Video summary

I Tested Letting Claude Trade For A Month and Made $102k

Main summary

Key takeaways

Finance

Performance / account setup

  • Challenge account: Robinhood
  • Starting capital (start of May): ~$66,000
  • Ending capital (end of month): ~$169,000
  • Gain: ~$102,000 (stated) / ~155% return over the past 30 days
  • Claim: “Every single trade plus adjustment” was made by Claude, including daily monitoring and changes (human kept “inside the loop”).

Explicit disclaimer / caution

  • Not financial advice — no guarantee of profits.
  • Caution: do not copy the exact trades/setup, since entries/prices were time-specific to the beginning of May.
  • The key takeaway is to understand the methods, not the specific execution.

Trading strategy framework used in May

Overall approach: two-pronged

  1. LEAP options (long-dated calls)

    • Rationale: leveraged upside with a more durable holding window; avoids short-term “perfect timing” risk.
    • Time-decay control: monitored Greeks, especially theta.
    • Contract constraint: LEAPs with >45 days left to expiration to reduce time-decay pressure.

  2. Small-cap equity shares (not options)

    • Rationale:
      • Smaller stocks can be illiquid, making it harder to enter/exit options at fair prices.
      • Options may be too expensive due to high IV.
    • Leverage use: primarily through LEAPs on larger names with more reasonable option pricing.

Research → execution workflow (how Claude was used)

  • Claude was used first as a non-deterministic analyst (qualitative reasoning; not presented as backtestable).
  • Then a daily monitoring system was built using:
    • Deterministic components: fixed rules / system logic
    • Non-deterministic Claude: daily news/analysis-driven judgments

Input “context” requirement (important methodology)

Claude’s outputs improved when the user provided:

  • account size
  • risk tolerance
  • holding periods throughout the month

Option / risk construction details (explicit contract examples)

  • Claude selected specific strike prices and expirations (not just “buy calls”).

  • Example LEAP contracts (both out of the money):

    • Nokia: expiration Jan 15, 2027, $20 strike
    • ServiceNow: expiration Jan 15, 2027, $120 strike

  • Portfolio balancing logic:

    • Consider strike selection, option cost vs IV, and how far out expiry must be to support the thesis.
    • Offset time-decay risk on aggressive OTM positions by balancing exposures across other names.

Tickers / instruments mentioned

Options strategy candidates / positions (mentioned in screen)

  • OSCR
  • NBIS
  • ServiceNow (shown as “ServiceNow”)
  • HIMS
  • MP (described as “government-backed”)
  • Nokia

Also referenced via monitoring / examples

  • “micro shares” (shares alongside options; no additional symbols beyond those above)

Macro / risk inputs mentioned (used in monitoring)

  • VIX (“macro gate” reference)
  • market breadth
  • credit spreads

Data sources mentioned for building the dashboard

  • Yahoo Finance: used as a free source; computes Greeks locally because Yahoo doesn’t provide them
  • Optional alternatives: FMP, Polygon

Score / screen logic and portfolio construction rules

Claude’s screening criteria (as described):

  • Companies had sold off but had a real reason to recover, such as:
    • “beaten down” with an actual catalyst
    • re-rating “on the calendar”
  • Allowed both:
    • oversold recovery setups
    • momentum / “riding high” setups (if upside was justified)

Options specifically required:

  • reasonably priced options (accounting for the IV environment)

Claude scored candidates across:

  • catalyst timing
  • IV environment
  • correlation to the broad market

Diversification method:

  • Combine names with different catalyst windows and different sectors
    • Example time-horizon layering:
      • OSCR: near-term catalyst window
      • NBIS: medium-term window
      • ServiceNow: long-term LEAP window
    • Example sector layering:
      • healthcare vs AI infrastructure vs enterprise exposure
  • Nokia was treated differently: included a probability distribution across price levels and expected value (details not quantified in the subtitles).

Name-specific notes given:

  • ServiceNow: described as “clean” after crashed post-earnings; calls it “cheap,” enabling favorable LEAP pricing.
  • MP: government-backed; catalyst exists but described as slow, so Claude recommended not a concentrated bet (smaller position size vs others).
  • Nokia: built a probability distribution and EV across price levels (details not quantified in the subtitles).

Daily monitoring / portfolio dashboard methodology (4-layer system)

The video describes building an “options monitoring system” run once per day.

Stated cost / timeline

  • Runs daily; stated cost: about $1/day (because Claude is queried once per day).
  • Daily data snapshots are taken to detect new strikes/expiries versus the prior day’s chain.

Framework: “four layers” (step-by-step)

Layer 1 — Data & valuation

  • Pull spot price and full options chain
  • Match held contracts to the chain
  • Compute Greeks locally (Yahoo doesn’t provide Greeks)
  • Save daily snapshots to detect new strikes/expiries
  • Output: P&L per position + distance to targets

Layer 2 — Portfolio analytics

  • Evaluate the book holistically:
    • allocation across names/sectors
    • total Greeks
    • estimated daily time-decay (theta bleed)
    • IV environment per position

Layer 3 — Market context & news

  • Deterministic macro score using:
    • VIX, market breadth, credit spreads
  • Claude news + context pass:
    • reads news for each held name
    • summarizes sentiment + technical indicators
    • determines what affects the position
  • Run frequency: once per day to manage token cost

Layer 4 — Alerts & dashboards

  • Detect changes vs prior chain:
    • new strikes
    • new expiries
  • Flag events:
    • when targets or stops are hit
    • large IV moves
  • Produce a one-screen daily dashboard with outputs and suggested actions
  • Optional push alerts: Telegram / iMessage

System execution detail

  • Hybrid setup:
    • deterministic code handles most monitoring/valuation logic
    • non-deterministic Claude runs daily for news/context and decision support
  • Inputs to the system are the portfolio positions (after initial trade entry onto Robinhood).

Presenters / sources

  • Presenter: Brendan (host of the channel; studied math/econ at UCLA; worked in investment banking at Raymond James for 3 years)
  • AI source used: Claude (Anthropic)

Original video