The retail trader’s agentic hedge fund

Hedge funds had the edge.This one is yours.

You do not have a research team, a Bloomberg terminal, or six hours a day to stare at charts. So we built the desk instead: 75 signals scored on every candidate, four agents arguing it out, and ten setups waiting by morning — each with the price to buy, the price to bail, and two places to take profit.

Free while in beta · Applications reviewed weekly · No card required

75
Model features scored per candidate, every run
5/5
Bull and bear setups waiting each morning, 50+ scored behind them
4
Timeframes read together: weekly, daily, 4-hour, session
7
Regimes, for the index, the sector and the stock
Setup · published 4:41 PM ET · immutableDesk setup fields · sample values

NVDA

AGRADE
BULL · BREAKOUT
$121.062.24 (+1.89%)

NVIDIA Corp · Technology · XLK regime: strong bull · origin: Morning Top

Last$121.06
Day+2.24 +1.89%
Session+1.12
ATR(14)2.68% 0.3×
vs 8-EMA (D)+1.44%
Entry triggerPT1 / PT2StopCall wall

Price ladder

PT2$134.00
PT1$126.00
Last$121.06
Entry$118.40
Stop$114.90
R:R to PT1
2.17R
State
ACTIVE · day 2
Signal score
91
Short interest 1.8%Call wall $125Put wall $115IV 42.1%Earnings in 19dSweeps 3Congress 2 buysSqueeze Pro midFull research & agent debatePaper trade this setup

The pipeline, running

You get ten names. We looked at five thousand to find them.

Every night the universe is filtered down before anything is scored, scored before anything is gated, and gated before four agents are allowed to argue about what is left. Most of the work is deciding what not to show you.

0

Universe

US equities and ETFs clearing volume and liquidity minimums.

0

Liquidity & trend gate

Hard pass/fail filters. Thin names never reach the model.

0

75-feature scoring

Bull and bear models run separately. Composites and penalties applied.

0

Regime gate

Index, sector and stock regime resolved. Fighting the tape costs you.

0

Agent debate → published

Four agents argue. Five bull, five bear survive with full levels.

Stage counts are representative of a typical run, not a fixed quota.

This runs whether you are watching or not.

The work finishes before the open. You read four numbers and decide.

Scoring one candidate

Nobody can watch 75 signals at once. You don’t have to.

Signal groups · this candidatesample values
Trend & EMA92
Momentum88
Squeeze Pro74
Fibonacci81
Volume surge95
Relative strength90
Short squeeze63
Options flow86
Congress flow58
News sentiment71
Composites93
Risk math84
Penalties−12
Compositesample values

0

signal score

A10th decile of tonight’s run

Grades are relative to the night they were published, not to an absolute bar. On a thin night an A is the best of a weak field, and the report says so rather than pretending otherwise.

The agent war room

The bear case you would never think of, argued before you risk a dollar.

Bull case · technical analyst
Bear case · portfolio manager
Synthesis + risk officer
Verdict
Bull · take the plan
Confidence
High
Entry trigger
$118.40
Stop
$114.90
PT1 / PT2
$126.00 / $134.00
Planned R:R
2.17R
Invalidation
Close below 8-EMA
Gap risk
Earnings in 19d

The risk officer signs off last and only on loss: what gaps, what cannot be exited, and what would invalidate the plan. If it objects, the setup does not publish.

Written to be representative of a run. Not transcribed agent output.

Why R matters more than win rate

You can be wrong half the time and still finish ahead.

Ten setups · risking 1R eachillustrative
4 reach PT1 at +2.17R
+0.00R
1 runs to PT2 at +4.46R
+0.00R
2 close flat, no trigger
+0.00R
3 stop out at −1R
+0.00R
Net on the batch+0.00R

This is arithmetic, not a track record. It shows what a fixed R model does to a batch when six of ten trades do not reach the first target — nothing here claims these are our results.

Outcome distributionillustrative shape
34%PT2
78%PT1
46%Open
30%No trigger
60%Stopped

A published plan is immutable; what happens to it is appended as events. A setup can therefore reach PT1 and later stop out, and both facts stay in the record. These proportions are illustrative shape, not measured rates.

Paper trade it for a month before you risk a dollar.

Take every setup at its published levels and see what the model does to your account.

One setup, start to finish

You will always know where a call ended up — including the ones that failed.

NVDA · Feb 17 – Feb 26sample values
Event logappended, never rewritten
PlannedFeb 17 · 4:41 PM$116.80
ActiveFeb 18 · 10:14 AM · entry tagged$118.40
PT1Feb 21 · 1:07 PM · day 3$126.00
PT2Feb 26 · 11:52 AM · day 6$134.00
ClosedFull move captured · +4.46R

A published plan is never edited. Had price closed below $114.90 instead, the log would read Invalid at −1R and stay in the report next to the wins.

Why swing trading goes wrong

The problem was never the charts.

I don’t know where to get in.

One entry trigger, decided before the open and never moved. If price does not reach it, there is no trade — that is the plan working, not failing.

fixed entry

I hold too long and give it all back.

The stop and two targets are set at the same moment as the entry, before you are emotionally in the position, with alerts on all three.

stop · PT1 · PT2

Every drawdown I switch strategies.

One scoring system, run the same way every night. Setups carry an archetype and are graded on that day’s deciles, so a weak night looks weak instead of looking normal.

one system

I can’t tell a real breakout from a fakeout.

That is what the 75 features, the regime gate and four arguing agents are for. Most candidates die at one of those three, and you never see them.

75 features

It’s too sophisticated. I won’t know how to use it.

You read one line: buy above X, stop at Y, targets at Z. Everything else is there for when you want to know why, not because you need it to place the trade.

four numbers

I have a job. The move happens while I’m in a meeting.

The work finishes before the open, so the plan is waiting when you wake up, and near-real-time notifications tell you when price tags a level.

alerts

Market regime

The same breakout that paid last month is the one that traps you this month.

Every candidate is scored against the regime of the index, its sector and itself. A clean setup in the wrong regime is a worse trade than a scrappy one in the right regime, and the gate treats it that way.

Regime · 24 weekssample values
SPY
QQQ
XLK
XLE
XLF
IBIT
24 weeks agothis week
Strong bullMild bullLow vol driftNeutral transitionChoppy noiseMild bearCrash bear
Sector rotation · this weeksample values
XLK+2.8%
SMH+3.6%
XLC+1.9%
XLY+1.2%
XLF+0.8%
XLI+0.4%
XLV−0.3%
XLP−0.6%
XLU+0.9%
XLRE−1.4%
XLE−2.2%
XLB−0.9%
GLD+1.6%
SLV−3.1%
USO−1.8%
IBIT−4.2%
ARKK−1.1%
IWM+0.2%

Tonight’s report

Filter it down to the way you actually trade.

Nightly report · Feb 19, 2026

Sample rows — replace with a real published dispatch before this page goes live.

#Symbol20dSideGrdScoreArchetypeLastEntryStopPT1PT2R:RState
1NVDABullA91Breakout$121.06$118.40$114.90$126.00$134.002.17RActive
2ARMBullA-87Momentum$103.44$101.20$97.40$110.00$118.502.32RActive
3ONBullB+82Breakout$81.90$81.10$77.00$88.50$94.201.80RPlanned
4CLSBullB76Swing$64.18$63.40$60.20$69.80$74.502.00RPT1
5AEHRBullC+68Reversal$22.31$22.00$20.40$25.10$27.601.94RPlanned

Showing 5 of 5 rows in this report.

This is the board, one night, filtered five ways.

The live one updates nightly, with alerts when price tags a level.

The work is finished before the open, because you have a job.

Weekends · weekly bars

The higher-timeframe view

  • structure_weekly · mom_weekly
  • adx_14w · roc_12w
  • aggregate indicators: signal + strength
  • regime read per index and sector

Every night · daily bars

Tomorrow’s trade plans

  • structure_daily · structure_h4
  • vwap_session · session ATR
  • full 75-feature evaluation per candidate
  • five bull, five bear, 50+ scored behind

The evaluated set

Every feature, named. Count them yourself — we would rather you checked.

Model features · scored per candidate75 features · separate bull and bear models

Trend & EMA structure9

price_vs_ema8_pctprice_vs_ema21_pctprice_vs_ema50_pctprice_vs_ema200_pctema_stacked_bullema_stacked_bearstructure_dailystructure_weeklystructure_h4

Where price sits against every EMA, and whether the ribbon agrees across three timeframes.

Momentum7

rsi_14roc_14roc_12wadx_14wmom_weeklychange_1d_pctsqp_mom_dir

Rate of change read daily and weekly, with trend strength, so a fast move is not mistaken for a trend.

Volatility compression5

sqp_barssqp_momentumsqp_probabilitysqp_firedadr_1d_pct

How long the squeeze has been building, and whether it has actually fired.

Fibonacci3

fib_ratiofib_bounce_readyfib_zone

Retracement depth and whether price is sitting in a zone that has held before.

Volume3

vol_surgesv_ratioshort_volume_ratio

Whether the move is carrying participation or is a thin drift.

Relative strength2

rs_reldaily_doc_spy_change

Performance against the index on the same day, not in isolation.

Short squeeze fuel6

short_interestdays_to_coversqueeze_probabilitysqueeze_scoresi_days_to_coversi_short_interest

How much stock is sold short and how long it would take to cover.

Congressional flow7

congress_buys_180dcongress_sells_180dcongress_net_180dcong_buys_180cong_sells_180cong_net_180cong_buy_pressure_180

Disclosed congressional trades over six months, netted and pressure-weighted.

News & sentiment5

news_countnews_scorenews_buzznews_bull_pctnews_bear_pct

Coverage volume and its direction, so a catalyst is separable from noise.

Composite buckets13

bull_comp_trendbull_comp_proximitybull_comp_momentumbull_comp_rsbull_comp_volumebull_comp_patternsbull_comp_penaltiesbear_comp_trendbear_comp_proximitybear_comp_momentumbear_comp_crestbear_comp_patternsbear_comp_penalties

Bull and bear models run separately, each with its own penalties. Nothing is netted.

Risk math7

dist_entry_pctdist_stop_pctdist_tgt1_pctdist_tgt2_pctrr_ratio_plannedrisk_scoreda_final_score

Distance to every level, and the planned reward against the risk before anything is published.

Gates & filters5

option_liquidity_passbull_filter_passbear_filter_passbull_filter_pass_v3bear_filter_pass_v3

Hard pass/fail. A name that fails here never reaches the model.

Price & session3

pricebiasvwap_relation

Where the last print sits relative to the session, and which way the book is leaning.

Total75

Add the counts above. They come to 75, and they are read straight from the model’s own feature list.

Agent context · assembled per symbolretrieved, not scored

Key levels

Pivots, prior swing highs and lows, and the levels the plan is built on.

Regime gate

The resolved regime for the index, the sector and the stock itself.

Golden sweeps

Confirmed sweeps with premium, expiry and direction.

Fundamental ratios

The handful that matter for a multi-day hold, not a full model.

Aggregate indicators

The combined signal and its strength, as the desk reads it.

Squeeze Pro state

Compression stage in words, so the agent can reason about it.

Patterns

Named chart patterns detected on the daily and four-hour.

Events & sentiment

Earnings dates, headlines, and anything that dates the trade.

Confirmed-entry history

How this name has behaved after prior triggers.

These are two different code paths. Agent context is written into the prompt, not scored, so it is not part of the 75.

How a setup gets made

Four agents, in order — the research team you cannot afford to hire.

Stage 01

Technical analyst

Commits to a direction for the next one to five sessions, and names the level that proves it right and the one that proves it wrong.

directionbreakout_levelbreakdown_levelrisks

Stage 02

Portfolio manager

Argues the other side for as many rounds as the run allows. Its job is to find the reason this does not work.

level_realisminvalidationsstructure_notes

Stage 03

Synthesis

Resolves the disagreement into the plan you see, and states how confident it is rather than hiding the uncertainty.

summarytriggersconfidence

Stage 04

Risk officer

The last pass is only about loss: what gaps, what cannot be exited, and what would invalidate the whole thing.

gap_risksliquidity_hazardsinvalidations

The rest of the desk

Everything you would otherwise be paying four subscriptions for.

Full charting

Candles, EMAs, levels and your own drawings, on the same chart the setup was built from.

Near-real-time alerts

Entry tagged, target hit, stop breached — pushed when it happens, not at the close.

Grounded copilot

Ask any setup why. It answers from that setup’s own research, not from the open internet.

Market structure

Index and sector regime, breadth, and where the tape is relative to its own history.

Sector rotation & heatmap

What is being bought and what is being left, at a glance.

Golden sweeps scanner

Confirmed sweeps with premium, expiry and direction, filterable.

Options advisor & scout

Contract selection for a plan you already have, rather than a trade idea generator.

Paper portfolio

Take the setup at its published levels and track it without risking anything.

Related companies

The names that move with this one, so a sector call is not accidentally six of the same bet.

Congress, insiders, filings

Disclosed trades and filings attached to the symbol, dated.

Research cohorts

Morning, ETF and full-universe runs, kept separate so they can be judged separately.

Report archive

Every past dispatch, including the calls that failed.

All of it, one login, free through the beta.

Charting, alerts, the copilot, the paper portfolio and the archive.

Straight answers

You have scrolled past a hundred posts claiming this. The difference is you can go read tonight’s.

What this actually is

  • Ten fresh setups a night, plus the fifty-odd scored behind them.
  • 75 model features, a nightly daily-bar run and a weekend weekly-bar run.
  • Losses in the same table as the wins, with the same fields.
  • The whole argument, including the case against the trade.
  • Alerts on entry, on both targets, and on the stop.

What it is not

  • Not a signals group, a Discord, or a guru with a P&L screenshot.
  • Not automated trading. Nothing is placed on your behalf, ever.
  • Not a black box — if you cannot see why, that is a bug.
  • Not a backtest dressed up as a track record.
  • Not a claim that every setup works. It is an edge, not a guarantee.

Latest writing

Read the work before you trust the work.

All insights →

How to Trade Violent Market Selloffs: Lessons from Today's Sharp Drop

Markets can drop violently on geopolitical news. Learn practical swing trading strategies for high volatility days including risk management, entries, psychology, sector rotation, and how to position for the rebound.

Read →

What is a Breakout in Trading?

Learn what a breakout is in trading, how to identify valid breakouts vs fakeouts, and the entry strategies professional swing traders use to capture explosive moves.

Read →

How to Read a Stock Chart for Beginners

Learn how to read a stock chart like a pro. This beginner's guide covers candlesticks, volume, moving averages, and the key patterns every swing trader needs to recognize.

Read →

Beta cohorts · applications reviewed weekly

Don’t take our word for it. Take tonight’s ten.

We are opening the desk a cohort at a time. Tell us how you trade, and we will tell you when your seat is ready — no card, no commitment, free through the beta.

How access works

  1. Request access. Thirty seconds. Name, email, and how you trade.
  2. We review weekly. Applications are read on Mondays and seats go out in order.
  3. You get the desk. Nightly reports, the live board, alerts, charting, the copilot and a paper portfolio.
  4. You tell us what breaks. Beta members steer the roadmap and keep preferred pricing afterwards.

Not ready to apply? The insights are free and already public — start there.

Request access

We will only email you about your application and the insights you asked for. Nothing here is financial advice.