Words the app uses

The publisher's standing call on a stock — Buy, Hold, Trim, or Avoid. It's an opinion, not advice.

Every stock we cover carries a single label — Buy, Hold, Trim, or Avoid — that summarises the editorial line on the evidence as of the last refresh.

Buy reads as materially favourable, Hold as balanced, Trim as mildly unfavourable with risks rising, and Avoid as materially unfavourable. The label is the same for everyone; it does not consider what you own or what you paid.

Tap a verdict pill anywhere in the app to jump to that ticker's full profile and read the supporting bull, base, and bear cases.

Browse every Buy in the screener →

Related · Confidence, Bull / Base / Bear, Why-now evidence, Citation, Publisher posture

How sure the model is in the verdict, 0% to 100%. Higher means the evidence pointed cleanly in one direction.

Confidence is a number from 0% to 100% that the verdict model attaches to its own call. A 90% Buy means the signals overwhelmingly pointed one way; a 55% Hold means the evidence was genuinely mixed.

It is not a probability of being right. Treat it as a measure of how clean the evidence was, not how likely the stock is to go up.

Related · Verdict, Debate model

Three narratives on every verdict: the optimistic case (bull), the most likely (base), and what could go wrong (bear).

Every full verdict ships with three short narratives.

The bull case is what the optimist sees — the upside if things go right. The base case is the most likely path given current evidence. The bear case is what breaks the story — the risks that could make the verdict wrong.

Reading all three is the point. A high-conviction Buy with a thin bear case is a different signal from a high-conviction Buy with a serious one.

Related · Verdict, Strongest counter-argument, Citation

Every verdict ships with the single strongest argument AGAINST the call, and why the system still landed where it did.

The bear case is one of three views of the stock. The strongest counter-argument is something different — it is the single sharpest argument against THIS verdict, written as a thoughtful skeptic would put it, plus the reconciliation that explains why the weight of evidence still landed on the chosen label.

It exists because the single most expensive failure mode in retail investing is reading only the side of the case that already matches what you believe. Publishing the publisher's own disagreement — on every verdict, not just the high-conviction ones — is a deliberate inoculation against that.

When the adversarial-review (debate) model runs, it sharpens this counter and the 'refined by debate' badge appears. Most verdicts carry a verdict-stage steelman; the debate stage upgrades it on the cases where being wrong matters most.

Related · Verdict, Bull / Base / Bear, Debate model

The handful of signals that pushed the verdict in its current direction. Each chip is tappable for a deeper AI explanation.

Beneath every verdict are a few coloured chips labelled with the specific signals that moved the model — an insider buy, a transcript tone shift, an unusual options print, a filing diff.

Green chips lean bullish, red chips lean bearish, grey is neutral context. Tapping a chip pushes a focused question into the AI sheet so you can dig into that one signal without losing your place.

Related · Citation, Form 4, Insider buy

A traceable claim → source mapping. Every verdict must point at the specific filings, prints, or data points behind it.

Verdicts are required to cite. Each claim the model makes is tied to a source (Edgar, FRED, OpenInsider, Quiver, transcript), a one-line claim, and the value pulled.

If a verdict can't cite, the pipeline doesn't ship it. This is what separates the verdict from a generic chatbot summary.

Related · Compliance guard, Verdict model

Two or three falsifiable, dated, observable conditions whose firing would change the publisher's read. Things to watch — not predictions.

Every verdict ships with two or three triggers: specific conditions that, if observed, would push the publisher to revise the call. Each trigger states the condition in conditional form ("if gross margin falls below 70%"), names where to verify it (a 10-Q line, a vendor datapoint, a transcript section), and gives a concrete horizon ("next print", "by FY27 Q1", "within 90 days").

Triggers are tagged by direction. On a Buy or Hold the publisher mostly lists bear triggers — what would weaken the read. On a Trim or Avoid the bias flips to bull triggers — what would strengthen it. The reader returns to the app when one of their triggers fires.

Triggers are conditional commentary on the security — never forecasts of what will happen, never instructions to act. The Lingley compliance guard re-checks each trigger's condition, observable, and horizon for that posture before the verdict leaves the pipeline.

Related · Verdict, Bull / Base / Bear, Why-now evidence, Compliance guard

The coloured stripe down the left edge of each feed card: green = bullish framing, red = bearish, grey = neutral.

The small uppercase label above a section — borrowed from newspaper layout. It tells you what kind of section follows.

The pop-up chat — a bottom sheet on mobile, a right drawer on desktop — that you can ask anything about the stock you're looking at.

The chat is the same model pipeline as the verdict, exposed as an ask-anything text box. Replies stream a sentence at a time and arrive with the same citations a verdict carries — every claim points back to a filing, a print, or a data point.

On any stock profile, the chat already knows which ticker you're on and has pre-loaded that ticker's full data bundle, recent transcripts, feed history, and latest verdict into context. You can ask follow-up questions without re-naming the company — "why is the bear case thin?", "compare margins to AMD", "what changed in the last 10-Q?" — and the model has the inputs to answer specifically rather than generically.

The chat is intentionally blind to your portfolio. It knows what stock you're viewing, not what you own — that's the publisher-posture line, and the chat composer enforces it.

Related · Spatial context, Pin a ticker, Citation, Publisher posture

The fact that the AI chat tracks which page you're on. Open NVDA, ask "what's the bear case?", and it knows you mean NVIDIA.

Every stock profile registers its ticker with the chat so the model always knows what page you're reading. Walk from NVDA to AMD and the chat's context follows you — the next question lands against AMD's bundle, not NVDA's.

This is what lets the chat handle terse questions like "why now?" or "how does this compare to the last quarter?" without you having to restate the ticker every time. It is also why the chat composer never sees your portfolio: spatial context is about the page, not about you.

Related · AI chat sheet, Pin a ticker

Lock the chat to a specific company so the context stops following the page. Useful when you want to keep digging into NVDA while you browse elsewhere.

The chat's default behaviour is to follow the stock you're currently viewing. Pinning overrides that: the chat will keep answering against the pinned ticker even after you navigate to other pages or the feed.

Unpinning returns the chat to following the current page. The pin is the one-tap escape hatch for the times you want to keep a research thread open while you compare against other tickers.

Related · AI chat sheet, Spatial context

Three ways the homepage feed can be wired — primary filings only, tech-voice aggregates only, or both blended with filings tiered higher inside the same hour.

The feed has three operator-selectable modes. Legacy reads only primary-regulator filings — EDGAR 8-K / 10-Q, Form 4 insider trades, FRED macro releases. Influencer reads only anonymized tech-voice aggregates — paraphrased ticker mentions counted across a curated universe of ~50 voices with the identities stripped. Blend interleaves both and deduplicates per ticker: when two cards land for the same company inside a 60-minute window, the primary filing wins over the influencer aggregate.

The current mode is set by the operator in env (FEED_SOURCE) and is the same for every reader. The card layout doesn't change; the data source does. The sentiment stripe and verdict pill render the same way regardless.

Related · Tech voice aggregate, Sentiment stripe, Form 4

A feed card built from anonymized mentions across a curated universe of tech voices. Cards never name a person — they show our paraphrase plus a count.

The influencer pipeline ingests public movements from a curated list of roughly 50 tech voices across Bluesky, Mastodon, X (via Nitter), YouTube, Substack, and 13F filings. A Claude extractor pulls ticker mentions, sentiment, and conviction; an aggregator buckets them per ticker per window and emits a single anonymized card.

By design, the public surface never names a person — only the publisher's own paraphrase plus a count of voices and an anonymized category mix. The named-person tables (influencers, posts, signals) are service-role-only; only the aggregate table is anon-readable. A name/handle leak detector backs that up at the language layer.

Related · Feed mode, Sentiment stripe

The filter-and-sort table for the whole coverage list — pick a verdict, a sub-sector, sort by P/E or market cap, and read across.

The screener at /screener is the long view of the catalog. Filter by verdict (Buy, Hold, Trim, Avoid), by sub-sector (the editorial tech taxonomy — ai-infrastructure, semiconductors, cloud-platforms, enterprise-saas, and so on), or by a free-form ticker search, then sort by any column.

Two patterns it's good for: scanning every Buy in a sub-sector to triangulate why the publisher leans the same way across peers, and sorting by P/E or market cap to find outliers worth a deeper read.

Open the screener →

Related · Verdict, Sub-sector, Market cap

Multi-ticker chart that rebases every line to the same starting value so percent moves are directly readable across stocks at very different prices.

The compare view at /compare takes a comma-separated list of tickers and renders each one's price series rebased to 100 at the start of the window. Lines above 100 are up, below are down, and the spread between two lines is the relative performance gap.

Rebasing is the point: without it, a $400 stock would dominate the axis even when it barely moved. With it, the chart reads as percent move, not as dollar price.

Compare NVDA, AMD, MSFT →

Related · Rebased, Sparkline

The standing explainer of how verdicts are built — the data sources, the signal blend, the three-model AI pipeline, and the compliance layer in front.

The methodology page at /methodology is the long-form companion to the glossary. It walks through every data source the pipeline reads (EDGAR, FRED, OpenInsider, Quiver, transcript providers), every signal module that scores them, the three-model Claude chain that turns the score sheet into a verdict, and the non-AI rules layer that audits the output before it ships.

When a reader wants to know not just what a verdict says but how it was made, this is the page to send them to.

Read the methodology →

Related · Citation, Router model, Verdict model, Debate model, Compliance guard

Your holdings and watchlist live in your browser's localStorage. The server never sees them, the AI never sees them, and we don't link your broker.

Tickers, share counts, cost bases, watchlists, even profile names — every byte of portfolio data lives in your browser's localStorage. There is no holdings table on the server, no Plaid linkage, no broker handshake. Clear your browser storage and the portfolio is gone; we have no copy.

This is a deliberate compliance posture, not a missing feature. The instant the server holds your holdings, the publisher becomes an adviser and the regulatory bar moves from "publish opinion" to "register as an RIA, run KYC, file Form ADV." Keeping the portfolio client-only is what lets the verdict stay an opinion, the chat stay blind to your positions, and the product stay shippable.

Practical consequence: the AI chat never knows what you own. "Should I sell my NVDA?" gets the same answer as "what does the publisher think about NVDA right now?" — opinion on the security, not advice for you.

Open your portfolio →

Related · Publisher posture, Watchlist, Shares, Cost basis

The publisher publishes opinion on securities, not advice tailored to you. The same verdict reaches every reader; no one's situation, holdings, or goals are inputs.

This product runs under the publisher exclusion in §202(a)(11)(D) of the Investment Advisers Act — the same posture courts upheld for Seeking Alpha in Lingley v. Seeking Alpha. The exclusion fits when three things are true: the content is impersonal, the same content goes to everyone who reads it, and the publisher doesn't render personalised advice to any subscriber.

Three rules drop out of that for the surfaces a reader sees. Verdicts are opinions on the security, never instructions to a person — no "you should buy." The AI chat is blind to your portfolio and your situation. Marketing language never claims to "predict" markets or describe itself as "AI-driven" investment advice.

The non-AI compliance guard at the end of every verdict and every chat reply enforces those rules in code, on every output, live or mock. Crossing them isn't a code change — it's a multi-month regulatory move into RIA registration.

Related · Compliance guard, Verdict, Local-only portfolio

A stock's short trading code on an exchange — NVDA for NVIDIA, AAPL for Apple. Unique per exchange.

A ticker symbol is the short identifier a stock trades under on a given exchange — typically one to five letters.

We use the U.S. exchange ticker throughout. The same company can have different tickers on different exchanges; if a company is dual-listed, the U.S. listing is the one we cover.

Browse the coverage list →

Related · Sector, Market cap

The broad industry bucket a company belongs to — Semiconductors, Banks, Energy, etc. Useful for comparing peers.

The last traded price per share. Updates intraday; ours is the most recent print at the time the page loaded.

How much the share price has moved since yesterday's close, in percent. Green = up, red = down.

See today's movers in the feed →

Related · Price, Sparkline, Today's P&L

When the company is expected to report its next quarter. The single most important forward catalyst — most of the chart's volatility lives in the days around it.

Public companies file a quarterly results release on a regular cadence — usually 3-6 weeks after each fiscal quarter ends. The release lands with audited numbers, fresh guidance, and a recorded call where management answers analyst questions. It is the single biggest scheduled event a stock has each quarter, and historically the days immediately around it carry several multiples of the average daily price move.

On every profile and rail row, the next-earnings chip shows two things: the calendar date (weekday + month + day), and the relative distance from today ("in 6 days", "tomorrow", "in ~6 weeks"). When the print is inside the next week the chip warms in colour so it catches the eye first. "~" in front of the date means the company has not yet formally announced — the date is a consensus estimate from the prior cadence and may shift by a few days.

"After market" prints land at 4 PM ET and play out the next trading session; "pre-market" prints land before the open and move the cash session itself. Either way, the read on the same data is very different two weeks before the print versus four days before — the chip is there so you never miss that context.

See NVDA's next-earnings cue →

Related · Price, Verdict, Spatial context

Share price × shares outstanding. The market's tag for the whole company's equity. Big = mega-cap; small = small-cap.

Market capitalisation is the share price multiplied by the number of shares outstanding. It is the market's current price tag for the company's entire equity stake.

Rough buckets: mega-cap ($200B+), large-cap ($10B–$200B), mid-cap ($2B–$10B), small-cap ($300M–$2B), micro-cap (below $300M). The bigger the cap, the more liquid the stock and the more analyst coverage it tends to have.

Sort the screener by market cap →

Related · Price, P/E ratio

The tiny line chart next to a price. It shows the recent price path — shape over precision.

See it in context on NVDA's profile →

Related · Rebased, Price

Each price series divided by its starting value so they all begin at the same point. Makes percent moves comparable across stocks with very different prices.

A $1 move on a $20 stock is a 5% move; the same $1 on a $400 stock is a quarter of a percent. Drawing both raw price lines on one axis would let the higher-priced stock dominate the chart even when it barely moved.

Rebasing fixes that. Every series is divided by its first close in the window — so each line starts at 100 — and the axis then reads as percent moved from the start. Lines above 100 are up, below 100 are down, and a flat line is unchanged.

Use it for relative direction across the window, not for trade timing — the underlying prices are still what you pay; the chart just removes the scale problem when several tickers share a frame.

Compare three tickers rebased →

Related · Sparkline, Price

Price ÷ earnings per share. How many dollars investors pay today for each dollar of yearly profit.

The price-to-earnings ratio divides the current share price by the company's earnings per share over the last year. A P/E of 20 means investors are paying $20 for every $1 of annual profit.

High P/Es typically signal that the market expects strong growth; low P/Es can signal a value play or a business in trouble. Always compare a P/E against peers in the same sector, not across sectors. P/E breaks entirely when EPS is negative — for unprofitable growth companies, reach for price-to-sales or EV/EBITDA instead.

Sort the screener by P/E →

Related · EPS (earnings per share), PEG ratio, Price-to-sales (P/S), EV/EBITDA, Market cap, Revenue YoY

Net income divided by diluted shares outstanding. The per-share slice of a year's profit.

Earnings per share is the company's net income divided by its diluted share count — the share count after every stock option, RSU, and convertible has been treated as already exercised. It is the denominator inside P/E and the headline number markets react to on earnings day.

GAAP EPS follows accounting rules to the letter; adjusted (non-GAAP) EPS strips out items management deems non-recurring — most often stock-based compensation. The gap between the two is where most accounting controversy in tech lives. Read both, and read alongside FCF margin for the cash check.

Related · P/E ratio, SBC intensity, FCF margin

Market cap divided by trailing annual revenue. The valuation ratio that still works when a company has no earnings to divide into.

P/S is market cap divided by revenue over the last twelve months — what investors are paying per dollar of sales. It is the valuation ratio of choice for growth-stage companies, where earnings can be negative or trivially small and P/E becomes meaningless.

Read it within a sub-sector, not across: software businesses routinely trade at 10–20× sales, hardware at 2–5×, banks below 3×. A 20× P/S on a SaaS business in a hyper-growth phase reads as expensive but not absurd; the same number on a hardware company is a statement.

Related · P/E ratio, EV/EBITDA, Revenue YoY

Enterprise value divided by EBITDA. The cleaner cross-company valuation ratio because it includes debt and ignores capital structure and tax differences.

Enterprise value (EV) is market cap plus debt minus cash — the price an acquirer would have to pay for the whole business, not just the equity. EBITDA is earnings before interest, tax, depreciation, and amortisation. Their ratio answers "how many years of pre-financing operating profit am I paying for the whole company?"

EV/EBITDA is preferred over P/E when companies in the same comparison have very different debt loads (a heavily indebted firm flatters P/E by inflating interest expense out of net income), different tax rates (multinational vs. domestic), or heavy depreciation (capital-intensive vs. asset-light). Most tech screens use it alongside P/E rather than instead of.

Related · P/E ratio, Price-to-sales (P/S), Market cap

P/E divided by expected earnings growth, in percent. A way to ask whether a high P/E is justified by how fast earnings are growing.

PEG divides the P/E ratio by the company's expected annual earnings growth rate in percent. Peter Lynch popularised the heuristic that a PEG of 1 is fairly priced — a P/E of 30 on a company growing at 30% — while a PEG under 1 is structurally cheap and over 2 is structurally expensive.

The heuristic breaks at the extremes. A company growing earnings at 80% won't grow at 80% forever, so a PEG of 0.5 on that base doesn't mean what it would on a 15% grower. Treat PEG as a screen, not a verdict — useful for sorting a peer list, not for picking a stock outright.

Related · P/E ratio, EPS (earnings per share), Revenue YoY

How much a stock tends to move when the market moves. Beta of 1 = market-like; 1.5 = 50% more volatile; 0.6 = much steadier.

Beta is the covariance of a stock's returns with the market's, divided by the market's variance — typically calculated over the trailing 60 months of monthly returns, with the S&P 500 as the market proxy. A beta of 1 means the stock has historically moved one-for-one with the market; 1.4 means it has moved 40% more on average; 0.6 means it has been much less volatile.

Beta is historical and mean-reverting — high-beta stocks tend to drift back down toward 1 as their business matures, and beta calculated through a regime change (a pandemic, a sector rotation) often overstates what the next year holds. It is also not a measure of direction; a beta of 1.4 says nothing about whether the stock is going up or down, only how much it amplifies whichever way the market goes.

Related · Volume, Today's change

How many shares of a stock traded hands over the day. Higher volume on a move means the move had real conviction behind it.

Volume is the simple count of shares traded in a session. Most large-caps print millions of shares a day; the more thinly traded a name, the easier it is for a single block to move the price.

Volume is a context number for the price action, not a directional one. A 5% move on triple the usual volume is structurally different from the same 5% on light volume — the first looks like a position being built or unwound, the second can be a quiet drift on no real flow. Earnings days, index-rebalance days, and option-expiration Fridays routinely produce volume spikes that are not about company news.

Related · Today's change, Beta, Float

The share count actually available for the public to trade — outstanding shares minus insider holdings, treasury stock, and other locked blocks.

Total shares outstanding includes every share the company has ever issued. The float is the subset of that pool that public market participants can actually buy and sell — outstanding minus insider holdings, employee restricted stock, treasury shares the company has repurchased, and any other locked-up tranches.

Low-float names move more on a given dollar of buying or selling pressure, because the dollars chase a smaller pool of available shares. Short interest is quoted as a percentage of float for the same reason — what matters for a squeeze is how borrowed shares stack up against tradeable supply, not against the full share count.

Related · Volume, Short interest, Stock buyback

The highest and lowest closing prices over the last year. A quick sanity check on where the current price sits in its recent history.

The 52-week high and 52-week low are the highest and lowest closing prices over the trailing year. Together they bracket the band the stock has actually traded inside recently.

A stock pressed against its 52-week high is in either an established uptrend or about to breach a resistance level traders are watching; a stock near its 52-week low is either oversold and reverting or about to make a new low. The range alone says nothing about which scenario is in play — it sets the context for whatever else you're reading on the page.

Related · Price, Today's change, Sparkline

A cash payment a company sends to shareholders, usually quarterly. Mature businesses pay them; most high-growth tech doesn't.

A dividend is a cash distribution paid out of company earnings, declared by the board on a per-share basis and paid quarterly in the U.S. by convention. Owning a share entitles you to the dividend if you owned it before the ex-dividend date.

Mature, slow-growth businesses — utilities, consumer staples, big banks, integrated energy, the older megacap tech names — return cash via dividends because they can't reinvest it all at decent rates of return. Hyper-growth tech largely doesn't pay them, on the same logic in reverse: a dollar reinvested in product compounds faster than the same dollar handed back to shareholders.

Related · Dividend yield, Ex-dividend date, Stock buyback

Annual dividend divided by current share price. The cash return rate a buyer locks in today, before any change in the stock price.

Yield is the company's trailing or forward annual dividend per share divided by the current share price, quoted as a percent. A 3% yield on a $100 stock paying $3 a year tells you what cash return you get from holding it for a year if neither the dividend nor the price changes.

Yield rises when the share price falls. A yield that suddenly screens unusually high relative to peers often signals a stock the market is pricing for a dividend cut, not a windfall — read the payout ratio (dividends as a share of earnings) before celebrating the number.

Related · Dividend, Ex-dividend date

The cutoff for the next dividend. Buy on or after this date and the previous owner gets the dividend, not you.

When a company declares a dividend, it names a record date (who's on the books gets paid) and an ex-dividend date (the trading day that determines who counts as on the books). Buy a share on or after the ex-date and the seller, not you, receives the upcoming payment.

On the ex-date, the share price typically drops by roughly the dividend amount in the opening print — the cash that's about to leave the company has effectively already left the share. Yield-chasers who buy the day before and sell the day after to "capture" the dividend usually end up flat after tax.

Related · Dividend, Dividend yield

A company repurchasing its own shares on the open market. Reduces the share count, which lifts per-share metrics like EPS.

A buyback is the company spending cash to retire its own shares. It is one of two ways to return cash to shareholders (the other being dividends), and the preferred channel for most tech megacaps because it is more tax-efficient (no per-share tax event for shareholders) and more flexible (the board can pause, accelerate, or stop a buyback without the market-signalling fallout of cutting a dividend).

Mechanically, a buyback shrinks the denominator in EPS — fewer shares for the same net income means each remaining share owns a bigger slice. That can flatter the per-share trend even when the underlying business is flat. The honest read is to compare a buyback's pace to the simultaneous pace of SBC dilution; many tech buybacks effectively just mop up the shares that were granted to employees.

Related · SBC intensity, EPS (earnings per share), Dividend

Slicing each existing share into multiple new shares. The economic stake doesn't change — only the per-share price and the share count do.

A 4-for-1 split turns every one share you own into four new shares, each worth a quarter of what the old share was worth. Your dollar exposure is identical; the company is identical; only the price and the share count have moved.

Splits are cosmetic. The historical rationale was to keep the per-share price in a range retail buyers found psychologically buyable, which mattered when round lots were a hundred shares and fractional trading didn't exist. With $0-commission fractional trading at every major broker, the rationale is largely vestigial — a split today is usually a signal about prior price appreciation, not about the business itself.

Related · Price, Shares

Revenue minus cost of goods, divided by revenue. The share of every sales dollar that survives the direct cost of producing the thing sold.

Year-over-year revenue change: this quarter's sales versus the same quarter a year ago.

Research-and-development spending as a percentage of revenue. The share of every sales dollar a company reinvests into its next product.

R&D intensity is annual R&D expense divided by revenue, expressed as a percent. In hardware-heavy or platform tech, single-digit R&D intensity at scale (Nvidia's ~9%) means the company is profiting on a moat already paid for. Double-digit or higher (Meta ~27%, Snowflake ~45%) means meaningful future product is still being built and the income statement is reflecting that.

Read the trend, not the level — R&D intensity climbing while revenue growth slows is a different story from intensity that compresses as the business scales. Compare within sub-sector: SaaS at 20% looks normal; a megacap consumer-internet company at 27% is unusual and worth understanding.

Related · FCF margin, Capex intensity, SBC intensity

Free cash flow divided by revenue. How many cents of cash the business actually keeps for every dollar of sales, after capex.

FCF margin is operating cash flow minus capital expenditure, divided by revenue. It is the cleanest one-line read of cash quality — a high-margin software business should print 25%+ FCF margins at scale; a capex-heavy cloud or chip business will run lower because the buildout absorbs the cash.

FCF margin tells you whether the GAAP income statement is hiding something the cash flow can't. Stock-based compensation flatters GAAP earnings but doesn't move cash, so a company with high SBC and low FCF margin is a different business from one where both lines agree.

Related · R&D intensity, Capex intensity, SBC intensity

Capital expenditure as a percentage of revenue. How much of every sales dollar the company is sinking into long-lived assets — data centres, fabs, factories.

Capex intensity is capital expenditure divided by revenue. For platform tech in the AI build cycle it's the single most-watched line: hyperscalers running capex intensity in the mid-20s%+ (vs. historical 12–15%) are betting that future revenue justifies today's spend.

High capex intensity is not inherently bad. The right way to read it is alongside the backlog (RPO) and the customer base — capex pulling forward to meet contracted demand reads very differently from capex pulled forward on hope.

Related · FCF margin, RPO (remaining performance obligations), Customer concentration

Stock-based compensation as a percentage of revenue. The non-cash cost that drives the gap between GAAP and adjusted earnings.

Stock-based compensation is the value of equity grants to employees, expensed on the income statement but not actually paid in cash. SBC intensity (SBC ÷ revenue) is the standard cross-company read.

SBC dilutes shareholders even if it doesn't move cash. A growing SBC line means a growing share count over time — the per-share earnings denominator gets bigger. Watching SBC intensity alongside FCF margin is the cleanest way to know whether 'adjusted' earnings are flattering the picture.

Related · FCF margin, R&D intensity

Contracted future revenue the company hasn't recognised yet. The forward look that backlog metrics give into cloud and enterprise-software trajectory.

RPO — remaining performance obligations — is contracted revenue the company has signed but not yet recognised. For hyperscalers and enterprise SaaS, RPO is a stronger near-term growth read than current revenue: it tells you what's locked in, not what's already booked.

Compare RPO growth to revenue growth. RPO meaningfully outrunning revenue means the next several quarters of acceleration are funded. RPO growth that lags revenue is the early warning of demand cooling.

Related · Capex intensity, Net revenue retention, Customer concentration

Dollar-weighted expansion in an existing customer base over a year, net of churn. The SaaS-quality metric.

Net revenue retention (NRR) is how much revenue a cohort of customers from a year ago is spending today, including upsells and downgrades, net of churn. 110% NRR means existing customers are spending 10% more than a year ago even after losses; 90% means the base is shrinking.

A SaaS business with consumption-based pricing (Snowflake, Datadog) prints elevated NRR when usage grows — that's the bull case in one number. Look at the trend: NRR collapsing from 160% to 120% over two years is not a disaster, but it is the end of a hyper-growth phase.

Related · RPO (remaining performance obligations), Customer concentration

How much of revenue comes from a few large customers. In tech, often phrased as 'hyperscaler exposure' for chip and infra suppliers.

Customer concentration measures how dependent a company is on a small number of buyers. Broadcom and Nvidia disclose top-customer concentration in the 30–45% range for their AI silicon businesses — meaningful, and the bull case carries a single-customer-deceleration risk that diversified businesses don't.

Concentration is not the same as fragility. A multi-year contract with a top-three hyperscaler is concentrated revenue, but it's also funded backlog. Read concentration alongside RPO and the customer's own capex disclosures.

Related · RPO (remaining performance obligations), Capex intensity

The editorial tech taxonomy this publisher uses — ai-infrastructure, semiconductors (ex-AI), cloud-platforms, consumer-internet, consumer-hardware, enterprise-saas, data-platforms, cybersecurity, devtools, fintech-and-payments.

The screener and the AI both read from a fixed sub-sector list rather than free-text 'sectors.' A reader who wants to study the AI inference stack picks ai-infrastructure + cloud-platforms; a reader studying SaaS quality picks enterprise-saas + data-platforms; a reader on payments rails picks fintech-and-payments.

Sub-sectors are deliberately narrower than the GICS categories most market data uses. The bet is that going narrow lets every surface — signals, peers, learn paths, AI — speak the same vocabulary. The taxonomy was reshaped on 2026-05-16 to put AI infrastructure on its own footing (every reference framework from Coatue to Roundhill CHAT now treats accelerated compute as a first-class object) and to broaden fintech-infra into the wider fintech-and-payments bucket used by FINX, ARKF, and Tiger Global.

Filter by sub-sector in the screener →

Related · Sector

A company's quarterly financial report filed with the SEC. Lighter than the 10-K, filed three times a year.

The 10-Q is the unaudited quarterly report every U.S. public company files for the first three quarters of its fiscal year. It contains the same kinds of financial statements as the annual 10-K but in less detail.

We diff each new 10-Q against the previous one and surface material changes — that's the 'lazy prices' signal, named for academic research showing markets are slow to react to subtle changes in repeated filings.

Related · 10-K, 8-K, Lazy prices

The audited annual report. The most thorough document a public company files — risk factors, financials, management discussion.

Related · 10-Q, 8-K, Lazy prices

An out-of-cycle disclosure filed when something material happens — an acquisition, a CEO change, a major contract win or loss.

Related · 10-Q, 10-K, Form 4

A filing required within two business days when a company insider — exec, director, big shareholder — buys or sells shares.

Form 4 is the SEC filing that company insiders use to report their personal trades. By rule, it must be filed within two business days of the trade.

Open-market insider buys are usually the strongest insider signal — insiders sell for many reasons (taxes, diversification, plan triggers) but typically only buy when they expect the price to rise.

See recent insider activity in the feed →

Related · Insider buy, 10b5-1 plan, 8-K

A pre-set sale schedule insiders adopt to sell stock on autopilot. Distinguishes routine sales from conviction-driven ones.

A 10b5-1 plan lets insiders pre-commit to a schedule of sales (or, less commonly, buys) at fixed prices or dates — set up while they don't have material non-public information.

When the plan triggers a sale weeks or months later, it doesn't signal anything new about the insider's view. We mark plan-driven sales separately from discretionary ones so neither gets mistaken for the other.

Related · Form 4, Insider buy

A company insider buying shares with their own money on the open market — historically one of the stronger bullish tells.

See recent insider activity in the feed →

Related · Form 4, 10b5-1 plan, Insider sell

An insider reducing their position. Much noisier than a buy — insiders sell for taxes, diversification, scheduled plans, divorce, college tuition.

An insider sale shows up on the same Form 4 a buy does, but the signal asymmetry is wide. People with non-public information have many reasons to sell that have nothing to do with their view of the stock — equity grants need to be exercised before they expire, tax bills come due, life happens. They have one reason to buy: they expect the stock to be worth more.

We mark plan-driven (10b5-1) sales separately from discretionary ones so neither gets misread. Even discretionary sales by a single insider carry less weight than a cluster of discretionary sales by multiple insiders inside a short window — the same pattern logic as buys, just on the other side.

Related · Form 4, 10b5-1 plan, Insider buy, Cluster buy

Three or more different insiders buying within roughly a 30-day window — the pattern carries more weight than any single buy on its own.

A single insider buy is informative; a cluster — multiple distinct insiders buying inside a short window — is much harder to explain away as diversification, plan triggers, or personal liquidity. The pattern has historically been one of the more durable retail-accessible alpha sources in academic and practitioner research.

We flag a cluster when at least three distinct insiders show open-market buys in the same rolling 30-day window. Plan-scheduled buys are excluded so the count reflects genuine conviction.

Related · Insider buy, Form 4, 10b5-1 plan

A required filing every member of Congress submits within ~30 days of a personal stock trade. Reports a member, chamber, buy/sell, and a dollar-amount band.

The Stop Trading on Congressional Knowledge (STOCK) Act of 2012 obliges members of the U.S. House and Senate — plus their senior staff — to publicly disclose personal stock trades within roughly 30 days. The disclosures name the member and chamber, the security, the trade direction, and the trade size in broad amount bands ($1K–$15K, $15K–$50K, and so on up to $50M+).

We read these as data, not as politics. There is no party labelling in our surface, no chamber-level partisan framing, and no attribution of motive. The reason to track them at all is that members sometimes vote on policy that materially affects the names they trade, and the disclosures are public record — a soft information edge at best, not a directional call.

Related · Form 4, Federal contract awards, Government activity

Cumulative U.S. government contract awards on record for a company — defense, intelligence, cloud-of-record, and IT-services line items combined.

Federal contract awards are a structural read on a company's recurring U.S. government revenue. A large book means the firm has cleared the moats that take years to replicate — security clearances for the workforce, FedRAMP authorisation for the platform, and a position on multi-year IDIQ (Indefinite Delivery/Indefinite Quantity) vehicles that future task orders flow through.

Among tech tickers, the largest federal books cluster around defense / intel-adjacent platforms (Palantir, the major semiconductors), the IT-services primes (IBM, Oracle), and the hyperscalers competing for cloud-of-record contracts (Microsoft, Amazon, Google). Most of the rest of the tech catalog has no federal exposure at all — that is unremarkable, not bearish.

Related · STOCK Act disclosure, Government activity

The share of a stock's float that's been sold short — borrowed and sold by traders betting the price falls.

Short interest measures how many shares have been borrowed and sold in the hope of buying them back lower. It is published every two weeks by the exchanges and updated more frequently by data vendors.

Rising short interest can mean traders see a problem the rest of the market hasn't priced in — or it can set up a short squeeze if the price rallies and shorts have to cover. Reading the number alone is not enough; the trend matters more than the level.

Related · Days to cover, Cost to borrow, Utilization, Short squeeze

How many trading days of average volume it would take shorts to buy back every borrowed share. Higher means a slower, more painful exit.

Days to cover divides short interest by the stock's average daily volume. A reading of 5 means it would take the entire short side roughly five normal trading days to close out — assuming they were the only buyers in the tape, which they aren't.

The number is useful as a relative read: 2–3 days is fluid, 6+ days starts to crowd the exit. It does not predict a squeeze; it tells you how tight the door is if one starts.

Related · Short interest, Short squeeze

The annualised fee a short seller pays the lender for the shares they've borrowed. Cheap on most names; expensive when supply gets tight.

Brokers charge a borrow fee on every shorted share, quoted as an annualised percentage of the position's value. Liquid large-caps cost a fraction of a percent; hard-to-borrow names can run into the double digits.

Rising cost-to-borrow alongside high utilization is the cleanest read of supply pressure on the short side — the market is making it more expensive to stay short. Falling cost reads the other way.

Related · Short interest, Utilization

The share of lendable inventory that is currently lent out. 90%+ means there is almost nothing left to borrow.

Utilization measures how much of the available stock lending pool is already in use by shorts. It is a real-time gauge of borrow scarcity — at 100% utilization, the next short can't open without finding someone new to lend.

High utilization on its own doesn't move the price. It does mean a fresh wave of buying — earnings beat, takeout rumour, sector rotation — can force forced covering at any price, because there are no replacement shares to short into the rally.

Related · Short interest, Cost to borrow

When a stock rallies on news or buying pressure and shorts are forced to cover, their buying accelerates the move into a feedback loop.

A short squeeze happens when a crowded short position has to close in a hurry — usually because the stock is rallying and lenders are recalling shares or borrow costs are spiking. The covering itself is buy pressure, so the move feeds on itself until shorts have flattened.

Crowded short interest, high utilization, rising cost-to-borrow, and a multi-day cover ratio together describe a stock where a squeeze is mechanically possible. Whether one actually triggers depends on a catalyst — there is no number that predicts the spark.

Related · Short interest, Days to cover, Cost to borrow, Utilization

Large, dollar-weighty options prints — calls or puts — that often signal where institutional money is positioning before the cash market reacts.

Options flow is the stream of unusually large options trades — typically tagged when premium spent on a single contract is large enough to suggest an institutional book, not a retail bet. A bullish call sweep can mean a fund is paying up for upside exposure; a put block can mean someone is hedging or pressing a short.

We read flow as a confirmation signal, never a primary driver. A single big print isn't a thesis — it can be a hedge against an existing book, a roll, or a misclassified close. The signal weights premium, so a $4M print counts more than five $250K prints, and tags the tape as bullish, bearish, mixed, or quiet rather than chasing single contracts.

Related · Options premium, Call / put ratio

The dollar cost paid for an options contract. The size of a print — premium times contracts — is how flow desks rank how serious the bet is.

Premium is the cash price of an options contract — the upfront cost of buying optionality. A trader paying $4 a contract on 1,000 contracts is laying down $400,000 in premium; on a $20-strike stock that's a much bigger statement than the underlying notional suggests.

Flow trackers sort prints by premium spent because it filters out small bets. A million-dollar premium print is a real position; a thousand-dollar print is noise. Our options-flow signal aggregates by premium for the same reason.

Related · Options flow, Call / put ratio

The premium spent on call contracts vs. put contracts over the same window. Above 1 = call-heavy tape; below 1 = put-heavy.

Calls bet the stock goes up; puts bet the stock goes down or hedge a long position. The dollar ratio between the two — premium spent on calls vs. puts — is the cleanest single read of which way the options tape is leaning.

A call-heavy ratio with bullish-tagged prints reads as conviction on the upside. A put-heavy ratio reads the other way — though it can also mean hedging into earnings, which is why we look at the tagged sentiment too, not just the call/put split.

Related · Options flow, Options premium

How positive or negative management sounds on an earnings call, scored from the transcript text. Tonal shifts often precede guidance changes.

Academic finding that markets are slow to react to subtle changes in filings. We diff each 10-Q/10-K against the last and surface the deltas.

Our internal blend of Quality, Momentum, and Value signals into a single 0–100 score. Used as a tiebreaker, not a verdict on its own.

QMV stands for Quality, Momentum, and Value — three signal families that decades of academic research have identified as drivers of long-term returns.

The composite scores each stock 0–100 on the blend. We use it as one input into the verdict pipeline alongside filings, insider activity, and transcript tone — never as a standalone call.

Aggregated and time-decayed signal from retail message boards. Noisy on its own, useful as a crowd-mood overlay.

Off-exchange venues where institutions trade large blocks without revealing the order to the public tape. Roughly a third of US equity volume.

A dark pool is an alternative trading system that matches buyers and sellers privately, then prints the trade to the consolidated tape after the fact. The point is to move a 500,000-share block without the order itself moving the price — visible orders on a public exchange tip off other traders to lean against the trade.

Dark-pool volume is real signal about institutional positioning, but it's lagged and ambiguous: the print confirms a large transaction happened without telling you who bought, who sold, or why. We read it as confirmation alongside options flow and Form 4 activity, never as a standalone driver.

Related · Options flow, Form 4

Composite of federal contract awards (recurring U.S. government revenue) and congressional STOCK Act disclosures. Low weight in the verdict — quiet but real.

Government activity blends two public-record reads on a ticker's political-economy exposure: cumulative federal contract awards (a structural read on recurring U.S. government revenue, treated as one-way positive) and dollar-weighted congressional STOCK Act disclosures over the recent window (treated as a soft directional lean).

The signal carries a low weight (~8%) in the composite — it is real signal but quieter and more lagged than fundamentals or insider buys. A large federal book nudges the score positive; absence of one is neutral. Congressional flow tilts the direction modestly when it concentrates in one direction. The publisher reads disclosures as data, not as politics: no party labelling, no motive attribution, no chamber-level partisan framing.

Related · STOCK Act disclosure, Federal contract awards

The number of units of a stock you hold. One share = one slice of ownership in the company.

What you paid per share, on average. Determines whether your position is up or down — and what you owe in tax when you sell.

Your cost basis is the average price you paid per share, including any commissions. It's the line you draw to compute profit or loss.

If you bought the same stock at different prices, most brokers report a weighted average. Cost basis matters for tax: a sale above cost is a capital gain; below it, a loss.

You enter cost basis by hand here — the app doesn't connect to your broker. The number you type lives in your browser only; see Local-only portfolio for why that's deliberate.

Open your portfolio →

Related · P&L, Shares, Allocation, Local-only portfolio

Profit and loss — what your positions are worth now versus what you paid for them. Green = gain, red = loss.

How much your tracked positions have moved in dollars and percent since yesterday's close.

What share of your tracked portfolio sits in a given position or sector. Concentration risk lives here.

Which positions actually drove today's gain or loss — a bigger position moving a little can outweigh a small position moving a lot.

Attribution answers the question, "what made my portfolio do what it did today?" It splits the day's net P&L into per-position contributions so you can see which tickers pulled you up and which dragged you down.

The size of a contribution depends on both how much the stock moved and how much of it you hold. A 0.5% move on your largest position often contributes more dollars than a 5% move on your smallest one — attribution makes that visible at a glance.

Open your portfolio →

Related · Today's P&L, Allocation

Tickers you're keeping an eye on without putting money into them yet. Promote any to a tracked position with one tap.

Open your portfolio →

Related · Shares, Cost basis, Local-only portfolio

The first AI stage — a fast, cheap model that classifies what you asked and extracts any tickers from your query.

Related · Verdict model, Debate model

The mid-tier model that turns the raw signal report into the structured Buy/Hold/Trim/Avoid verdict, with citations.

Related · Router model, Debate model, Citation

The top-tier model that adversarially reviews high-stakes verdicts and sharpens the strongest counter-argument that ships with every call.

Every verdict carries a strongest counter-argument — the single sharpest case AGAINST the call, plus why the evidence still landed there. The verdict-stage model writes the first draft of that counter on every run.

On verdicts where being wrong matters most — low confidence, or the label conflicts with the signal lean — a stronger model takes a second pass. It either affirms the call or revises it, and in either case re-writes the counter to the form a thoughtful skeptic would actually use. When that happens, the 'refined by debate' badge appears on the counter.

Routing the debate selectively keeps cost and latency in check. Publishing the counter on every verdict makes the publisher's internal disagreement visible regardless.

Related · Router model, Verdict model, Strongest counter-argument, Compliance guard

A non-AI rules layer every verdict and every chat reply passes through before reaching you. Catches advice-speak, guarantees, and portfolio references.

The compliance guard is plain TypeScript — no model, no probabilistic call. It runs as the last step of every verdict and every chat reply, live or mock. Three bright lines: it blocks second-person trade instructions ("you should buy"), it blocks portfolio / holdings / balance references (the chat can't talk about what you own because it doesn't know), and it blocks prediction or guarantee language.

When the guard catches something, it sanitises the text in place rather than failing the response — "you should buy" becomes "the publisher leans bullish", and the reader still gets an answer. The line that the guard enforces in code is the same line that lets the product ship as a publisher rather than register as an adviser; see Publisher posture.

Related · Verdict model, Publisher posture, Citation

When API keys aren't configured, every stage falls back to deterministic sample output so the UI still works. Marked with a `mock` tag on the verdict.

Every adapter and every model stage has a deterministic fallback. With no Anthropic key, the router, verdict, and debate stages each return canned output; with no FMP, Quiver, or transcript keys the data adapters return fixtures of the same shape. The compliance guard runs over mock output just like it runs over live output — the contract is stable across the boundary.

Mock mode is what makes the codebase walking-skeleton runnable with zero config and what keeps the smoke scripts deterministic in CI. A verdict produced in mock mode carries a `mock` tag so it can never be mistaken for a live editorial call in a cache or an export.