AI-Native Investment Firm
Built for the next research cycle.
Markets do not suffer from a lack of information.
They suffer from too much of it.
Third Wave Capital was built with AI woven through the research process itself — reading, comparing, testing, remembering — so that human attention goes where it earns the most.
The final call is human. It always will be.
Why AI-native
The bottleneck has moved.
For most of market history, the edge was access. Better calls. Cleaner data. Earlier datapoints. A model with one more tab than the next person.
Some of that still matters. It always will.
But information stopped being scarce. Markets produce more of it every quarter — filings, transcripts, datasets, commentary — and an analyst's day does not get longer.
The constraint now is attention. What deserves another hour? What changed since the last filing? What is the market still treating as true? Where has the evidence moved faster than the narrative?
Third Wave Capital is built around that shift. We did not add AI to a research process. We built the research process around it.
The turn
The most sophisticated investors in the world are re-tooling.
Bridgewater now runs a multi-billion-dollar strategy in which machine learning does the analytical heavy lifting, with humans setting the boundaries and owning the risk. Man Group's agentic systems propose signals, write the code and run the backtests before a researcher ever sees the output. At Balyasny, nearly every investment team works alongside AI daily — analysis that took days now lands in hours.
Even the sceptics have turned. AQR spent years doubting machine learning in public; it now hands roughly a fifth of its flagship trading signals to the machines.
The largest firms are spending fortunes rewiring processes built over decades. That is the hard way to become AI-native.
We were founded in the AI era. There was nothing to bolt on — and nothing to unlearn.
95%
of hedge fund managers now use generative AI in some form — up from 86% two years earlier.
58%
expect AI to move deeper into the investment process itself over the next year, nearly triple the share in 2023.
6%
let the machine make the final investment decision. On this, we side with the majority — deliberately.
AIMA global hedge fund manager survey, 2025; industry reporting, 2026.
Every firm now says it uses AI. The question that matters is where the machine stops and the investor begins.
The rest of this page answers it.
Internal process
Research should compound.
A thesis should not disappear into an old memo.
A useful datapoint should not need to be rediscovered six months later.
A mistake should not vanish once the position is closed.
Our internal process is designed to keep the firm's work alive: what we read, what we believed, what changed, what surprised us, and what we got wrong. Most firms hold that in the heads of whoever was in the room. Ours holds it in the process — where it accumulates.
Markets rhyme often enough that forgetting is expensive.
Research memory
Everything the firm reads, writes, tests and learns stays alive and queryable — so the next piece of work starts where the last one ended.
Diligence discipline
Every idea is forced through the same questions: what matters, what changed, what is priced, and what would change the view.
Primary sources
Filings, transcripts, data and market signals sit at the centre of the process. Every claim traces back to a source a human can check.
Parallel work
Ten lines of inquiry can run at once instead of one. Conviction gets earned against more evidence, sooner.
Scenario work
The base case matters. So do timing, downside, positioning and the path to being wrong.
How an idea moves
Same discipline. More surface area.
The process is the one good fundamental investors have always run. What changes is how much of it the machines can carry — and how little of it gets skipped when time is short.
Source
Machine-widenedAn idea can arrive from a screen, a filing, a chart, a sector theme, or something that simply looks odd. The sweep runs wider than any one analyst ever could.
Evidence
Machine-widenedFilings, transcripts, guidance history, data. Machines read all of it — and hold it against everything the firm has read before.
Model
Machine-widenedNumbers rebuilt from primary sources, not inherited from someone else's spreadsheet. History checked. Assumptions chosen, and owned, by the analyst.
Thesis
Machine-widenedWritten down in plain language: what changed, who benefits, what is already priced, and what would prove the view wrong.
Risk
Machine-widenedThe bear case run properly. Crowding, timing, positioning, and what the downside looks like if we are early, late, or mistaken.
Decision
Human-ownedA person puts capital at risk. That is not a workflow step to be automated. It is the job.
Review
Machine-widenedEvery outcome — skill, luck, timing, or mistake — is written back into the firm's memory, so the next idea starts smarter than the last.
Machines widen every stage of the process. One stage never leaves human hands.
What changes
More questions before conviction hardens.
The lazy version of AI in investing is just more output.
More screens. More summaries. More memos. More noise.
The more interesting change is the ability to ask better questions earlier — and to ask all of them, every time, not just the ones there was time for.
- ▸Compare management's guidance to what they actually delivered, every quarter, eight years back.
- ▸Read every peer transcript this season. What did competitors say about pricing?
- ▸Which estimate stopped moving — and when did it stop?
- ▸Rebuild the bear case as if we were short. What is the strongest version?
- ▸What did we believe the last time we owned this name, and what did we get wrong?
Questions like these used to cost a week of analyst time each. Now they are the default — asked of every idea, on every revisit.
This does not make investing easy.
It makes lazy research impossible to excuse.
How we think
The process matters most when markets move.
Start with what would change the view
Before we put capital to work, we try to write down what would make the thesis weaker, crowded, late, or wrong.
Consensus can still be right
We are not contrarian for sport. Sometimes the obvious answer is the right one. The question is whether the price, evidence and timing still leave room for return.
The story has to connect
A chart is useful. A pattern is useful. A backtest is useful. None of them is a thesis. The work is connecting what changed, who benefits, how it shows up in numbers, and when the market may care.
Keep the trail
Before a position, write the view clearly. During a position, track whether the evidence is improving or deteriorating. After a position, work out whether the result came from skill, luck, timing, or a mistake worth remembering.
Signal over theatre
Markets produce noise. So does the internet. The job is not to react to everything. It is to know what matters, what changed, and what is just another loud headline.
Judgment is the constraint
AI widens the field of research. It reads more, compares more and remembers more. It cannot own the risk. That stays with us.
The edge is in what the process notices before the market cares.
Third Wave Capital is an AI-native investment firm — not because we bought the tools. Everyone has the tools.
Because the firm was designed so that memory, evidence, repetition and judgment compound.
The machinery matters.
The questions matter more.
AI supports our research process. It does not replace professional judgment. Nothing here constitutes investment advice. See our full disclaimer.