The stock-picker known as @aleabitoreddit spent Saturday on two fronts: a fresh research thread on Asian holding companies trading below the value of what they own, and a long retrospective defending his record against early skepticism.

The NAV-discount screen. His most substantive new work was a survey of "holdco" structures — parent companies whose market caps sit below the value of their listed stakes (a NAV discount). He worked through several: Priortech, which owns 21% of metrology firm CAMT; Bit Digital, parent of crypto/AI name WYFI, which he dismissed as a messy, diluted discount rather than a clean one. He cooled on the Korean candidates — Iljin Holdings (a transformer and DC-cable parent tied to his AI-power-grid theme) and Simmtech Holdings, a PCB-substrate parent trading at roughly one-sixth of its stake value — citing chronic distrust of Korean corporate governance and arguing those are better suited to activist investors than passive NAV-unlock bets.

The standout was Wistron (Taipei: 3231), which he called the most exciting find since Sivers. Wistron is a Taiwanese electronics manufacturer that owns about 35% of AI-server builder Wiwynn; he flagged roughly 144% year-over-year quarterly revenue growth and, crucially, independent growth rather than a static holding. He also noted Sin-American Silicon (5483), which owns ~47% of wafer maker GlobalWafers (6488) at a steep discount, though with slower standalone growth. His tentative conclusion: he may concentrate further into ACMR and WUS — semiconductor-equipment and PCB names with upcoming H-share listings and, in WUS's case, an active activist campaign — on top of existing positions. He stressed this was still in-progress research, not a final call.

The victory lap. A second, widely-seen post recited a list of ideas that drew heavy backlash before being validated. The throughline reinforces his existing themes rather than adding new ones. On AI photonics — his dominant area — he cited AXTI (indium-phosphide substrates, the material for high-speed lasers), IQE (epiwafers), SOI/Soitec (silicon-on-insulator substrates for co-packaged optics), and above all SIVE/Sivers, the tiny Swedish maker of indium-phosphide light sources he treats as the central laser chokepoint for co-packaged optics. He pointed to institutional buying and partnerships with JBL (Jabil) and GFS (GlobalFoundries) as validation, and name-checked AAOI (Applied Optoelectronics, his Made-in-America transceiver pick) and LITE (Lumentum, the peer he wants Sivers to become). MRVL (Marvell), his custom-ASIC and optical hub, and AEHR (burn-in test for memory/HBM packaging) also appeared.

Beyond photonics he reiterated long-running calls: RKLB (Rocket Lab, his generational space long), RPI (Raspberry Pi, which he framed as edge-AI hardware and which recently posted ~58% forward growth against far lower consensus), EWY (his leveraged Korean memory proxy), and NBIS (Nebius), his preferred neocloud long versus IREN as a short. He took a swipe at the IREN camp for once accusing him of spreading propaganda.

A brief third post addressed his X revenue-sharing income, where he noted his RDDT (Reddit) network of idea-sharing peers — itself one of his core mispriced-software longs.

Nothing today amounts to a position change, but the holdco screen is a genuine new line of research worth watching for Monday.