Moods AI
The macro research desk for self-directed investors
Built and operated end to end

Moods Alpha OS

It watches primary sources, writes a trade plan, paper-trades only when the rules pass, and records what worked. The examples below show the actual plays, including the losses.

Closed paper trades
5353 executable decisions
Net paper result
-$206.72-0.2% of starting capital
Benchmark alpha
-0.4%Aggregate result versus each trade's benchmark
Live capital at risk
00 paper positions in this snapshot
Actual decisions

What the system traded, in plain English

The complete closed ledger now mirrors the operator book. Use the featured view for full explanations or open all closed trades for the synchronized record. These are paper outcomes, not broker fills or a claim of live performance.

Paper, not liveAudit OK53/53 ledger sync
Closed paper case study

LULU Jul17'26 120/140 Call Spread

LONG CALL SPREAD / catalyst 2026-06-25

WIN
Legacy reconstructedHistorical reconstruction frozen / SHA-256 7ead174f2718; not ex-ante proof.Legacy review derived
Realized P&L$616.00
Trade return+89.3%
Book impact+0.6%
Held8.8d

2026 annual meeting / contested proxy vote

The play

LULU proxy fight creates a dated upside convexity setup into the June 25 vote.

Why then

The company filed definitive proxy materials and the June 25, 2026 vote date is now the hard catalyst. Same-day price action remains inside the no-chase window, and options exist through the vote.

Entry rule

Enter while LULU < $127 and spread debit <= $7.60, preferring limit near $6.90 mid; same session or next regular session only if still inside no-chase and debit limits

Maximum loss

$7.60 per spread (debit paid); shadow thesis stop on underlying close below $112 or 50% spread premium loss

Exit plan

Max loss: $7.60 per spread (debit paid); shadow thesis stop on underlying close below $112 or 50% spread premium loss. Time stop: 2026-06-26. Profit take: Underlying $132-$138, spread +80-100%, or settlement/vote-support pop before annual meeting. Take profit if LULU trades $132-$138 before the vote or the spread gains roughly 80-100%. Close by June 26 if the vote/settlement path does not produce follow-through. Exit immediately if the source read changes from governance catalyst to reputational noise.

What happened

The written profit target was reached and the position closed automatically.

What the system learned

A dated catalyst paired with a defined-risk option expression produced asymmetric upside. This trade is evidence, not strategy validation.

Selection policy: Up to four closed, executable paper decisions: best convex winner, best linear winner, worst loss, and latest close. Duplicates are removed and the newest eligible trades fill empty slots.

Honest read: early evidence, not a validated strategy.

53 closed paper marks produced -$206.72 (-0.2% of starting capital) and -0.4% aggregate benchmark alpha. The convex winner is useful evidence, but the sample is not yet large or repeatable enough to support a strong performance claim.

Evidence standard

How much should you trust the result?

The automation is operating, but repeatable strategy performance is not yet validated. Historical paper results remain visible, while only new decisions captured before entry can promote an algorithm.

Collecting evidence
Prospective paper closes17/30Required for strategy promotion
Resolved probability forecasts0/20COLD START
Canonical diagnostics60One fixed test per captured decision
Lane champions00 look-ahead violations
What learning can do

Reorder research effort after enough prospective evidence.

What learning cannot do

Rewrite facts, change sizing or risk limits, count diagnostics as official P&L, or enable live trading.

How it works

Six auditable steps from source to result

Agents gather and rank evidence. Deterministic gates control identity, spend, risk, execution eligibility, and attribution, so a model cannot silently invent a position or rewrite a result.

01

Source intake

2 new / 0 research / 21 of 21 healthy

Official releases, SEC filings, agency material, and monitored sources enter an immutable raw ledger first. Generic filings and weak company-wire notices stay context-only; only deterministic hard events can enter research.

02

Event identity

13 canonical events

Deterministic semantic identity collapses duplicate alerts and tickets into canonical events with evidence and competing hypotheses.

03

Agent queue

2 jobs queued

Lease-safe Postgres jobs rank the next information task. Free workers are allowlisted; model work cannot run when its spend contract is false.

04

Risk gate

0 canonical decisions awaiting review

Duplicate source tickets do not consume portfolio capacity. No paper or live entry is allowed without position sizing, max loss, stop logic, event time stop, and liquidity context.

05

Paper execution

53 executable closes

Approved ideas move to a $100k ledger using whole shares or contracts, explicit costs, current marks, and the same no-chase contract used at entry.

06

Attribution loop

23 economic wins

Every closed paper mark stores P&L, alpha versus benchmark, process result, failure bucket, and the exact exit reason.

InterfaceReact + Vite
AgentsNode.js workers
StateJSONL + Supabase
QueuePostgres leases
RuntimeHetzner + systemd
ControlsRLS + spend gates
Live system replay

What the agents are doing now

The public view is deliberately anonymous: it shows event counts, queue contracts, worker state, and capital gates without publishing active research.

01
Source ledger3,953 alerts1,001 ticket rows projected
02
Canonical events1316 actionable hypotheses
03
Event router329 completed routes0 deferred / 0 failed
04
Agent queue2BUILD DOSSIER, BUILD DOSSIER
05
Capital gate0 paper0 live trades
Projectionalpha_event_graph_v2SUCCEEDED / Sep 15, 08:45 PM UTC
Queue contractBUILD DOSSIERPriority 85 / estimated cost $0.00
Free workerAllowlistedClaims market-refresh jobs only
Paper ledger

An attributed paper ledger

Realistic paper ledger starts with $100,000, sizes from current paper equity, rounds equities to whole shares and options to whole contracts, and counts skipped minimum-size trades as $0 P&L.

-$206.72 realized+35.7% decision hit rate
Starting capital
$100,000Realistic paper ledger base
Ending capital
$99,793-0.2% portfolio return
Closed marks
5353 executable / 42 paper decisions
Economic win rate
+43.4%23 positive outcomes across the paper ledger

Equity curve

-0.4% alpha
Start: 100.00Index: 99.79
Open technical appendixSimulation controls, convex funnel, risk gates, lane economics, aggregate attribution, and recent automation runs.
Simulation lab

Official P&L and missed-opportunity learning stay separate

Diagnostic orders and finalist observations are excluded from official paper P&L.

0 virtual orders0 diagnostic open
Working limits
00 observed natural-debit fills
Finalist observations
150 pending / 15 closed
Diagnostic contribution
-0.6%Standardized 1% sleeves; excluded from official P&L
Blocked winners
311 undirected observations carry no P&L

Execution controls

Official ledger isolated
Fill rule

Observed natural debit at or below the written limit

Midpoint fills

Prohibited

Unknown direction

Magnitude tracked; signed P&L excluded

Official positions

Unmodified by diagnostics

Decisions captured
2,6173,371 historical snapshots are clearly excluded from learning
Rule variants
27,6483%, 5%, and 7% entry bands plus pullbacks across six holding horizons
Prospective outcomes
6,4948,677 open / 2,598 waiting for an entry
Look-ahead violations
0Quarantined automatically and never allowed into strategy promotion
Convex focus

From dated event to bounded convexity

Research priority, trade quality, and execution readiness are scored separately. A high-priority event is not presented as a trade until its scenario, expected value, direction, dossier, and execution gates clear.

3 screened0 research finalists0 qualified
Events screened
33 watch / 0 rejected
Research finalists
00 still in quality review / 0 non-positive EV
Qualified setups
00 paper ready / 0 deadline decisions
Prospective evidence
17/300 prospective convex closes; legacy results are not validation
Why separate this sleeve

Convexity is measured independently

Free deterministic screening expands the dated-event funnel first. Scenario review and observed option-chain evidence remain mandatory before any paper position can open.

Scenario review

0 finalist(s)

Need first chain

0

Observed chains

0

Observed structures

0

Limit watch

0

Executable structures

0

Paper ready

0

Clean median hold

8.8d

Risk management

Controls that stop bad trades

0 paper positions are open and 0 live trades are authorized. Missing risk fields, spend limits, and service isolation remain visible.

No hidden live risk

0 live trades and 0 open positions in the published dashboard.

Defined max loss required

No ticket can enter paper until sizing, max loss, entry logic, and a time stop are explicit.

Paid API budget visible

614 paid-call ledger entries MTD; heartbeat flags none.

Legacy API firebreak

Old public-site paid APIs are paused while the alpha lab owns paid research calls.

OpenClaw isolated

OpenClaw is inactive and disabled; this page is the Moods Alpha paper process, not live crypto execution.

Post-trade accountability

53 closed paper marks with benchmark alpha, process labels, and failure buckets retained in the private ledger.

Lane governance

Opportunity sourcing scorecard

Each lane must earn its cost. The latest scorecard keeps active lanes visible and shows which monitors are capped because they produced too many no-trade or low-yield outputs.

12,812 runs / 24h
LaneActionTicketsUsefulPaid callsNo-tradeTop reject reason
CATALYST CONVERGENCEInsufficient recent evidence for promotion or removal.WATCH17000%-
EARNINGS MISREADInsufficient recent evidence for promotion or removal.WATCH16473 / useful0%-
REGULATORY CALENDARCold-start evidence budget: 15/16 runs and 15/24 paid calls used.WATCH130150%-
DILUTION SHOCKInsufficient recent evidence for promotion or removal.WATCH4000%-
SOURCE REACTION PROBERealized sample has not cleared minimum P&L, alpha, profit-factor, and sample gates.WATCH3300%-
FEDERAL AWARD INFLECTIONInsufficient recent evidence for promotion or removal.WATCH111 / useful0%-
MERGER CVR ARBInsufficient recent evidence for promotion or removal.WATCH1000%-
EARNINGS VOL EXPLORATIONCold-start evidence budget: 8/10 runs and 40/100 paid calls used.WATCH00400%-
EARNINGS ANTICIPATIONCold-start evidence budget: 7/10 runs and 21/70 paid calls used.WATCH00210%-
Accountability

Outcome accountability

The public snapshot exposes aggregate decision quality and benchmark alpha while individual positions, sources, and exit records remain private.

Closed marks5353 executable decisions
Economic wins23+43.4% economic win rate
Decision hit rate+35.7%42 paper decisions closed
Benchmark alpha-0.4%-$206.72 aggregate paper P&L
Automation

Recent operating loop

Hetzner timers: 4 configured, failed services: 0.

0 failed servicesold APIs paused
Sep 15, 08:45 PM UTCanaphora:autonomous-loopOK
Sep 15, 08:45 PM UTCalpha:heartbeatOK
Sep 15, 08:45 PM UTCalpha:decision-resolverOK
Sep 15, 08:45 PM UTCalpha:event-storeSUCCEEDED
Sep 15, 08:45 PM UTCalpha:autonomyOK
Sep 15, 08:45 PM UTCalpha:decision-resolverOK