Codifying a Tacit Tradition: A Position Paper on the elix-researches Program for Vietnamese Chuyên Tin / IOI-Pipeline Pedagogy

Vietnam has placed at the International Olympiad in Informatics every year since 1989 [1] and operates a national network of seventy-seven specialised informatics (chuyên Tin) classes feeding the pipeline — an elite sliver of order 103 active students per cohort year, less than ∼ 0.01% of the MOET Circular 32/2018 Tin học population. This paper is about that sliver, not the general K-12 population. The pedagogy that produces the country's IOI result is carried almost entirely as practitioner-tacit knowledge in textbooks, on the VNOI wiki, and through an alumni-mentor network; a Scopus / Web of Science / VCGate scan of Vietnamese- and English-language venues over 2000–2025 (documented in M1 §M1.2) returns no peerreviewed publication of its theory or its evidence base. This paper introduces elix-researches, a theoretical, document-analytic, and simulation research program whose object of study is precisely this tradition — chuyên Tin / VNOI / IOI-pipeline pedagogy — viewed under a competitive-programming lens. The program does no cohort work, does not touch minors, and does not gate execution on legal or regulatory review; admissible methods are documentary, corpus-analytic, simulative, LLM-in-the-loop, expert-adult elicitation, and meta-analytic. We commit, here and explicitly, to the chuyên Tin / IOI-pipeline target population under ADR‐002‐target‐population‐commitment.md, and bracket the broader Layer-C non-chuyên question as deferred (not denied; reversibility conditions codified in the ADR). We name the program's first-mover contribution as a four-layer novelty stack: a Vietnamese-canon ↔ IOI Syllabus 2024 crosswalk (N1), a four-way comparative pedagogy matrix across Vietnam / USACO / CSES / Codeforces (N2), a variation-axis coverage schema operationalised against VNOJ (N3), and — conditional on T3 resolution and pending ADR-001 ratification in the V1.5 subproject — the Spine-Cluster framework for Layer-D proceduraltuning under Anderson ACT-R + Chi-Glaser + Ericsson bounds (N4, forthcoming). The empirical surface is novel-problem generation through an LLM-in-the-loop pipeline (E4.2), expertadult elicitation against IOI alumni and chuyên coaches, and Spine-Cluster simulation under BKT / ACT-R-grounded learner models. We position the program against Vietnamese literature (sparse, practitioner-dominated, no programmatic line on K12 informatics pedagogy) and international literatures (CSTA, CAS / PRIMM, Brennan-Resnick, Olympiads in Informatics) and report three honest seams: the T-1 target-population question (now resolved by commitment), the T-2 memoised-recursion /