QBO-A02 · Quobo Labs fleet

Quobo Labs Stocks

Labs applicationOpen beta

Tech-stock research that shows its work: standing, cited Buy/Hold/Trim/Avoid verdicts, report cards that publish the losers, and a portfolio the server can't read.

Join the beta →

Chapter one

The problem

A retail investor who cares about tech stocks lives in a split world. The brokerage app shows a chart and a news wire and explains nothing. Real research terminals are priced for professionals. In between sits a feed economy of confident calls. The moment of pain is daily and specific: a stock you own is down 4% at lunch, your feed is a firehose of hot takes, and there is no fast way to answer — what actually changed, does it matter, and what does someone with a considered, evidence-cited view think?

The stranger part is the accountability gap. Whoever called it loudest yesterday is never graded today. Winners get victory laps; misses quietly disappear. Nobody keeps a scoreboard, so nobody has to be right.

And underneath, a quieter pain: every app that offers to know your portfolio wants your brokerage login and your balances on its servers. A lot of people just want to model what they hold, privately, and have the research come to them — without handing an app their financial life.

The mechanics

How it works

A ranked feed of what's actually moving
Price moves, SEC filings, insider buys, short-interest spikes, congressional trades, and anonymized chatter from a curated set of working technologists — ranked by recency, magnitude, and how many independent signals agree. Names you hold float to the top, with extra urgency on bearish news; that boost happens in your browser, because that's where your holdings live. A synthesized morning read names the day's story — and declines to render on quiet days rather than invent one.
The feed — today's read up top, kind and sentiment filters, movers ranked alongside
A ticker profile built from primary sources
Every covered name gets a deep page where each metric carries a rank chip against its sector peers, so “is a 62 P/E good?” answers itself. Below that, auditable evidence sections: what changed on the earnings call between quarters, the insider Form 4 history, short interest and cost to borrow, congressional trades and federal contracts, options flow, and the year-over-year change in the company's own filing language.
A standing verdict, with its history attached
Each profile ends in a published house view — Buy, Hold, Trim, or Avoid — with a confidence score, bull, base, and bear cases, and citations back to the evidence sections above it. One model writes the verdict; a second re-litigates it adversarially whenever confidence is low or the call disagrees with the quantitative signals. A trajectory strip shows how the label has moved over time, so a changed mind is visible, not memory-holed. It reads like a publication with an opinion, not a broker with a suggestion.
A verdict — label, confidence, bull/base/bear cases, and the strongest counter-argument, refined by debate
Report cards that publish the losers
Every signal in the stack gets a public, falsifiable report card: the basket of stocks it flagged, measured against a market benchmark at 30, 60, and 90 days. Signals that trail the benchmark are listed right alongside the ones that beat it. That is the point — research willing to be measured.
A portfolio the server structurally cannot read
Model your holdings by hand, or import them from a broker screenshot — the OCR runs in your browser, and the image never leaves the page. Holdings live locally and sync end-to-end encrypted, so the server stores opaque bytes it cannot decrypt. No brokerage logins, no balances on anyone's servers — and the research still gets personal: your names rank first, and verdict changes on them are flagged, downgrades before upgrades.

Field notes

Would this fit your life?

A backend engineer with RSUs and a red morning

The moment: A name she holds is off 4% by lunch, and the group chat has three contradictory explanations.
What they do: The feed has already floated the ticker to the top. On its profile she skims the since-you-last-viewed strip, sees which evidence sections are firing, and checks the verdict — still Hold, confidence off a few points, the full label history visible. One question to the assistant, already pinned to the ticker, walks the bear case with citations.
The payoff: From red number and rumors to what changed, and whether the house view moved, in about three minutes — inside an app that can't place a trade.

A self-taught dip buyer chasing down a rumor

The moment: Social media says insiders are loading up on a name he watches. He wants to check before he acts.
What they do: He opens the ticker's insider section and reads the actual Form 4 record — cluster buys versus routine scheduled sales — cross-checks the congressional-trading section, then pulls up that signal's public report card to see how insider-buying flags have really performed, sample sizes included.
The payoff: A rumor becomes a checkable primary-source record plus a measured base rate. The anti-hype workflow.

A privacy-conscious designer with two brokerages

The moment: She wants her real positions reflected in the app — and refuses on principle to connect a brokerage to anything.
What they do: She screenshots her broker's positions table. The import runs OCR in her browser — the image never leaves the page — she corrects one misread row, and her holdings sync end-to-end encrypted. The feed immediately re-ranks with her names on top.
The payoff: Full personal relevance — ranking, alerts, P&L — with zero broker credentials and holdings the operator provably can't read.

A skeptic burned by the last hype cycle

The moment: A friend recommends “an AI stock app.” He arrives looking for the catch.
What they do: He reads the disclaimer, then goes straight to the report cards: every signal's record against the benchmark, trailing signals printed with the leaders. He opens a verdict and finds the models named, the change history shown, and a one-tap account deletion in settings.
The payoff: Nothing promises outperformance; everything invites audit. That turns out to be the pitch.

No varnish

From the lab notebook

Where it stands

A functional beta, free and email-gated. The whole surface — feed, ticker profiles, verdicts, the assistant, screener and compare, the encrypted portfolio with in-browser screenshot import, a ninety-six-term glossary with guided learning paths, and the public report cards — runs end-to-end today. The AI pipeline, the compliance guards that sanitize every output, and the feed ranking are real, tested code with CI gates behind them.

What's still rough

Some evidence sections — earnings-call deltas, options prints — run on curated deterministic fixtures until the data vendors are wired in, and most deployments run in that fixture mode. Sign-in is an interim email-keyed session with no verification yet; the migration to proper magic-link auth is planned and written up. Coverage is a hand-built catalog of about twenty-two tech names, not a full-market database.

What's next

The technologist-voice feed — chatter from a curated set of people who build the technology, aggregated anonymously — is fully built but sits behind a flag until it has a month of live history and written counsel review; publishing opinions about stocks is a business where you check with the lawyers first. After that: live data wiring vendor by vendor, and app-store submission for the native shell. A paid tier is on the roadmap, but nothing charges today.

Next in the fleet

QBO-A03

Quobo Labs Schedule

In the lab

An AI chief of staff that interviews you: you talk, it files what you say into a self-correcting memory that writes your to-dos, plans its own next meetings, and briefs any fresh AI on your world.

Continue the tour →