Robotics & Embodied AI
RoboLLM
Robot workbench — webcam hand-teleoperation, humanoid motion mirroring, and LLM-driven robot control experiments on ROS 2.
Explore projectA map of things I’m
building & learning.
Harness & meta-harness engineering, AI agents and robotics — shipped end-to-end.
01Robotics &
Embodied AIPerception, action & learning02Agent SystemsTools for dependable intelligence03ProductsIdeas made usefulA MAP OF INTERESTS, ALWAYS EVOLVING
I'm Santapong Sondhi, a software, AI and robotics engineer. I build tools, explore how intelligent systems interact with the world, and write about what I learn along the way.
A little more about meFEATURED PROJECTS
All projects 
Agent Systems
Accretion
Observable, deterministic meta-harness for local AI coding agents — supervises Codex and Claude Code through one control plane with static strategy selection, verifier-gated bounded loops, isolated Git worktrees, a durable event trace and the ACR-ARCH selector benchmark.
Explore project
Products
StockFlow
Smart-Inventory — stock and cash-flow SaaS for Thai SMEs. Pools inventory from Shopee, Lazada, TikTok Shop and LINE OA into one place and tells you tonight's cash position, with a nightly LINE summary.
Explore projectFROM THE FIELD JOURNAL
All writing Thinking out loud. Testing in practice.
SEO and GEO for a Personal Site: The Afternoon Playbook
How this site went from misattributed to ranked #1 for my name and legible to AI answer engines in an afternoon: canonical domain, per-page metadata, JSON-LD, a dynamic sitemap, Search Console as a diagnostic — then the GEO layer: AI-crawler allowances, llms.txt, RSS, and content concrete enough to quote.
Embodied AI vs Cross-Embodiment AI: Where the Line Actually Sits
Embodied AI is the field: intelligence closing a perception–action loop against physics. Cross-embodiment (X-embodiment) is a scaling bet inside it — and the RT-X numbers say it pays off asymmetrically: 27.3%→75.8% at 55B params, but worse-than-solo at 35M. Diversity beats volume, human video is the cheapest body, and the evaluation harness matters at every scale.
AI Proposes, Safety Decides
RoboLLM's four principles in practice: one ROS 2 bridge for MCP and dashboard, a physical-arm chain where no model ever touches PWM, gesture pick-and-place and fist-triggered wall welding with collision-verified standoffs — and why it is the same shape as a meta-harness.
CURRENT INTERESTS
More about me Three threads, one curious mind.
Robotics & Embodied AI
How perception, learning and action come together in embodied systems.
Explore the work
Agent Systems
How to make agent workflows observable, bounded and easier to trust.
Explore the work
Products
How engineering becomes useful software for everyday work.
Explore the work