Braden Dennis — research hub

Braden Dennis · founder and CEO of Fiscal AI, the AI-native financial-data platform (~150,000 users, ~50 employees) competing with Bloomberg and FactSet. Canadian, engineering background, 30, a stock podcaster for a decade — a practitioner-operator rather than a market strategist: quality/GARP names held for decades, conviction carried by a handful of company-specific KPIs rather than by price.

Sections: stock index · overall thesis · the product · transcripts. Last updated 2026-SEP-09.

▲ Positive

Ticker	Name	Current thesis	Research	Seen in	Total $k
MA	Mastercard Incorporated	Held on identical terms to Visa and always named with it — owned "for a long time", monitored on one payments dashboard (transaction volume + cards in force across V/MA/AXP), price judged reasonable, holding horizon decades.	QT · SA · STK · FA	2026-JAN-28	7.0
UBER	Uber Technologies	The change-your-mind case study: dismissed for two years as "a ZIRP phenomenon, venture-backed subsidized, terrible unit economics", then bought when take rates doubled while total trips tripled — pricing power and usage confirmed together, before the street conceded it could earn a dollar.	QT · SA · STK · FA	2026-JAN-28	2.0
V	Visa Inc.	Long-term holding, potentially for decades — "probably two of the most brilliant businesses ever created" with Mastercard. The whole thesis is reduced to two tracked metrics: combined total transaction volume (~$27–30tn) and total cards in force; interchange-cap headlines are noise unless those break.	QT · SA · STK · FA	2026-JAN-28	0.8

► Neutral / referenced

Ticker	Name	Current thesis	Research	Seen in	Total $k
ABNB	Airbnb, Inc.	Under active research — "starting to look pretty interesting" — but his own gross-bookings cross-comparison contradicted the prior: Booking, not Airbnb, has grown fastest off the COVID lows. No position stated.	QT · SA · STK · FA	2026-JAN-28	—
AMZN	Amazon.com, Inc.	Illustration only — the stand-in for "a company that just reported" in both the incumbents' manual data-factory description and the two-days-to-two-minutes latency claim. No view expressed.	QT · SA · STK · FA	2026-JAN-28	—
AXP	American Express Company	Tracked, not owned — the third name in his payments dashboard (total transaction volume + cards in force across Visa, AXP and Mastercard). He says "them both" of V/MA and "these three companies" of what he monitors.	QT · SA · STK · FA	2026-JAN-28	—
BKNG	Booking Holdings Inc.	The surprise output of his travel-aggregator chart: fastest gross-bookings growth off the COVID lows, ahead of Airbnb and Expedia. Stated as a fact about the numbers, not a stance — a lead, not a call.	QT · SA · STK · FA	2026-JAN-28	—
Bloomberg	Bloomberg L.P. (private)	Private; the industry's benchmark incumbent and the origin of his category label — "Bloomberg killer products have not actually been able to take material market share" because they license data rather than own it. Financial data sized at $46bn/yr with five incumbents as its "arms dealers".	—	2026-JAN-28	—
EXPE	Expedia Group, Inc.	The third leg of his saved "travel aggregators" gross-bookings comparison (with BKNG and ABNB), named only as a comparator. No view.	QT · SA · STK · FA	2026-JAN-28	—
FDS	FactSet Research Systems Inc.	The incumbent he is trying to displace, not a stock view: chapter three is "can I build AI native FactSet better, faster, cheaper", and the manual divide-and-conquer data process he describes is what he says AI removes.	QT · SA · STK · FA	2026-JAN-28	—
Fiscal AI	Fiscal AI (private; venture-backed) — his own company	His own private, venture-backed company (~150,000 users, ~50 employees, all with equity) — a financial data business consumed as a terminal, an Excel add-in and a data-feed API. Chapter three is owning the data content itself (96TB, 2.4m SEC filings, two-day latency cut to two minutes) so every cell can be clicked through to the filing. Listed for the record, not as an investable stance.	—	2026-JAN-28	—
LUV	Southwest Airlines Co.	UI demonstration only — pulled up to show a COVID collapse-and-recovery as a chart, and as the example for querying a single earnings transcript with a custom AI prompt. No view expressed.	QT · SA · STK · FA	2026-JAN-28	—
TSM	Taiwan Semiconductor Manufacturing Co.	UI demonstration only — the name sitting in the dashboard notifications panel during the demo ("Taiwan Semiconductor just posted this event… it's their Q4"). No view expressed.	QT · SA · STK · FA	2026-JAN-28	—

▼ Negative

Ticker	Name	Current thesis	Research	Seen in	Total $k

Overall thesis

In one line: Own a small number of brilliant businesses for decades, and let a short list of company-specific KPIs — not price, not headlines, not a score — decide when to buy, when to ignore the news, and when to leave.

Quality at a reasonable price, held for decades. His stated style is "somewhere in that growth at a reasonable price type area"; absent a steady history of topline growth he is "very very unlikely… going to be a long-term shareholder", special situations and spinouts aside. On Visa and Mastercard — "probably two of the most brilliant businesses ever created" — the horizon is "decades potentially."
Conviction comes from KPIs, not from the tape. He reduces each holding to a handful of disclosed operating metrics and treats everything else as noise: for the card networks, combined total transaction volume (~$27–30tn) and total cards in force; for Uber, total trips and take rate. Users who work the same way build "these are the eight things I care for every company" as a per-name template — the feature he built for them describes his own process.
Willing to be publicly wrong, and to reverse on data. He called Uber "a ZIRP phenomenon, venture-backed subsidized, terrible unit economics" for about two years, then bought it when take rates doubled while trips tripled — pricing power and usage confirmed at once — while "the consensus on the street was that Uber's never going to make a dollar."
The exit is written in the same units as the thesis. A disrupted business shows up in the KPI first, "or even worse, they stop reporting or stop disclosing the number… I would be very much so looking for the exit before other people are." A discontinued disclosure is itself the sell signal.
Screen to exclude before you screen to select. "It's really helpful to filter not what's in the screen, but filter out what's maybe not a good use of my time." His one named sector exclusion: metals and mining — "I don't like junior mining. I think it's just so boom bust, hard to make money. It kind of feels more like a lottery ticket unless you're in the industry."
Explicitly anti-technical-analysis and anti-verdict. The platform is "completely useless if you're looking to draw some technical analysis on a chart"; he calls that "astrology more than investing". He also refuses to build the scoring feature customers ask for: "I don't want to build a scoring system that tells you if a stock is good or bad or valued or undervalued" — and is equally blunt that "letting GPT manage your portfolios hasn't really been linked to good returns."
Trust the number only when you can trace it. Owning the data content rather than licensing it is his whole competitive argument, and the reason is auditability: a buy-sider "needs to be able to click through to the filing." "There's no such thing as a perfect data set in the world… trust is built over me showing you instead of me telling you."
Asset ownership as the stated mission. The formative observation is that the best-off people in his large extended family were not the highest earners but the ones acquiring assets: "you can benefit from this system much much better if you own assets… you feel like the system's cheating you if you don't own assets." He frames capitalism as non-zero-sum, and gives every full-time employee equity in his own company.

The product — Fiscal AI

What it is: an AI-native financial research and data platform founded and run by Braden Dennis, venture-backed, with ~150,000 users and ~50 employees, competing directly with Bloomberg and FactSet in what he sizes as a $46 billion a year industry. Its differentiator is that it owns the data content rather than licensing it — 96TB stored, 2.4 million SEC filings, and an AI reasoning layer that compresses the industry's two-day post-earnings latency to two minutes. Customers are professional analysts and self-directed individual investors, whom he says use it identically; other fintechs license the same feed. There is a free plan (now with 10 years of history) alongside the paid tiers.

Grounded only in what he says on 2026-JAN-28 — this is the first appearance archived here, so the picture is necessarily partial and will be revised as more are processed.

Offering	What it is	How he runs it	Seen in the index
Terminal	The web platform itself: overview page (top-level ratios, business description, trailing-month news, bulls-bears), financials, research estimates, news, ownership, industry, dividends.	"The guts of where most time is spent" is the financials tab — three statements, ratios, KPIs, adjusted and custom metrics, forward estimates, 20 years of history, switchable to quarterly. Click any number and "boom, it builds you the exact data visualization."	V, LUV, AMZN
Excel add-in	The same data pulled into a spreadsheet.	One of the three ways data is consumed industry-wide, alongside a terminal and a feed: "If you want to whip that into Excel, no problem."	—
Data-feed API	The database on tap — the layer other companies license from him, including consumer brokerages: "say you're the Robinhood of country X, Y and Z."	"We're pulling our own API which many other companies also license from us directly." Owning the content is what makes the feed possible at all; the two-days-to-two-minutes latency claim is about this pipeline.	—
KPIs & segments layer	Company-specific operating metrics and revenue segments (by geography, by line item) that sit outside the three financial statements.	The feature "has become a little bit synonymous with what we do", built since chapter one because he saw Substack analysts pulling these by hand into Excel. The platform also flags a discontinued metric — which is his own exit trigger.	V (16.7tn transaction volume), UBER (trips, take rate)
Filings & investor relations	The "unstructured content" — PDFs, reports, earnings-call transcripts and audio — in one viewer, with a side panel for one-shot or custom AI summaries scoped to that single document.	"I don't want to reach in the mind of the entire LLM. I want to reach into the mind of the LLM's extraction of this specific document."	LUV
Cell-to-filing traceability	Every number links back to the exact place in the filing it came from — for US, Canadian, UK and European companies.	The compliance requirement that used to lose him buy-side deals: "I need to be able to click through to the filing." "Trust is built over me showing you instead of me telling you."	V ($40,000,000,000 net revenue)
Screener	A filter over the universe, with sector and industry exclusions.	Used to remove rather than to select — "filter out what's maybe not a good use of my time." He demonstrates it by excluding metals and mining.	—
Charting / cross-comparison	Saved multi-company charts of one metric across an industry.	His own saved sets: payments (transaction volume across V/MA/AXP), trading platforms, and travel aggregators (gross bookings across EXPE/BKNG/ABNB).	V, MA, AXP, ABNB, BKNG, EXPE
Dashboard & watchlists	Personalised home base: watchlists with user-chosen columns, an ownership/portfolio breakdown aggregating your positions "if it was a specific business", and an event notifications panel.	He sorts by forward EV/EBIT; a Taiwan Semiconductor Q4 posting is the live notification example.	TSM
Free plan	A permanently free tier, recently upgraded from four or five years of history to 10 years.	"Of all the ones that exist there, no one is offering 10 years of historicals. There is not one of them doing it."	—
Deliberately not offered	No scoring, no valuation verdict, no technical analysis.	"I don't want to build a scoring system that tells you if a stock is good or bad or valued or undervalued." Technical analysis he calls "astrology more than investing" — the platform is "completely useless" for it by design.	—

How it serves retail investors:

The same product for both cohorts. A professional running billions and an individual running their own money "are looking at the same things… the use case is identical" — so the individual is not sold a stripped-down tier of a professional tool.
Company KPIs that were previously hand-assembled. The metrics that decide a business are disclosed but scattered; he built the layer because thoughtful independent analysts were pulling them into Excel by hand, "because no one had done it for them."
Auditability, not authority. Rather than asking users to trust a cleaned number, every cell traces to the filing — the same standard a buy-side compliance team demands.
No verdicts to lean on. Refusing scores and ratings forces the judgment to stay with the investor: "humans are really good at synthesizing that and making good decisions."
A real free tier. 10 years of history at no cost, positioned explicitly against the other free platforms.
Speed parity with professionals. Results available minutes after the release rather than days later removes a structural information gap that used to be priced.

Transcripts

One dated page per appearance — each has its stock table, talking points, and the saved transcript. Newest first.

Date	Title / analysis page	Show	Link	Transcript	Actionable insights
2026-JAN-28	Braden Dennis (Fiscal AI CEO): How One Frustrated Investor Democratized Wall Street's Data	Talking Billions (Bogumil Baranowski)	▶ YouTube	transcript	actionable insights

To process — backlog

Appearances not yet processed — newest first. None queued yet.