A private wealth office and trading desk

Your whole financial life, run to one risk level.

One desk to plan, invest, trade, track and file. Set how much a bad year may cost you. Every idea, rebalance, trade and alert is then measured against that number, and nothing reaches your book until it clears a research bar most ideas fail.

  • Plan
  • Portfolio
  • Wealth
  • Tax
  • Research lab
  • Strategy accounts
  • Ask
Your goal at a risk levelSample household
17.8%chance of $2,000,000 by Sep 2031
-$110,990a 1-in-20 bad year
16%volatility a year, the accepted level
$0$1M$2M$3M$4M202620272028202920302031goal $2,000,000
Ideas checked against the accepted 16%
  • TJXcuts risk
  • ANETfits
  • UBERfits
  • TSMbreaches
$768,412 book plus $2,500 a month. In the terminal the curve uses your own book, goal and measured Sharpe. A range of outcomes, never a promise.
342
studies the lab ran in week 42. None passed the bar, so none reached a book.weekly job, Oct 12 to Oct 16
0 of 4
well-known ways to follow smart money beat the market for real.lab runs 176 to 179, 2026-09-30
24 + IL
official IRS form PDFs filled, plus the Illinois resident return.IL-1040, IL Sch NR, IL Sch M, IL Sch ICR
0
orders the AI can place. You confirm each one with your authenticator code.house rule 1
The office

Everything you own, on one desk.

The brokerage book sits beside the house, the rental, the 529, the bills and the tax bill. Each figure names its source and date, and a stale one says so.

  • Net worth across accounts, property and private stakes, kept current from the FHFA index and statements.
  • Drop a 1099, a K-1 or a card statement. It reads the file, checks each figure against its page and builds the return for your CPA.
  • A written policy you sign once, and a scenario studio that replays 2008 and 2022 against your plan every night.
  • Every date with money on it, synced to your own calendar.
Sample household
ItemValue
Brokerage book, 3 accountsE*TRADE via SnapTrade and a Google Sheet$768,412
Home (primary)FHFA house price index, 2026-06-30$998,000
Rental condo, Unit 4BFHFA house price index, 2026-06-30$458,500
Savings (emergency fund)bank statement, 2026-09-30$50,000
529 plan (college)Plaid, read-only, 2026-10-02$41,000
HSA, investedFidelity HSA statement, 2026-09-30$28,500
Checkingbank statement, 2026-09-30$14,000

Seven of the household's items. Debts and the rest are on Wealth, Net worth.

The bar
0of 4 passed

Famous ways to follow smart money, tested out of sample.

Each ran against a 50/50 blend of the S&P 500 and Nasdaq-100 over the same sessions. Beating the market on paper is not proof.

What was tested

The strategies tested and their yearly returns against the blend
StrategyPer yearBlend
Copy Nancy Pelosi's buysrun 179, 955 sessions, t -2.043.2%27.9%
All Congress buys filed within 10 daysrun 177, 965 sessions, t -1.0718.7%26.5%
Concentrated managers' 13F best ideasrun 178, 1,013 sessions, t -0.0621.5%22.1%
Insider buys of $1M or morerun 176, 766 sessions, t 0.2724.4%22.8%

A t of 3.0 is needed to pass. Lab runs 176 to 179: congressional periodic transaction reports, SEC Form 4 and 13F filings, run on 2026-09-30.

Five checks every idea must clear

  1. t > 3Beats the market for realIts edge over the blend has to be three times its noise.
  2. 0.95Survives the searchDeflated Sharpe counts every variant ever tried, so luck from trying many things comes out.
  3. < 20%Not overfittedThe chance the best backtest is a fluke, measured directly.
  4. < 10%Few false alarmsAcross everything the lab has tested, false discoveries stay rare.
  5. as ofNo peeking aheadEach day uses only what was public that day. A planted leak fails the test.

See how we test with every threshold read from the code and the count of ideas the lab turned down.

The lab also tests itself

  • A noise marketEvery rule family runs on seeded pure-noise prices and must fail. One that passes there is fitting noise.
  • Random entriesExit rules re-run on random entry days. An exit that wins either way is not an edge.
  • Double costsA survivor is re-run at twice the trading cost, never less than 5 basis points.
  • A hostile reviewerA separate AI reads an imported strategy for lookahead, missing costs and survivor-only data. A lab run checks each finding.
  • AI earns its weightAn AI input counts only after its calls are graded and show measured skill.
The workbench

From a thesis to a strategy account, on the same bar.

Bring a strategy you read about or write your own. If it passes, give it dollars and let it trade on paper while the record builds.

  1. Bring the idea

    Type a thesis and edit it as a workflow, pick one the AI proposed, import a TradingView strategy or run your own Python in a sandbox.

    Ask, Workflows, Import
  2. The lab tests it

    A walk-forward out of sample, the five checks and the self-tests. Each run lists its trades, drawdown and results by market regime beside buy and hold.

    daily, hourly, 15-minute and 1-minute bars
  3. Give it an account

    Open a strategy account with a dollar amount. It trades each session, or the moment an event you pick happens: earnings, a filing, a CPI print, a price level.

    paper first, always
  4. Watch the record

    Each account reports daily against the house baseline, with its own survival window. A daily review flags one that drifts from its backtest.

    live only after the record clears the bar
What the lab can test
  • Stocks and ETFs
  • Option strategies on ORATS volatility
  • Long-short books
  • Coins
  • ML classifiers
  • Tick history
  • Prediction-market odds
Every automated account runs inside these limits
  • at most $25,000 an order
  • stops after a 5% loss in a day
  • never holds more than its dollars
  • kill switch and pause
  • every order in that day's notice
Ask

Say it in plain words. It drafts. You decide.

Ask reads every page and can draft any change the desk can make: a trade, a rule, a budget line, a scenario. Each arrives as a card that says how it fits your risk level. Nothing runs until you click, and a trade also needs your code.

Live trading unlocks after118 of 252 sessions

Sample. The gate also needs a walk-forward that beats SPY and QQQ net of AI spend.

AskSample conversation

Trim AVGO and put it into ANET, inside my risk level.

Two drafts. Both keep the book inside your 16% band.

SELL 30 AVGO, limit $341.50cuts risk

Volatility 17.8% to 17.6%. 1.33% of the book.

BUY 72 ANET, limit $141.20fits

Adds 1.3% of the book. The most that stays inside your band is 2.3%.

Authenticator code
Between visits

It keeps working when you close it.

  • A desk agent on your schedule

    It works the goal and lists what needs you: a rebalance, a harvest, an order to draft. It proposes. You click.

  • Your phone knows first

    Install StockGaines and get a push for a signal, a fill, an alert or the morning board.

  • One honest letter every Saturday

    What worked, what failed and what the lab found, with the market beside your book.

Nothing here moves money on its own. It watches, proposes and tells you, and you decide.

StockGaines, week 42Oct 12 to Oct 16

Your book held up better than the market.

Paper book
0.28%
Market blend
-3.76%
  • The lab ran 342 studies: none passed the bar.
  • The Oct 10 top 10: 6 of 10 beat the S&P 500.
The weekly Ideas block (ideas/weekly.json), written by the weekly_ideas job
House rules

Built like money is at stake, because it is.

  • The AI drafts. You decide.

    The AI has no tool that places an order. A chat trade is a card you confirm with a fresh authenticator code. A strategy account trades only inside the dollars and limits you set.

    10-minute expiry, one confirm per draft
  • Every number has a source.

    Each figure names where it came from and when, down to the lab run or the filing.

    lab run 179, 2026-09-30
  • Bank links read, never move.

    Plaid and broker data links are read-only. Your keys are sealed in your own vault.

    read-only
  • Every import can be undone.

    A document batch, a bulk edit or a rule change comes back out in one click.

    Undo this import
  • Tax is an estimate.

    The return is prepared for your CPA to review and file. There is no e-file.

    Estimates, not tax advice.
  • Research, not advice.

    StockGaines is not a registered adviser and promises no returns. Past results do not predict future ones.

    not a registered adviser
By invitation

Bring your goal. The desk will tell you the odds.

StockGaines is open by invitation while the record builds. Enter your code to set up your desk, or sign in.