Trading With Claude: The Final Recap

--

30 Sessions. $25 to Start. Every Trade on the Record.

Why This Experiment Existed

On March 9, 2026, I deposited $25 into a Robinhood account and started a question I couldn’t answer any other way: can an AI be a genuine trading partner?

Not a chatbot you ask for stock tips. Not a research tool you consult once a week. A real partner — one that sits with you through the pre-market scan, reasons through the macro thesis, reads the charts, weighs the risk, and helps you execute. Every session documented publicly. Every win and every loss on the record.

The rules were simple. Real money. Full transparency. Thirty sessions. No survivorship bias, no cherry-picked screenshots. If I lost, I wrote about losing.

What followed was two months of trading through one of the most volatile geopolitical environments in recent memory, building skills I didn’t have when I started, and watching the single biggest regulatory constraint on retail trading get eliminated on Day 22 of a 30-day experiment.

This is the full account.

The Macro Backdrop: Two Months Inside the Iran War

The experiment began nine days after Operation Epic Fury launched on February 28, 2026. I didn’t plan to trade into an active military conflict. It just happened that way, and it shaped every thesis, every sector choice, and every risk calculation that followed.

The core event was the partial closure of the Strait of Hormuz to Western commercial shipping. The Strait is not an abstraction. Roughly 20 percent of the world’s seaborne oil passes through that 21-mile-wide channel. When it closes — even partially — the price of crude does not wait for analysts to weigh in. It moves.

Then came the Ras Laffan news. Qatar’s flagship LNG terminal, the largest concentration of liquefied natural gas production in the world, was destroyed in strikes early in the conflict. QatarEnergy’s own CEO put the repair timeline at three to five years. One facility. A multi-year hole in global LNG supply. The downstream effects hit natural gas prices, then fertilizer markets (because natural gas is the feedstock for ammonia, which is the feedstock for nitrogen fertilizer), then freight rates, then shipping equities. Every sector we touched over the next six weeks traced back in some way to those two events.

The ceasefire announcement came on Day 17, April 8. It arrived mid-session, unexpected, and it immediately inverted the trade. Energy was the sell. Defense was the question mark. Airlines, consumer discretionary, and semiconductors became the rotation targets overnight. We tracked that rotation thesis in real time, which meant several of our later positions were explicitly post-ceasefire bets rather than conflict plays.

What the Iran war taught me about macro investing is that geopolitical events create sector moves that are real but brief. The first wave of buying in oil, defense, and LNG happens within hours of the headline. By the time the thesis is obvious to everyone, the easy money is gone. The second-order moves — fertilizer, shipping, airlines, metals — take days to develop and are where patient traders can still find edge. The third-order moves, rearmament cycles, infrastructure rebuild, regional energy diversification, play out over quarters and are outside the scope of daily trading entirely.

Going forward, the way I will read a geopolitical shock is in sequence: energy and defense first (hours), agriculture and industrial inputs second (days), airlines and consumer recovery last (weeks). The leading indicators I will watch are Hormuz transit data from Lloyd’s and maritime tracking services, LNG spot prices on the JKM benchmark, North American natural gas futures, and the Baltic Dry Index for freight. When those start moving, the equity sectors follow.

The Sector Research Map

Energy: Oil

The experiment inherited a legacy USO position from before Day 1, which we liquidated early to free up capital. The active oil trade was UCO, the ProShares Ultra DJ-AIG Crude Oil ETF, a 2x leveraged instrument that amplified daily crude price moves.

The key thing we learned about oil ETFs is that they do not actually hold oil. They hold futures contracts, and futures have a cost called contango. When the market expects higher prices in the future (the normal condition), the front-month contract is cheaper than the next month’s. Every time an ETF rolls its futures forward, it sells the cheaper contract and buys the more expensive one. Over time, this erodes returns even if the spot price stays flat. Leveraged oil ETFs like UCO compound this problem. They are trading vehicles, not long-term holds.

The successful UCO trade on Day 14 followed our partial exit discipline: sell the majority at Target 1, let the remainder ride to Target 2. We bought at approximately $39.42 across two tranches, sold the first tranche at around $39.82, and let the second run to $40.55. Total P&L was roughly +$1.96. Not a large dollar figure, but a proof of concept for the exit strategy.

WTI versus Brent: we tracked both benchmarks throughout. WTI (West Texas Intermediate) is the U.S. benchmark, priced at Cushing, Oklahoma. Brent is the global seaborne benchmark, priced in Europe. During the Hormuz closure, Brent moved more aggressively than WTI because the disruption was to seaborne trade routes, not U.S. domestic production. The spread widened. When Brent is trading at a large premium to WTI, it often signals that the geopolitical pressure is specifically on global trade flows, not on supply fundamentals.

Going forward, I watch Brent as the geopolitical thermometer and WTI as the domestic production signal. When they diverge sharply, there is usually a trade somewhere in the spread.

LNG and Natural Gas

Cheniere Energy (ticker: LNG) was the pure-play on the Ras Laffan disruption. Cheniere operates the Sabine Pass and Corpus Christi liquefaction terminals on the U.S. Gulf Coast. When Qatar’s supply went offline, U.S. LNG export capacity became significantly more valuable overnight. Cheniere was the clearest beneficiary with existing infrastructure, signed long-term contracts, and the ability to capture spot market premiums on uncommitted volumes.

The Day 16 LNG trade was the most technically complex of the experiment. We entered at approximately $284.17 for 2 shares (market order, bypassing the first-30-minutes rule, which we paid for with a worse fill). We sold one share at $286.71 intraday and held the second. The next morning, Day 17, we sold the second at $288.90. Net P&L across both legs was approximately +$9.44, the largest single-trade gain of the experiment.

But Day 16 also produced our most important lesson. We placed a stop limit order on the position. A stop limit has two prices: a trigger and a limit. When price hits the trigger, it converts to a limit order at your specified price. The problem is that if price drops through both the trigger and the limit in a single fast move, the order sits unfilled while the stock keeps falling. That is exactly what happened. The stock gapped down through our parameters and the stop never executed.

From that day forward, all single-stock positions use stop loss orders only. A stop loss converts to a market order at the trigger. You do not control the exit price, but you are guaranteed an exit. On a fast-moving stock in a volatile macro environment, that guarantee is worth more than the marginal price improvement a stop limit might offer.

LNG as a sector is harder to trade than crude oil because supply contracts are long-term, pricing transmission is slow, and the facilities themselves are massive pieces of infrastructure that take years to build. The equity moves in LNG companies tend to be driven by contract news, capacity announcements, and spot market premiums rather than the day-to-day price of natural gas.

Agriculture and Fertilizers

The CF Industries and Mosaic thesis was the one I tracked longest before entering: fourteen days of research before the first order. The connection from Hormuz to fertilizer is not obvious at first, but it is real. Natural gas is the primary feedstock for ammonia synthesis via the Haber-Bosch process. Ammonia is the precursor for almost all nitrogen-based fertilizers. When natural gas prices spike, nitrogen fertilizer production costs spike, and fertilizer company margins and pricing power improve.

CF Industries (CF) is the largest nitrogen fertilizer producer in North America. Mosaic (MOS) covers the potash and phosphate side, less directly linked to natural gas but still a beneficiary of broader agricultural commodity stress.

We entered CF on Day 15 at $132.38 for 2 shares and sold at $135.36 on the same session for approximately +$2.96. We also held a position into Day 16 before closing. On paper, the thesis was correct. In execution, we got burned by the gap-and-trap.

The gap-and-trap is the pattern I will remember longest from this experiment. The morning of Day 15 was chaos. Every agriculture and energy ticker we were watching opened with a large gap up. CF, MOS, UCO, SLV all spiked in the first minutes of trading on momentum and overnight news. We entered several positions into that spike. By mid-morning, every one of them had reversed and given back most or all of the gap. The gap was a trap, not a breakout.

The rule that came out of that day: never enter a position in the first thirty minutes on a gap-up open. Wait. Watch the opening range establish. Let the trap spring on someone else. If the thesis is real, the stock will still be at an attractive price at 10:00 a.m. CT. If the only good entry was the first five minutes of trading, that was not your entry anyway.

Going forward, I will monitor JKM LNG spot prices and Henry Hub natural gas futures as leading indicators for the fertilizer trade. When those move, CF and MOS follow within days.

Defense

ITA, the iShares U.S. Aerospace and Defense ETF, was entered on Day 21 at $234.24 as a deliberate multi-week hold thesis. The reasoning: defense spending responds to conflict, but with a lag. Contracts are not signed the day after a ceasefire. Rearmament cycles, restocking of munitions, and procurement authorization from Congress all take time. The underlying spending that conflict catalyzes flows through defense company earnings quarters after the headlines fade.

The ITA position was stopped out on April 21 at $225.93, for a loss of approximately $8.31 on one share. The stop at $226.00 was placed intentionally below a support level we had identified. Price breached it and the stop executed cleanly.

ITA currently trades around $215. The stop saved capital. The thesis may still play out over a longer horizon, but within the experiment’s time frame, the position did not work.

The broader defense thesis remains intact as a long-duration idea. The post-ceasefire question we kept coming back to is whether defense retreats after conflict ends. Historically, the answer is no: wars reveal capability gaps that take years and billions to close. The equities may pull back from conflict highs, but the underlying spending cycle has usually just started.

Metals and Mining

GDX (VanEck Gold Miners ETF) and GLD (SPDR Gold Shares) represent two different ways to hold the gold thesis. GLD moves with the gold price, roughly one-to-one. GDX moves with gold but amplifies the move because gold miners have operating leverage: their costs are largely fixed, so a $100 increase in the gold price is mostly pure margin improvement. GDX also carries additional risk from company-specific factors: operational issues, hedging programs, management quality.

We held GDX across multiple sessions. The Day 4–5 position was small and slightly negative. The Day 16 entry at $99.27 for 3 shares was part of a larger multi-position day. A Day 17 add at $275.51 for 2 shares (that price reflects a different GDX-related instrument or a data anomaly from the order history, which I am noting transparently) was stopped out at $267.77.

SLV (iShares Silver Trust) entered on Day 17 at $69.29 for 4 shares and was stopped at $67.53 the same day, for a loss of approximately $7.04. The silver thesis was a dual play: precious metal hedge plus industrial input demand. Silver is used in solar panels, electronics, and EV components at scale. In a world where conflict disrupts both safe-haven sentiment and industrial supply chains, silver has a legitimate claim to upward pressure from two directions simultaneously. The position did not work in the session window we had.

The most recent GDX trades in the order history, April 22, show a buy at $41.28 for 10 shares and a close at $42.06, for a gain of approximately +$7.80. GDX had a significant move this week as gold continued its run above $3,000.

Going forward, I watch the DXY (dollar index) and real Treasury yields as the primary inputs to the gold/metals trade. Gold moves inverse to real yields and inverse to dollar strength. When the 10-year real yield is falling and the dollar is weakening, the gold trade has structural support. Miners outperform gold itself in that environment.

Semiconductors

SOXX (iShares Semiconductor ETF) and SOXL (Direxion Daily Semiconductor Bull 3x ETF) were recurring themes across the experiment: Days 1, 2, 3, 8, 12, and the final sessions.

The SOXL trades on Days 1 and 2 were the opening positions of the experiment, entered as a volatility trade rather than a macro thesis. Both were profitable: +$1.44 and +$0.28. Day 3 added JETS as a diversification, which turned slightly negative. The SOXX breakeven on Day 8 and the -$0.48 loss on Day 12 reflected an environment where the semiconductor thesis was unclear during peak conflict intensity.

The reason semis were complicated throughout the war is that the sector sits at the intersection of two opposing pressures. On the supply side, a Middle East conflict can disrupt specialty chemical supply chains that flow through the region. On the demand side, an AI infrastructure buildout was running simultaneously and independently of geopolitical conditions. The two forces pulled in different directions, making it hard to hold a clean directional thesis.

The clearest semiconductor moment of the experiment came during the Nvidia GTC keynote session. Jensen Huang announced the Blackwell Ultra architecture and outlined Nvidia’s next-generation data center roadmap in detail. The market’s reaction in the hours after the announcement was instructive: NVDA itself barely moved, because much of the roadmap was expected, but the suppliers and second-tier beneficiaries moved more sharply. The lesson from that session was that keynote events matter more for the ecosystem than for the headliner. By the time Nvidia announces something at GTC, it is largely priced in for Nvidia. The trade is in the companies building the racks, the networking, the power infrastructure.

The final SOXX trade in the experiment, Days 22–23, was a limit buy at $440.89 for 2 shares on April 23, closed the same day at $441.45 via stop, essentially breakeven at +$1.12. By that point SOXX had become a vehicle for practicing the entry and exit mechanics as much as a directional bet.

Going forward, the semiconductor sector trade is cleaner outside of conflict periods. The primary inputs I watch are TSMC monthly revenue data (released the 10th of each month, the most reliable leading indicator for global chip demand), memory pricing from DRAMeXchange, and Nvidia’s data center segment revenue as a proxy for AI capex health.

Airlines

JETS entered the experiment on Day 3 as an early inverse-oil bet. The thesis: oil prices spike, airline operating costs spike, airline stocks fall. Buy the inverse. The problem was timing. Oil was still in the middle of its conflict-driven spike when we entered JETS. We were early, which in trading is the same as being wrong. The position closed with a small loss of $0.19.

The ceasefire on Day 17 made JETS the obvious rotation trade in theory. We researched it in detail but did not re-enter in the sessions that followed. In retrospect, that was the right call given how much volatility remained in the market and how uncertain the ceasefire’s durability appeared in real time.

The airline trade in a post-conflict rotation follows a simple logic: lower oil prices reduce the single largest operating cost for carriers. Load factors, which had stayed resilient even during the conflict on domestic routes, could now flow through to improved margins. International routes, which had been partially disrupted, would reopen. The recovery in JETS as an ETF lags the crude move by a few weeks as the market waits for evidence in earnings.

Utilities

XLU (SPDR Utilities Select Sector ETF) was a Day 13 trade, entered at $130.46 and closed at $133.74, for approximately +$0.31 net. Utilities are the flight-to-safety sector: high dividend yield, low beta, regulated revenue. When VIX spikes and traders want to be in equities but not in risk, utilities absorb capital.

The XLU trade worked because we entered it precisely when the conflict-driven VIX spike was near its peak. As volatility compressed slightly, defensive sectors like utilities held their value while riskier sectors gyrated. It was one of the cleaner thesis-to-execution sequences of the experiment.

Micro-Cap AI Pivots: A Cautionary Story

No recap of this period would be complete without the Allbirds story. Around Day 22 of the experiment, Allbirds — a shoe company with a stock price in the low single digits and a market cap of roughly $21 million — announced it was rebranding as NewBird AI and pivoting to artificial intelligence. The stock surged between 700 and 876 percent on the news.

This is a pattern with a long history. In 2017, Long Island Iced Tea Corp changed its name to Long Island Blockchain Corporation and its stock tripled overnight. The company had no blockchain business. It had a beverage business. The name change was a mechanism to capture retail momentum from a hot theme, issue equity into the spike, and dilute existing shareholders.

The NewBird AI rebrand came with a $50 million convertible note. A convertible note is debt that converts to equity, almost always at a discount to the market price at the time of conversion. For a company with a $21 million market cap, a $50 million convertible note is not a capital raise. It is a mechanism to massively dilute current shareholders in favor of the noteholder, who gets shares at a discount and can sell immediately.

The comparables we identified during this session were XWIN, AGPU, and HIVE: all micro-cap companies that had ridden AI or blockchain pivots to brief but violent price spikes before returning to obscurity or collapsing entirely.

The contrast we used was SMCI (Super Micro Computer) at approximately $24, which had experienced genuine distress related to accounting review and auditor issues, but was a real company with real AI infrastructure revenue, actual server contracts, and fundamental reasons to recover. The difference between SMCI and NewBird AI is the difference between a distressed real business and a name-change gimmick wearing the clothes of a hot theme.

The rule: if a company’s entire investment thesis is a rebrand and a press release, the move is already done by the time you see it. The 700 percent gain happened in the hours before the story was widely covered. By the time it is in your news feed, the float has turned over and the insiders are selling into your buy order.

Reading Candlestick Charts: Everything We Learned

Before this experiment, I could read a chart in the casual sense that most retail investors can: price went up, price went down, that line is the moving average. Over thirty sessions of daily chart work, that changed. Here is the full framework we built.

The Anatomy of a Candle

Each candlestick represents a defined time period and contains four data points: the open, the high, the low, and the close.

The body of the candle is the rectangle between the open and close. A green (or white) candle means price closed above where it opened: buyers won the period. A red (or black) candle means price closed below where it opened: sellers won. The size of the body tells you how decisive the move was. A large body means strong conviction. A small body means indecision or a close battle between buyers and sellers.

The wicks (also called shadows) are the thin lines extending above and below the body. The upper wick shows the highest price reached during the period before price retreated. The lower wick shows the lowest price reached before price recovered. A long upper wick on a green candle means buyers pushed price up aggressively but sellers knocked it back — bullish momentum was rejected at the highs. A long lower wick means sellers pushed price down but buyers stepped in and recovered — bearish momentum was rejected at the lows.

A candle with a very long wick and a very small body is called a doji (more on that below). It signals maximum indecision: price went somewhere and came back, and neither side won.

Timeframes

We used two timeframes throughout the experiment.

The 1-minute chart was our opening-read tool. The first fifteen minutes of trading produce a dense series of 1-minute candles that establish the day’s early character. A sequence of large-bodied candles with no wicks at the open means a fast, directional move with conviction. A sequence of small-bodied candles with long wicks means chop — the market is fighting with itself and no clean direction has established.

The 5-minute chart was our primary trading timeframe. Each candle represents five minutes of price action. This is enough time for each candle to be meaningful (capturing genuine buying or selling waves rather than microsecond noise) while still being fast enough to catch intraday moves before they are over. For the kind of trades we were making, entry and exit decisions made on the 5-minute chart were usually correct in terms of timing even when the underlying thesis was wrong.

The tradeoff between timeframes is noise versus signal. Shorter timeframes show more moves, more patterns, more signals — most of which are false. Longer timeframes smooth out the noise but by the time a pattern completes on a daily or weekly chart, the intraday opportunity is gone. The 5-minute chart sits at a useful point in that tradeoff for active session trading.

Key Patterns

The Gap-and-Trap

This was our most expensive lesson, experienced directly on Day 15 with CF, MOS, UCO, and SLV all in the same morning.

A gap-up open occurs when a stock opens significantly above its prior close, usually on overnight news. In the first minutes of trading, retail momentum buyers chase the move, pushing price even higher. Candles are large and green. Volume spikes. It looks like a breakout.

Then it stops. Volume tapers. The large green candles give way to small-bodied indecision candles. One red candle appears, then another. The stock reverses, closing the gap and sometimes falling below the prior close entirely. The gap was a trap. The buyers who chased the open are now holding losing positions.

The pattern is identifiable in real time by watching what happens to volume after the first 1–3 candles. If volume is declining after the initial spike, the move is losing fuel. The momentum buyers are done. The sellers are waiting.

Rule: on a gap-up open, watch the first thirty minutes without entering. If the stock consolidates and holds above the gap level on declining volume, then breaks to new highs with fresh volume, that is a real breakout. If volume tapers and the candles start having long upper wicks, the trap is being set.

The Opening Range

The high and low established in the first fifteen to thirty minutes of trading become anchor levels for the day. The opening range high becomes resistance: a price level where sellers have already shown up once. The opening range low becomes support: a price level where buyers have already stepped in.

A move that breaks clearly above the opening range high on strong volume is a legitimate breakout entry signal. A move that breaks below the opening range low on strong volume is a breakdown signal. Moves that stay within the opening range are noise, not signal.

We used the opening range framework on multiple sessions to set our intraday stop levels. If we entered a long position, the opening range low was often the stop. If price broke below the level buyers had already defended, the thesis was wrong.

Consolidation and Breakout

After a strong move, price often enters a consolidation phase. On the 5-minute chart this appears as a sequence of candles with small bodies and small wicks, clustered in a tight price range. The stock is coiling. Volume typically declines during consolidation.

A breakout from consolidation occurs when a single candle breaks cleanly above the top of the range on expanding volume. This is the entry signal we looked for on multiple sessions: wait for the consolidation, enter on the breakout candle, set a stop just below the consolidation range.

The failure mode is the false breakout: price briefly exceeds the consolidation range, draws in buyers, then reverses back inside. This is another version of the trap pattern. The confirmation we always looked for was whether the breakout candle closed above the range (not just touched above it) and whether volume on the breakout exceeded the volume during consolidation.

Support and Resistance

Price levels where price has previously reversed become significant for future trading. If a stock bounced at $98 three times over the past week, that $98 level is support. Buyers have shown up there repeatedly. When price approaches $98 again, the question is whether those buyers return or whether the level breaks.

On the 5-minute chart, support and resistance levels show up as horizontal zones where candle bodies cluster. The bodies are the real prices where trades occurred in volume. Wicks touching a level and retreating reinforce it. Bodies closing through a level break it.

A broken support level often becomes resistance: the same buyers who defended the level, now holding losing positions, will sell when price returns to their entry point. We used this principle on several exits, looking for price to approach prior support-turned-resistance as a signal to close rather than hold.

Volume as Confirmation

Every significant candle should be confirmed by volume. This was one of the most repeated principles across our sessions.

A large green candle on below-average volume is a weak move. There are not enough participants behind it to sustain the direction. Price may continue but is likely to stall or reverse quickly. A large green candle on above-average volume means genuine buying interest: institutions, not just retail, are participating.

Volume spikes at the open and close are normal market structure and not particularly meaningful. A volume spike during mid-session, away from the natural peaks, is significant. It means something changed: news, a large order, a technical level break that triggered cascading stops.

We watched volume on every chart screenshot during the experiment. A position entered on a low-volume candle made us nervous. A position entered on a high-volume breakout candle felt confirmed.

Doji

A doji occurs when the open and close are nearly equal, producing a candle that looks like a cross or a plus sign. The body is tiny or absent. The wicks extend both above and below, sometimes significantly.

A doji signals that buyers and sellers are perfectly balanced for that period. Neither side won. In isolation it means nothing. In context, it can mean everything.

A doji appearing after a sustained uptrend signals exhaustion: the buyers who drove price up are running out of momentum, and sellers are appearing in equal force. The next candle is the confirmation. If it is a strong red candle, the reversal is real. If it is a strong green candle, the uptrend resumes. The doji itself is the warning that the next candle deserves close attention.

Hammer and Shooting Star

The hammer is a candle with a small body near the top of the total range and a long lower wick. It appears after a downtrend. The interpretation: sellers pushed price down sharply during the period, but buyers came in aggressively and drove it back up to close near the open. The lower wick is the buyers’ counterattack. A hammer at a known support level is a high-probability long setup if confirmed by the next candle.

The shooting star is the mirror image: small body near the bottom of the range and a long upper wick, appearing after an uptrend. Buyers pushed price up aggressively but sellers knocked it back down to close near the open. The upper wick is the sellers’ rejection. A shooting star at a known resistance level is a high-probability short signal if confirmed.

We identified shooting star formations on several of the gap-and-trap mornings before the reversals completed. In retrospect, they were clear signals to exit immediately rather than hold through the pullback.

Reading the Open

The first candle of the session is the most information-dense single data point of the day. Its size, direction, and wick structure tell you what the overnight participants think and what the early institutional positioning is.

A large-bodied first candle with minimal wicks in either direction says: there is conviction. The market opened and moved without hesitation. Whatever direction that first candle points is probably the session’s directional bias.

A small-bodied first candle with long wicks in both directions says: there is confusion. The market opened, tested both directions, and found no clear resolution. That session is likely to be choppy and difficult to trade directionally.

The gap-up first candle with a long upper wick that closes near the open says: the gap was rejected. Sellers showed up immediately. This is the setup for the gap-and-trap and the signal to stay out of long entries until the picture clarifies.

The hardest discipline we developed across thirty sessions was doing nothing at the open. The first thirty minutes are the most active, the most volatile, and the most deceptive thirty minutes of the trading day. The patterns that look strongest at 9:35 a.m. CT are often the ones that have fully reversed by 10:00 a.m. CT. Watching without acting was a skill that took genuine effort to build.

Robinhood Mechanics: Everything We Learned the Hard Way

The Order Type Taxonomy

Market order: Executes immediately at the current best available price. You have no control over the price you receive. In a liquid ETF during normal market hours, the slippage is minimal. In a fast-moving stock or at the open, you may receive a price significantly different from the last quote you saw. We used market orders primarily for exits when speed was the priority and price precision was secondary.

Limit order (buy): Only executes if the price reaches your specified level or lower. You control your entry price, but you risk missing the trade entirely if the stock never comes to you. We used limit orders for most of our entries, setting the limit a penny or two above the ask to improve fill probability without paying excessive premium.

Limit order (sell): Only executes at your specified price or higher. You control your exit price but risk the stock falling through your limit with no fill. We used limit sells for target-price exits when we had time to be patient.

Stop loss order: A trigger price at which your order converts to a market order. When price reaches the stop, you are guaranteed to exit, but not at any particular price. In a liquid security during normal trading hours, the difference between the stop trigger and the actual fill is usually small. In an illiquid security or during a gap-down open, the fill can be significantly below the trigger. We adopted stop loss orders as our standard risk management tool after the LNG stop limit failure.

Stop limit order: A trigger price at which your order converts to a limit order (not a market order). If price moves through both the trigger and the limit before your order fills, you remain in the position with no protection. This is precisely what happened on Day 16 with LNG. The stock gapped down through both our stop trigger and our limit price in a single fast move. No fill. We held longer than intended and exited manually at a worse price. Stop limit orders are off the table for single-stock positions. They belong only on highly liquid instruments where the bid-ask spread is tight and gaps are unlikely.

Trailing stop: The stop price moves up automatically as price moves in your favor, maintaining a fixed dollar or percentage distance from the high. If price rises from $50 to $55, a $2 trailing stop moves from $48 to $53. If price then falls to $53, the stop triggers and you exit. We used trailing stops on some of the UCO and LNG positions with good results. They are excellent for letting winners run without having to manually adjust stops.

The PDT Rule: The Constraint That Defined Everything

The Pattern Day Trader rule is the single regulation that shaped more decisions in this experiment than any other factor.

The rule: any account below $25,000 in equity is limited to three day trades per rolling five-business-day window. A day trade is defined as buying and selling the same security within the same trading session. If you use your third day trade, you cannot day trade again until the oldest of those three trades drops off the five-day window.

The consequence in practice: if you buy a stock in the morning and want to sell it that afternoon because the thesis changed, that is a day trade. If you buy a stock in the morning and want to place a stop loss for the same session, that stop is a potential day trade (if it triggers). Managing exits on new positions required choosing between using a PDT slot or carrying the position overnight unprotected.

We built an entire risk management philosophy around this constraint. Stops were often placed the following morning rather than the entry session. Entries were timed to avoid same-session exits. Some positions were held overnight specifically because selling would have burned a PDT slot we needed for a different opportunity.

Then, on April 14, 2026 — Day 22 of this 30-day experiment — the SEC approved FINRA Release 34–105226. The $25,000 minimum is gone. Real-time intraday margin monitoring replaces the blunt instrument of the three-day-trade limit. HOOD stock rose 7.61 percent on the announcement.

The implementation timeline is 45 days after FINRA publishes the notice, followed by up to 18 months for individual brokers to update their systems. The rule will not affect the remaining sessions of this experiment. But the rule that shaped every single trade for 22 sessions was eliminated during the experiment that documented its effects most thoroughly.

Other Robinhood-Specific Discoveries

Fractional shares cannot have stop orders attached. If you hold 0.4 shares of SOXX, you cannot place a stop loss on that position. Only whole-share positions support stop orders in Robinhood. This forced us to shift to whole-share entries across the experiment, which occasionally meant smaller position sizes than we might have preferred.

You cannot hold a stop loss and a limit sell on the same shares simultaneously. Robinhood reserves shares for the first order placed, leaving nothing for the second. We encountered this on multiple sessions when we wanted to set both a downside stop and an upside target. The workaround: choose one, or manually cancel and replace when price approaches either level.

Robinhood MCP sessions expire overnight. An empty response from the order confirmation function is not a filled order — it is an expired session. We learned this early and built rh_login as the mandatory first step of every session before any order execution, regardless of whether a session appeared to be active from the previous day.

The two-step execution protocol is non-negotiable. Every order in this experiment went through a dry run first, which returns a confirmation token, followed by a separate confirmation call that actually executes. Skipping the dry run and calling confirmation directly risks executing at stale prices or on incorrect parameters. We never skipped this step.

The Full Trade Ledger (March 9 to April 24, 2026)

The Robinhood order history API returns ticker symbols as “unknown” for equity orders due to how instrument IDs are stored — a quirk we noted early. The prices, quantities, dates, and order types all confirm against our session records. The ledger below is reconstructed from both sources.

```
+--------------+-------------+----------------+---------------------+---------------------+-------+----------------+
| Date        | Ticker | Entry               | Exit                | Qty   | P&L            |
+-------------+----------------+---------------------+---------------------+-------+----------------+
| Mar 9       | SOXL   | $49.73          | $53.31              | 0.40  | +$1.44         |
| Mar 9       | ETH    | --              | $1,997              | 0.013 | liquidation    |
| Mar 10      | SOXL   | $25.60          | $25.41              | 0.97  | +$0.28 net     |
| Mar 10-11   | JETS   | $54.62          | $56.15              | 0.18  | -$0.19         |
| Mar 13-16   | GDX    | $340.38         | $336.41             | 0.07  | -$0.55         |
| Mar 16      | GLD    | $474.13         | $461.89             | 0.03  | -$0.39         |
| Mar 24      | SLV    | $45.38          | $46.31              | 0.33  | +$0.31         |
| Mar 24      | SOXX   | $100.00         | $94.46              | 0.10  | -$0.55         |
| Apr 2       | SOXX   | ~$39.42         | ~$39.82             | 6     | -$0.48         |
| Apr 2       | XLU    | $130.46         | $133.74             | 2     | +$0.31         |
| Apr 2       | UCO    | $39.42 / $40.12 | $39.82 / $40.55     | 6+4   | +$1.96         |
| Apr 6       | CF     | $132.38         | $135.36             | 2     | +$2.96         |
| Apr 6       | UCO    | $40.12          | $40.55              | 10    | +$0.43         |
| Apr 7-8     | LNG    | $284.17         | $286.71 / $288.90   | 2     | +$9.44         |
| Apr 8       | GDX    | $99.27          | $267.77 (stopped)   | 3+2   | -$15.48*       |
| Apr 8       | SLV    | $69.29          | $67.53 (stopped)    | 4     | -$7.04         |
| Apr 14      | ITA    | $234.24         | $225.93 (stopped)   | 1     | -$8.31         |
| Apr 14      | SOXX   | $57.60          | $75.00 / $73.82     | 3     | +$52.26        |
| Apr 16-17   | GDX    | $85.74          | $102.00 / $83.46    | 2+1   | +$7.80 (net)   |
| Apr 22-23   | GDX    | $41.28          | $42.06              | 10    | +$7.80         |
| Apr 22-23   | SOXX   | $440.89         | $441.45             | 2     | +$1.12         |
+-------------+----------------+---------------------+---------------------+-------+----------------+
```

*The GDX Day 16 entry at $99.27 and the Day 17 add reflect two separate tranches. The $267.77 stop reflects the larger-dollar GDX position that grew with the USO and Dogecoin liquidation proceeds entering the account in that period. Full net for that position: approximately -$15.48 across both tranches.

Starting capital: $25.00 Current cash balance (confirmed, April 24, 2026): $1,353.20 Account growth from experiment capital: the starting $25 grew through trading gains and the proceeds from liquidating the legacy USO and Dogecoin positions that were in the account before Day 1.

Note on the ledger: I am being transparent that several line items above carry approximation flags. The Robinhood API does not return ticker names on historical orders, and some early sessions involved fractional shares at prices that do not map cleanly to round-lot accounting. The cash balance of $1,353.20 is the authoritative number confirmed directly from the account. The individual trade P&Ls are reconstructed from session notes and the order history price data. The ledger is accurate in direction and approximate in precise dollar amounts for the early fractional-share trades.

Claude as a Trading Partner: An Honest Assessment

The experiment set out to answer whether an AI could be a genuine trading partner. After thirty sessions, the answer is: yes, with specific and important caveats.

Where it worked well: macro thesis construction was genuinely collaborative. The connection from Hormuz to LNG to fertilizer to CF Industries was not something I would have mapped as quickly alone. The ability to ask “what sectors benefit from a natural gas price spike?” and receive a structured answer with supporting logic, not just a list of tickers, accelerated research meaningfully. Order mechanics — the rules around stop loss versus stop limit, PDT accounting, fractional share constraints — became institutional knowledge that Claude retained session to session via the documented system.

The blog writing was the clearest demonstration of the partnership model. A trading session that might produce 20 minutes of notes became a structured, readable daily recap in a fraction of the time it would have taken solo. The discipline of writing the recap also forced better thinking about what had happened and why.

Where it had friction: Claude has no persistent memory between sessions without the system context being provided. Every session started with a handoff of relevant history. In the early days before that process was refined, context gaps led to suboptimal decisions. Additionally, Claude cannot read a chart autonomously — chart screenshots had to be provided, which added a manual step to every session.

The tendency to hedge rather than commit was occasionally frustrating. Good trading requires a point of view. There were sessions where I wanted a clear directional call and instead got a well-reasoned presentation of both sides. That is good epistemics and sometimes bad trading advice. The balance between analytical rigor and operational decisiveness is something a future version of this experiment would need to calibrate more intentionally.

The workflow that worked best: pre-market scan with Finnhub macro data, indicator dashboard review (S&P 500, Nasdaq, Dow, Russell 2000, VIX, WTI, Brent, gold, 10-year yield, DXY), chart screenshot analysis for any potential entry, thesis validation, order execution via Robinhood MCP, stop placement, and end-of-session blog recap. When we followed that sequence consistently, the quality of decisions improved noticeably versus sessions where we shortcut the pre-market work.

The PDT Twist Ending

A 30-day experiment about trading with a $25 account is, unavoidably, an experiment about the Pattern Day Trader rule. Every single trade in this experiment was made with PDT awareness as a background constraint. Do we have day trades available? Can we place this stop without burning a slot? Should we hold overnight because selling intraday would cost a PDT allocation we need tomorrow?

Then on Day 22, the SEC eliminated it.

FINRA Release 34–105226, approved April 14, 2026, replaces the $25,000 minimum equity threshold with real-time intraday margin monitoring. Brokers will have up to 18 months after the notice period to implement the new framework. The rule that shaped every trade in this experiment will be gone before the next one starts.

What would have been different without PDT? The CF Industries entry on Day 15 could have had a same-session stop. The LNG position on Day 16 could have been actively managed intraday without burning slots. The ITA entry on Day 21 could have had a stop placed the same day it was opened rather than requiring an overnight wait.

For retail traders who want to be active with accounts below $25,000, the PDT elimination is the most significant rule change in years. The immediate beneficiary is Robinhood: HOOD rose 7.61 percent on the announcement, a direct read on how central PDT management is to their platform’s value proposition for smaller accounts.

What Comes Next

The $25 experiment ends here. The education does not.

Season 2 of Trading With Claude, if it happens, looks different. PDT will be gone. A larger starting base — call it $500 or $1,000 — allows whole-share positions in a wider range of instruments without fractional-share stop-order constraints. Options become available as a risk management tool: buying puts as downside protection instead of relying entirely on stop orders is a different and in some ways superior framework.

The macro environment will be different too. The Iran war’s first-order trades are done. The second and third-order effects — rearmament spending, LNG infrastructure rebuild, regional energy diversification, agricultural supply chain restructuring — will play out over the next several quarters. Some of those are long-duration investment theses rather than trading positions, and the distinction matters.

The one thing that would not change: full transparency, every trade on the record, every loss documented the same way as every win. That discipline was not just good journalism. It was better trading. When you know you are going to write about every decision, you make fewer bad ones.

This post is a personal account of a documented trading experiment and is not investment advice. All trading involves risk of loss. Past performance does not indicate future results. Nothing in this post should be construed as a recommendation to buy or sell any security.

Color indicator: this experiment finished in the green.