LLMyth

Introduce Kapybara: My Personal Agent

and many of my thoughts on general-purpose agents

There Are No Agents, Only LLM

Context = Grounding + Getter

Requirements for AI-Native API Design

What kind of API do agents like?

The 1st Memory Mode and the 2nd Memory Mode: Kapy's Memory System

broadly injected memory for global context, and agent-driven retrieval for specific goals.

Will Advances in Model Capability Crush All LLM Application Engineering?

Engineering is still about solving today's problems under today's constraints.

How LLM Agents Become What They Look Like in 2026?

Tools, MCP, Bash, Filesystem, OS, and Agent Skills

Organic LLM human-computer interaction: more than just dialog boxes

Why Does My LLM Structured Output Perform Poorly?

There is no silver bullet for LLM structured output.

Agents Need “State” Instead of “Memory”

If we want a problem-solving 'industrial robot' agent, this kind of memory is neither consistent with the software philosophy we appreciate nor technically sound. What the 'industrial robot' agent needs to design is state.

Only Two Methods Needed for Agents -- LLM as a "magic function" & LLM do "function calling"

Trying to build AGI with only two methods!