Frequently asked questions

How has AI changed investment research?

It collapsed the cost. A full investment report that took weeks of analyst time now takes hours: one evening I commissioned two, on Cisco and CAVA, and published both. A pre-market swing briefing lands every morning at 8 AM. Portfolio math, sizing scenarios and options structures, runs conversationally in seconds. Research used to be gated by labor; now it is gated by the quality of your questions.

What can AI not do for an investor?

The psychology. No model held Palantir through the 2022 drawdown; the research, the conviction, and the sitting still were all mine. A machine can tell you the thesis is intact, but only a human can watch a position fall 70% and not flinch. Conviction is evidence-gated and slow, and to go big on a ticker you still need to spend a ton of hours on it yourself.

Where is the investor's edge now?

Up the stack: the quality of the questions you ask, the framework you run the answers through, and the temperament you bring to the position after the report is filed. When research is free, the scarce resource is judgment. The differentiator is the human who decides what to do at 9:31 AM, and what not to do for the next three years.

TL;DR: AI collapsed the cost of investment research. A full report that took weeks now takes hours, a configured daily briefing arrives every morning, and portfolio math that needed a quant desk runs in seconds. What it did not collapse is the psychology: conviction, sizing, and sitting through pain are still human jobs, and they always were the actual job.

What can AI research do in one evening?

Tonight I commissioned two full investment reports, one on Cisco and one on CAVA. Financials in detail, segment breakdowns, technical deep dives, partnership maps, competition comparisons, five-pillar gradings, charts, diagrams, and print-ready PDFs. Both are now published on this site's research page.

A few years ago, that body of work was weeks of analyst time. It arrived in hours. Not rough drafts either: sourced, scored, and formatted to a house standard. I spent my time on the part that mattered, asking whether the Hyperfabric story and the new Palantir collaboration changed the Cisco verdict (they deepened it; they did not change it), and let the machine do the assembling.

This is the first change, and it is the biggest one. Research used to be gated by labor. Now it is gated by the quality of your questions.

What is the briefing that never sleeps?

The second change is quieter but more personal. Every trading day at 8 AM, a pre-market swing report lands with entries, stops, and targets for that day's setups. It is configured exactly the way I want it: long-only, US large and mid-caps, technical levels, no noise. I did not hire a research desk. I built a pipeline that behaves like one.

This used to be the edge of institutions. A Bloomberg terminal, a morning meeting, a team of juniors screening overnight. Now the screening, the charting, and the briefing are a configured routine that costs almost nothing to run. The retail investor of 2026 wakes up to the same morning stack the professional had in 2016.

How did the math become trivial?

The third change is the modeling. Portfolio balancing, scenario analysis, options structures: the arithmetic that once needed a spreadsheet jockey or a quant now runs conversationally. Ask what happens to the portfolio if a position doubles, halves, or gets cut in thirds. Ask what a cash-secured put at a given strike does to the entry math. The answers come back in seconds, with the assumptions visible.

This matters more than it sounds. Good sizing decisions were always math plus temperament. The math half just became free, which means the temperament half is now the entire game.

What can the machine not do?

Here is what AI has not changed, and I doubt it ever will: the psychology.

No model held Palantir through the 2022 drawdown. The research was mine, the conviction was mine, and the sitting still was mine. A machine can tell you the thesis is intact. Only a human can watch a position fall 70% and not flinch, because flinching is not a data problem. It is a stomach problem.

CAVA nearly doubled, then round-tripped back to about my cost. The audit said thesis intact, price pain rather than thesis pain. The machine ran the numbers in minutes. Living with the round trip for months was the human part. There is no prompt for that.

This is the durable truth about investing that no tooling touches: the engine's first job is to make mistakes small and convictions big, and conviction is evidence-gated and slow. Borrowing conviction is fine for starting a position. To go big on a ticker, you need to spend a ton of hours on it. Those hours are still yours to spend. AI can compress the research; it cannot compress the seasoning.

What is the new bottleneck?

Put it together and the investor's job has been redefined, not replaced. Information access is no longer the edge. Everybody gets the same machine. The edge moved up the stack: to the quality of the questions you ask, the framework you run the answers through, and the temperament you bring to the position after the report is filed.

That last one is the whole ballgame now. When research is free, the scarce resource is judgment. When every investor has an analyst that never sleeps, the differentiator is the human who decides what to do at 9:31 AM, and what not to do for the next three years.

The analyst is now a machine. The investor still has to be human. That was always the harder job anyway.

Research and opinion, not investment advice. Do your own due diligence before investing.