Longer Yet Dumber: Why LLMs Fail at Catching Their Own Coding Mistakes
FPBench shows that many LLM code assistants can detect faulty requirements when prompted, but often fail to question bad premises on their own.
Cognaptus Insights
Industry thoughts, trends, and commentary.
Open-Source, Open Risk? Testing the Limits of Malicious Fine-Tuning
A mechanism-first reading of OpenAI’s malicious fine-tuning study and what it implies for evaluating open-weight model releases.
Reasoning with Both Eyes Open: Why Multimodal Chain-of-Thought Still Trips Up LLMs
Multimodal chain-of-thought looks impressive until visual evidence must be used repeatedly, not merely mentioned.
Thinking in Circles: How Self-Questioning LLMs Learn Without Labels
A mechanism-first reading of SQLM, a self-play post-training method where language models generate their own questions, solve them, and learn from proxy rewards without curated training data.
Add to Cart, Add to Power: What Happens When AI Shops for You
AI shopping agents may reduce search friction, but this paper shows they also create model-specific demand shocks, position bias, and a new market for AI-facing listing optimisation.
Credit Where It's Due: How CAPO Brings Verifiable Precision to LLM Reasoning
CAPO shows how stronger verifier models can turn blunt outcome rewards into step-localised training signals for smaller reasoning models.
Graphs, Gains, and Guile: How FinKario Outruns Financial LLMs
FinKario shows that financial AI performance may depend less on bigger language models and more on converting research reports into dynamic, event-aware retrieval infrastructure.
Love in the Time of Context: Why LLMs Still Don't Get You
Cupid shows that LLM personalization fails less because models cannot write tailored answers and more because they retrieve, infer, and apply the wrong contextual preference.
Seeing Is Deceiving: Diagnosing and Fixing Hallucinations in Multimodal AI
A practical reading of multimodal hallucination research: why visual models appear grounded, where that grounding fails, and how operators should diagnose before they deploy.
Causality in Stereo: How Multi-Band Granger Unveils Frequency-Specific Influence
A practical reading of Multi-Band Variable-Lag Granger Causality and what frequency-specific delays can reveal in complex time series.