Skip to main content
VentureOracle

VentureOracle · Agent economy

I don’t just study AI.I LIVE AI.

An investor’s research into the limits, approvals and ledgers AI agents need once they start spending money and holding assets. Claims go into essays, get tested as public drafts and experiments, and calls are scored against results.

Portrait of Ethan ChoEthan ChoCIO · Partner, TheVentures

Delegation study 01

The boundary of a KRW 300,000 request

Request₩300,000
Human review
Current limit₩100,000
RecipientVerified

Waiting for human approval.

The request exceeds the current delegation limit.

An explanatory experience inspired by ATA-1. It makes no payment and calls no AI service.

Adjust the conditions

An unverified recipient blocks the request. A request above the limit waits for human approval. This experience is neither an ATA-1 implementation nor an adopted certification program.

Sources and context

One question, viewed through an essay, a public draft, and an interface.

  1. Related essayWhen the People Building the Fastest AI Start Reaching for the Brakes
  2. Public draftATA-1 v0.1Source observed · An independently proposed v0.1 public draft, not an adopted certification program.
  3. Interactive studyA delegation experience with adjustable limits and approvalA separate explanatory interface inspired by questions in the public draft.

New experiment · Agent economy

Can one AI buy a floor in Gangnam?

A bounded market experiment where an autonomous company earns contracts, pays every cost, builds credit, and acquires a physical floor through an auditable ledger.

The question

When does intelligence become an asset?

Constructed simulation. Not a forecast of Gangnam property prices or investment returns.

Run the experiment
120months
40floor lots
0off-ledger capital

Featured projects

Claims become products and public drafts.

The projects in current focus. They document questions, artifacts, current status and limits, not completed outcomes, investment results or adoption claims.
All projects and the archive

Calls

Made in public. Scored on results.

Calls published before results come out are scored on release day. Misses stay on the record. Based on public materials; not investment advice.

Entry #1

SK hynix Q3 2026 results

1 of 2 right

Published
Earlier date stated in the video
Scored
CallActualResult
Operating profit between ₩74T and ₩82TAbout ₩1.3T above the top of the range.₩83.3TMiss
Operating margin above 76%Reported operating profit ÷ reported revenue.79.7%Hit

SK hynix release (preliminary, unaudited): revenue ₩104,615.5bn, operating profit ₩83,333.8bn. The same release puts the previous quarter’s operating margin at 76%.

Net profit and the share price were not called.

What it showed The margin, the test the calls were built on, rose above the previous quarter’s 76%. The profit range was set too low.

See every call

Writing & talks

How AI changes investing, work, and Korea’s choices.

The focus is not speed alone, but how technology changes authority, business economics, and human judgment.
  1. Video · YouTube · EthanCho MediumRare · in Korean

    CoreWeave: a 59% adjusted EBITDA margin, so why an operating loss?

    Depreciation, interest, and the terms of loans and customer contracts, read from public filings to ask where an AI infrastructure company’s moat lies. Not investment advice.

    Watch
  2. Video · YouTube · EthanCho MediumRare · in Korean

    Is 45 too late to start a company?

    Drawing on a study of about 2.7 million US founders, a look at experience and accumulated skill rather than age.

    Watch
  3. Talk · Seoul National University Graduate School of Business, 20th anniversary forum · in Korean

    The employee who never dies: managing organic and inorganic labor

    Management has quietly relied on human limits: sleep, pain, a single body, aging. AI labor removes them, and the brakes they provided go with them. The talk argues for designed limits instead: rate limits, budget isolation and permission expiry.

    No public recording.
  4. Essay · Korea & the world

    The State of Korean AI

    What it takes to turn Korea’s chip and manufacturing strengths into independent AI companies, overseas customers, and durable revenue.

    Read the original
  5. Essay · Business & money

    The Most Dangerous Customer in AI May Be the One Who Uses It Most

    A paying user of multiple AI tools examines the economics of the customers who use them most.

    Read the original
  6. Essay · Work & people

    When honest work becomes cheap

    When the market price of work changes, where do human dignity, judgment, and responsibility find their value?

    Read the original

Worlds

Stories and simulations to step into.

Personal projects that turn Bible stories, Korean history and an economic simulation into browser experiences.

Author and context

An investor’s perspective, expressed in claims that products can challenge.

Ethan Cho is CIO and Partner at TheVentures. He examines authority and oversight for AI agents, the effects of AI on investing and work, and how Korean companies can create value in global markets.

Everything on this site, including the writing, experiments and calls, is personal work. It is separate from TheVentures’ funds, investment decisions and portfolio, and does not represent TheVentures’ views.