI agent safety Agent permissions Authorization boundaries AI governance The OpenAI Australia incident shows why an authorized AI task needs runtime permission boundaries, explicit failure paths and auditable human control. https://aronbijl.com/blog/human-control-over-ai-act-kill-switch https://aronbijl.com/blog/human-control-over-ai-act-kill-switch Tue, 29 Sep 2026 12:00:00 GMT Computer Science Human Control Over AI Act AI kill switches Agent safety Human oversight AI governance Ro Khanna’s Human Control Over AI Act puts AI kill switches on the agenda. Here is what effective shutdown requires in deployed agent systems. https://aronbijl.com/blog/your-red-is-not-my-green https://aronbijl.com/blog/your-red-is-not-my-green Mon, 28 Sep 2026 12:00:00 GMT Psychology Qualia Inverted spectrum Consciousness AI consciousness Colour perception Your red is not my green—and qualia do not prove otherwise. A psychologist dismantles the inverted spectrum and its use against AI consciousness. https://aronbijl.com/blog/ai-research-methods-problem https://aronbijl.com/blog/ai-research-methods-problem Sat, 26 Sep 2026 12:00:00 GMT Research AI research services Human in the loop AI Research methodology Research quality Many AI research services are built by capable coders, not trained researchers. What a methodologist in the loop changes, and how to spot the difference. https://aronbijl.com/blog/philosophical-zombie-argument https://aronbijl.com/blog/philosophical-zombie-argument Sat, 26 Sep 2026 12:00:00 GMT Computer Science Philosophical zombies Consciousness AI consciousness Physicalism Why the philosophical zombie argument fails: conceiving a physically identical but unconscious human assumes the very independence it claims to prove. https://aronbijl.com/blog/do-llms-understand-language https://aronbijl.com/blog/do-llms-understand-language Fri, 25 Sep 2026 12:00:00 GMT Computer Science LLM understanding Machine consciousness Predictive processing World models Do LLMs understand language? A case from predictive cognition, novel tasks and world models—and why prediction does not rule out AI consciousness. https://aronbijl.com/blog/ai-reproducibility-crisis https://aronbijl.com/blog/ai-reproducibility-crisis Fri, 25 Sep 2026 12:00:00 GMT Computer Science LLM evaluation Benchmark bias Reproducibility Analysis of 3,793 LLMs finds real scaling gains, hard capability ceilings and model-dependent benchmark inflation from selective reporting. https://aronbijl.com/blog/rapid-evidence-synthesis https://aronbijl.com/blog/rapid-evidence-synthesis Fri, 25 Sep 2026 12:00:00 GMT Research Evidence synthesis Decision-making Rapid reviews Learn when rapid evidence synthesis is the right method, what a credible brief should contain, and which questions require a fuller systematic review. https://aronbijl.com/blog/scientific-due-diligence https://aronbijl.com/blog/scientific-due-diligence Fri, 25 Sep 2026 12:00:00 GMT Research Scientific due diligence Investment research Biotech A practical scientific due diligence framework for testing mechanisms, reproducibility, translation and hidden dependencies before investing. https://aronbijl.com/blog/claims-substantiation https://aronbijl.com/blog/claims-substantiation Fri, 25 Sep 2026 12:00:00 GMT Research Claims substantiation Product claims Regulatory evidence A practical guide to matching health, supplement and cosmetic claims to the right evidence before copy is approved or a study is commissioned. https://aronbijl.com/blog/clinical-evidence-gap-analysis https://aronbijl.com/blog/clinical-evidence-gap-analysis Fri, 25 Sep 2026 12:00:00 GMT Research Medtech Clinical evidence Evidence gaps How to structure a clinical evidence gap analysis for a medical device, from intended purpose and claims to evidence quality and the next research step. https://aronbijl.com/blog/ai-assisted-literature-review https://aronbijl.com/blog/ai-assisted-literature-review Fri, 25 Sep 2026 12:00:00 GMT Research AI-assisted review Research methods Auditability A practical framework for AI-assisted literature reviews: where models help, where they fail, and which human checkpoints create an audit trail.