askOdin — AI Judgment Infrastructure for Capital Allocation

Press Release

For release ·

askOdin Publishes the Halvard Audit: A Synthetic AI Infrastructure IPO That Scores 0

A fifteen-file IPO data room for a fictional GPU neo-cloud scored 43 on presentation. Checked against itself, it scored 0.

SINGAPORE — October 26, 2026 — askOdin Pte. Ltd. today published The Halvard Dilemma, a forensic case study of a synthetic IPO for Halvard Computing, a fictional AI neo-cloud. askOdin drafted the whole data room with Claude, Anthropic’s model: fifteen files, including the IPO investor presentation, a financial model, a cap table, board minutes, an independent engineer’s report and a power purchase agreement summary. All entities in it are fictional.

On presentation, the primary deck scored 43 out of 100. askOdin Clarity then checked the deck’s numbers against each other and returned four findings, two of them Critical. The deck calls its FY26 to FY28 build “fully funded” on about US$24.1bn of capital, against US$28.89bn of capex in its own schedule, a gap of roughly US$4.8bn. And it markets US$14.4bn of “contracted revenue” whose footnote includes non-binding heads of agreement.

The funding gap cost 20 points. The JUDGE Protocol then applied its Narrative Masking override, a kill shot, and the Clarity Score came back 0.

Each Halvard slide is internally consistent; the contradictions only exist between slides, and between a slide and its own footnote. A summary reports what each slide says. Finding the gap means holding the capital sources on one page against the capex schedule on another.

“A polished data room is the easiest thing in the world to produce now,” said YekSoon Lok, Founder & CEO of askOdin. “We built this one with a frontier model, and it reads like a book an investment bank would market. That is the point. The question an investment committee has to answer is not whether the story is well told. It is whether the story agrees with its own numbers.”

The case study closes with four questions an investment committee should ask before allocating to an AI infrastructure IPO.

About askOdin

askOdin builds AI Judgment Infrastructure™ for private capital. askOdin Clarity tests the claims in a CIM or an investment memo against the financial model behind it, before the committee votes. Its PE & M&A desk tests buyout add-backs against the seller’s databook before the LOI; its venture desk scores decks and data rooms across five pillars on a 0–100 Clarity Score™. askOdin Crucible gives founders the same test for free.

The engine is deterministic and runs on the patent-pending RUNE, RAVEN and JUDGE protocols; a fourth filing, NORN, is on the roadmap. askOdin audits reasoning, not truth. Customer documents are never used for training, raw documents are held under a documented 30-day ceiling, and every output carries a Defensible Audit Log™.

Founded in 2025 by YekSoon Lok. Headquartered in Singapore. Four U.S. provisional patent applications filed. A member of NVIDIA Inception, Google for Startups and Cloudflare for Startups.

Media contact: [email protected] · All press releases and the press kit