Yosgi Yosgi

Making scan results land in the right place

How we split 2D detection and 3D verification into two stages to reduce false positives in digital twin asset annotation, and how that revealed the next recall bottleneck

How we turned free-form 3D exploration into measurable coverage scanning and improved valve inventory recall in a digital twin factory from around 40% to around 90%

A practical engineering recap of how a simple drawing text extraction task turned into a 2D-to-3D mapping workflow for digital twins, combining SVG parsing, vision models, geometry rules, registration …

A real-world MCP-as-Code refactor on where Code helps, where Tool still fits better, and why Agent runtime needs both.

In large-scale Agent systems, pure Tool-based approaches tend to collapse under context pressure, while pure Code-based approaches introduce excessive latency. Based on real-world experiments, this …

Tool and Code do not just differ in expressiveness; they also fail at different times. This article explains why multi-round Agent systems amplify that difference.

AI will not replace software engineers as a whole. It will remove much of the repetitive, mechanical work and give engineers more leverage to focus on judgment, systems, users, and building better …

Multi-tool agents become unreliable when the same task takes different tool paths each time. This article explains why Skills should preserve validated execution paths, not one-off results, and why …

How Cut 6,000 Tokens Down to 500