AI & the economy

Research in progress

Runs on Brain →

My independent study on how AI is reshaping work and markets. I write the paper. The system behind it finds sources, checks every one against the original, and keeps the argument honest.

  • 5research questions
  • 164sources found and graded
  • 25%minimum counter-evidence
  • 2channels: labour and capital markets

THE ARGUMENT

Substitution now, complement later

The claim the paper makes, and the test it has to pass.

  • Short run: substitution

    AI substitutes for entry-level white-collar work in exposed occupations, like software development and marketing.

  • Short run: demand for the trades

    The infrastructure buildout raises demand for the skilled trades that build it.

  • Long run: complement

    AI becomes a complement wherever it augments human judgment, distribution or proprietary data.

  • Markets ran ahead

    Capital markets have priced in substitution ahead of both the labour data and the cost economics. AI valuations are a bubble to the extent they assume AI labour is cheaper than human labour at costs deployments do not yet reach.

HOW IT WORKS

How the paper is built

How the paper is built: questions in, every source verified, the argument tested against its confounds, and only checked sources reach the page. The writing is mine.

Interactive diagram. Select a card to highlight what it feeds and what feeds it; press Escape to clear.

Who does the work:

UNDER THE HOOD

How the research holds up

  • 164 sources, every one graded

    The first research pass alone produced 91 verdicts: 49 use, 18 check, 17 fix and 7 drop.

  • It caught fabrications

    A made-up central bank letter, a misattributed figure and an invented quote were caught before any of them reached the paper.

  • Leads, not answers

    Research models are used to locate known items, never to fill gaps.

  • A counter-evidence quota

    At least 25% of sources must challenge the thesis.

  • Confounds named up front

    Interest rates, post-pandemic over-hiring and a software R&D tax change are each weighed before a labour claim is made.

  • One confound, four readings

    On interest rates: two papers for, a reply and an arbiter, read together before taking a position.

  • A test that can fail

    If the cost of AI falls fast enough to close the gap, the bubble claim weakens, and the paper says so.

  • Unquotable until checked

    The interview transcript cannot be quoted until I have checked it against the audio, spot by spot.

  • Two channels, on purpose

    Labour and capital markets only. Topics that pulled the argument off course were cut, with the reasons logged.

HOW I DESIGNED IT TO WORK

Sources are claims until verified

Two rules the whole study runs on.

“No source is cited until it is USE.”

On citation

“Claims to test, not evidence.”

On the tech manifestos

WHAT'S NEXT

Next, in order

  1. Check the interview transcript against the recording, then regrade it.

  2. Decide the paper’s position on the interest-rate confound.

  3. Retrieve the sources that need library access.