📚 Learning, 👨🏻💻 Building, 📝 Documenting...
Hi! I'm Duncan Leung and I currently lead Frontend Engineering at Airvet.
In a previous career I worked overseas in Beijing as a Product Manager at a skunkworks studio building digital experience projects for Ford Motor (China) and consulted on UX projects for Starwood Hotels. During this time I also mentored startups at Chinaccelerator and ran startup events and hackathons in Shanghai.
👨🏻💻 I'm a self taught software developer, and this is a place for me to jot down some notes. I mostly cover Javascript, Typescript, React, and software development.
LLM context rot isn't a bug — it's a mathematical property of how the attention mechanism works. Three interacting failure modes (softmax's sum-to-one constraint, RoPE's distance decay, and causal masking's primacy bias) combine to produce the U-shaped attention curve where models attend well to the beginning and end of context but lose the middle. Understanding the mechanism changes how you structure prompts and manage context.
LLM context windows have a marketed size and a useful size, and they're not the same number. Research shows quality degrades gradually based on absolute token count — not fill percentage — and a 1M-token model at 40% full is far more degraded than a 200K model at 40%. Here's the research behind the Smart Zone, the Warn Zone, and the Dumb Zone — plus a Claude Code statusline that tracks which zone you're in and how deep.
Opus 5 scores 96.0% on SWE-bench Verified vs Opus 4.6's 80.8% — a 15-point gap. But on blind user-preference coding leaderboards, Opus 4.6 ranks 4th (ELO 1547) while Opus 5 ranks 11th (ELO 1530). Synthetic benchmarks and human preference are telling opposite stories. I ran a two-round adversarial research workflow to figure out which signal to trust and landed on a mixed-model strategy.
The retrieval pipeline returned chunks that mentioned vaccination schedules but didn't contain the actual schedule data. The model saw evidence that was on-topic but empty of the answer - and correctly refused. The root cause wasn't the model, the reranker, or the query. It was the chunking strategy splitting tables and structured data away from the prose that named them.
Same question, same branch, same retrieval, same 10 evidence chunks with identical rerank scores - and one deployment refused while another answered with full clinical detail. The root cause wasn't retrieval. It was the LLM making a borderline judgment call on evidence that covered half the question well and the other half barely at all.
Stay updated, receive the latest post straight to your mailbox
