Start Here
A reverse-chronological archive is not a way in. This page is the way in.
If you have five minutes
The four posts that best show what this blog does — read one primary source properly, verify the claims, and report what actually held up:
- Ouroboros: The Agent OS That Hides the Answer Key From Its Own Workers — a spec-first agent OS where the grading command never reaches the worker. Includes a stdlib reimplementation of its two mathematical gates, plus one documentation defect found by reading the source instead of the docs.
- Twelve Concepts, One Missing Layer — worked through a 12-article agent curriculum and found the load-bearing layer missing from its own summary. Durable execution implemented and verified from scratch.
- Fable 5 Isn’t a Faster Chat Model — the substrate for self-improving agents, read as a systems question rather than a benchmark race.
- Prompt Repetition: A Simple Trick That Improves LLM Accuracy — when a cheap trick genuinely works, and the narrow conditions where it stops working.
By topic
Agents & harness engineering
The main thread. Agent loops, tool dispatch, spec-first workflows, evaluation, and the scaffolding that decides whether an agent survives contact with production.
- DeepSeek Harness: What If the Agent Loop Itself Were Just Another Plugin?
- jeo-code Puts Skills and Approval Gates Around Coding Agents — includes video
- One Memory Setup, Every Harness: Omnigent’s Hindsight Bridge
- OpenHarness Lets Its Allow List Outrank Your Deny Rules
RAG & retrieval
Retrieval that survives real corpora: reranking, hybrid search, graph memory, and the failure modes that only appear at scale.
- Cognee: Building Persistent Memory for Multi-Agent Systems
- From RAG to Context Layer: Ontology, LLM Wiki, HyGRAG
- STORM Parse vs Google Gemini File Search
- RAG or Fine-Tuning? Fine-tuning embedding models for retrieval
Models & papers
Architectures read closely enough to explain, not just cite.
- Diffusion Models, Visually: How Noise Becomes an Image
- Building LLaMA 4 from Scratch: Mixture-of-Experts
- NPGA: tracking and rendering measured separately
- LLM Course — Let’s build a simple LLM
Infrastructure & production systems
What it takes to run this material rather than demo it.
- The Full MLOps Blueprint: Monitoring and Observability
- The Production Generative AI Stack
- Signal-Decision Architecture: Semantic Routing at Scale
Multimodal & vision
Coming from game AI, this is where I started: systems that read a screen and act on what they see.
- 3D Gaussian Splatting vs NeRFs: what is the difference?
- 3D Language Gaussian Splatting (LangSplat) — includes video
- The giant leaps of open-source vision models
Developer tooling
- Unity CLI: From Editor Installs to Verifiable Game-Production Work
- Unity CLI Atomic Agent
- Antigravity CLI Migration: a production-minded checklist
From projects to technical evidence
If you prefer to start with something I built rather than a topic label, use Portfolio. It connects production AI products, public agent tooling, multimodal QA research, and game automation to the articles and repositories that provide the technical trail. The separate visual portfolio provides the full bilingual gallery and career timeline.
By format
Different posts do different work. If you prefer one mode over another:
| Format | What it looks like | Examples |
|---|---|---|
| Runnable code | Companion .py files you can download and execute; every assertion in the post was produced by running them | Ouroboros gates · Durable execution · Supertonic ONNX |
| Video | Embedded walkthroughs and demos | jeo-code harness · Gemini 3 multi-agent · SIMA 2 in 3D worlds |
| Diagrams | Mermaid architecture and flow diagrams | Production GenAI stack · Signal-decision architecture |
| Deep dives | 2,500+ words, single subject, primary sources only | Fable 5 · MLOps blueprint |
| Async & performance | Measurement-led, with numbers | Why async code can be slower — includes video |
How this blog works
A few conventions worth knowing before you read:
- Primary sources over summaries. When a post analyses a repository, I read the source, not only the README — and say so when the two disagree.
- Code is executed, not illustrated. Where a post claims code runs, the output shown is real output. Companion files are downloadable so you can check.
- Findings are dated. Star counts, version numbers, and benchmarks are recorded on the date noted. They will drift; the post says when it was true.
- Mistakes get published too. Several posts document where my own first calculation was wrong. That is the useful part.
- AI assistance is disclosed site-wide. AI can assist research, drafting, translation, diagrams, and code review. Scheduled audits may publish under standing approval after independent evidence review and automated checks, not per-page human approval. I remain accountable for the editorial rules and corrections. The editorial method explains this boundary; AI output is never evidence by itself.
- Legacy notes are reviewed separately. Older reference posts without enough original analysis are removed from search and advertising until they are rewritten or retired.
Publishing cadence
Active since 2024, publishing in research and project bursts. Cadence varies with a full-time engineering job and a PhD: bursts when a topic opens up, quieter when a project is consuming the week.
Who writes this
Jang Young Jeong — AI Product Engineer at Supercent, Ph.D. candidate in Game Engineering at Hongik University, 8 years of shipped AI at NCSOFT and Com2uS. Full background on About; reach me via Contact.