Prompted and Confused: When LLMs Forget the Assignment
A close reading of why LLM-generated optimisation models can look correct, compile occasionally, and still misunderstand the problem hiding in plain sight.
Cognaptus Insights
Industry thoughts, trends, and commentary.
Skills to Pay the Agent Bills: Why LLMs Need Better Moves, Not Bigger Models
SkillGen shows why the next gain in LLM agents may come from reusable procedural skills, not longer prompts or larger models.
Thresholds, Trade-offs, and the Art of Not Overthinking Your Robot
How calibrated symbolic uncertainty helps robots decide when to act, when to look again, and when confidence becomes expensive.
Tools of Habit: Why LLM Agents Benefit from a Little Inertia
AutoTool shows how agent systems can cut repeated tool-selection costs by learning when workflow habits are reliable enough to bypass another LLM call.
Value Collision Course: When LLM Alignment Plays Favorites
A mechanism-first reading of how human-feedback design choices quietly decide whose values an aligned model learns.
Ask, Navigate, Repeat: Why Socially Aware Agents Are the Next Frontier
FreeAskWorld shows why embodied AI needs interaction as an operational information channel, not just prettier simulation scenery.
Benchmarked Brilliance: How CreBench Rewrites the Rules of Machine Creativity
CreBench shows why evaluating AI creativity requires rubrics for ideas, process, and products—not another beauty contest for generated images.
Ghostwriters in the Machine: How Multi‑Agent LLMs Turn Raw Transport Data Into Decisions
A new multi-agent LLM framework shows how transport analytics can become stakeholder-ready reports, provided we remember it is automating interpretation, not operational judgement.
Graph Medicine: When RAG Stops Guessing and Starts Diagnosing
A mechanism-first look at how retrieval-augmented LLMs can turn clinical guidelines into structured medical knowledge graphs—and why the hard part is still clinical reliability.
LLMs, Trade-Offs, and the Illusion of Choice: When AI Preferences Fall Apart
A new preference-coherence test shows that many frontier LLMs can produce trade-off behaviour, but very few show stable preference structures across AI-specific scenarios.