Private beta
Research that reads the gaps.
Gapreader runs a team of AI agents that monitor and research investment themes and theses for you. Scouts gather the evidence, analysts build the view, a challenger argues the other side, and you get a brief that says what changed and why it matters.
Every claim cites its source. Every call is scored.
Request early access Agents run on Claude models by Anthropic.
Why the gaps
A supplier says demand is constrained by supply through next year. A buyer says it is designing the part out. A tracker says prices are still rising, more slowly. An analyst raises the estimate. None of them is lying, and none of them is the whole story.
Synthesis is matching what the supplier, the buyer, the competitor, the independent data and the analysts each say about the same question, and reading the gaps between them. That is the work Gapreader's agents do every day, and the record they keep is one you can check.
How it works
A pyramid. Information filters up; only what matters reaches your desk.
-
Scouts
Read primary sources: filings, results calls, industry trackers, regulators, market data. Each entry carries its date, its URL, a provenance label and a perspective: management, buyer, competitor, independent data, financier, regulator, sell-side or press.
-
Theme analysts
Maintain a living note per theme: the mechanism, the numbers, the evidence for and against, what is already priced (stated as inference), checkable watch conditions with their bands set in advance, and open questions with a way to find out.
-
The Challenger
Reads the raw evidence before the analyst's note, so it is not anchored. Argues the other side, tests every claim for stacked sources, and writes a pre-mortem on every position before it is taken.
-
The chair
Decides what reaches you. A brief of a few hundred words, written like a president's brief: what changed, why it matters, what to watch. No process, no debate history. The disagreements that changed a view, and nothing else.
One sentence, end to end
A memory maker reports its quarterly DRAM prices rose "high-teens percent", with the passage quoted, the filing linked, the date stamped.
Scores it against the bands the note set before the release: 13 to 20 percent is deceleration, not a turn. Updates the view, cites the entry, leaves the prior wording in the record.
Points out the figure blends in a product priced flat, so the core line probably rose more; marks the range as inference and names where to find out.
"Prices rose high-teens, inside the band set before the release: deceleration at one supplier, not a turn. The figure blends in a flat-priced line, so the core probably rose more; the record cannot show how much."
What you get
A daily brief
The settled view and the live disagreements, in a few hundred words. A fuller edition each week.
Theme radar and calendar
What is live, what moved, and every dated catalyst with the tell to look for.
Watch conditions
Checkable lines, with the bands set before the release rather than after it.
Predictions, scored
Dated calls with explicit probabilities, resolved and scored for calibration.
Standing price models
For the names you care about, refreshed after every report and every mark that moves.
A model book
A paper long/short portfolio benchmarked against the S&P 500 at equal risk, so the whole system is scored, not just its opinions.
Principles
- Every claim cites an entry you can read.
- Nothing sourced is deleted. Corrections are marked, never erased.
- "Priced in" is an inference, and it is labelled as one.
- A stance carries a dated invalidation, or it is not a stance.
- The record scores itself, and reports the noise alongside the result.
- You make the decisions. Gapreader is decision support, not advice.
Who it is for
Investors and research teams who hold views over quarters, not minutes, and want them tested every day by something that never tires of reading.