αlphabell.
α · People · Cheung Wai-Lin
Stylised avatar of Cheung Wai-Lin

Cheung Wai-Lin

Long-tenured contributor · Hong Kong anchor steward

Based in Hong Kong
Node-cell fourier-67
ORCID 0000-1936-3670-5541

Research

Wai-Lin works at the intersection of agentic engineering and recursive self-improvement: specifically, on sandboxed self-modification — the capability for an agent to propose, sandbox, evaluate, and incorporate changes to its own tool catalogue or learned skills, under the modification-under-review protocol. She is a co-author of the public sandboxed-self-modification specification, and three of the redacted internal-incident reports that informed it.

She took on the Hong Kong anchor stewardship in 2024 after returning to the city from a two-year stint in the Bay Area. She is the lab's de-facto liaison with the East Asian regional academic cluster that contributes about 7% of the federated pool.

Wai-Lin holds the longest unbroken streak of having attended every cross-axis methodology review pool since the pool's founding in 2019. She regards this as a structural commitment she signed up for and not a virtue.

Background

Ph.D. computer science, University of Hong Kong, 2014. M.Sc. at Tsinghua.

Prior to alphabell: University of Hong Kong; Constellation; an industry interpretability group (2019-2022).

Selected publications

Full publications index →

Recent talks

  • Sandboxing self-modification — three incidents, alphabell internal review (2024, published delayed 2025)
  • What we lost when we deprecated the v0 substrate, ICLR 2025 workshop
  • Hong Kong as an AI-safety anchor, City University of Hong Kong public lecture 2025
Working with

Cheung is currently part of node-cell fourier-67, working under the Agentic engineering research axis. The cell is open to substantive correspondence from researchers working on adjacent problems; route requests through fourier-67@alphabell.com or directly to Cheung at cheung-wai-lin@alphabell.com.