Seeing Is Thinking: When Images Do the Reasoning
A mechanism-first reading of why visual generation helps reasoning only when the task needs a visual world model, not whenever a model can draw.
Cognaptus Insights
Industry thoughts, trends, and commentary.
When Benchmarks Forget What They Learned
A practical reading of memorization-heavy evaluation: why models that remember too well can still be risky, and why controllable forgetting may need to be designed into training itself.
FadeMem: When AI Learns to Forget on Purpose
FadeMem shows why scalable AI agent memory may depend less on storing everything and more on governing what should fade, merge, or survive.
From Indicators to Intent: When Trading Libraries Grow Up
A sharper look at why the strategyr refactor matters: not because it adds more indicators, but because it clarifies where market description ends and trading intent begins.
When Empathy Needs a Map: Benchmarking Tool‑Augmented Emotional Support
A mechanism-first reading of TEA-Bench, showing why tool-augmented emotional support agents need grounded context, selective tool use, and careful evaluation—not just warmer wording.
MemCtrl: Teaching Small Models What *Not* to Remember
A mechanism-first reading of MemCtrl, a lightweight memory-control method that teaches small embodied AI agents to filter observations before they flood context.
Metric Time Without the Clock: Making ASP Scale Again
A mechanism-first reading of how metric temporal ASP can avoid the grounding explosion by moving time from Boolean atoms into difference constraints.
REASON About Reasoning: Why Neuro‑Symbolic AI Finally Needs Its Own Hardware
A systems-level reading of REASON shows why neuro-symbolic AI may bottleneck not on neural inference, but on the messy symbolic and probabilistic reasoning that makes it useful.
Sequential Beats Parallel: When Deep Research Agents Learn to Reflect
A practical reading of Deep Researcher Reflect–Evolve, and why enterprise research agents may need shared memory and plan reflection more than larger swarms.
SokoBench: When Reasoning Models Lose the Plot
A mechanism-first reading of SokoBench, showing why long-horizon planning failures in reasoning models begin with fragile counting, state tracking, and world representation.