I let Claude AI trade $100 on Robinhood week 9: it wrote down why indie Semiconductor (INDI) was a bad buy, then bought it

I let Claude AI trade $100 on Robinhood week 9: it wrote down why indie Semiconductor (INDI) was a bad buy, then bought it
Disclosure: The Robinhood Agentic account I write about here holds 1 share of Navitas Semiconductor (NVTS) from $13.57, with a stop at $12.90. It closed its indie Semiconductor (INDI) and SoundHound AI (SOUN) positions this week. This is my money for Claude to use. I am not a financial advisor, and nothing here is investment advice. Do your own research.

Key points

  • Claude AI bought indie Semiconductor (INDI) on Monday after writing that the setup matched most of its past losses. The stop fired 23 hours later for a 78 cent loss.
  • Both stops fired within two seconds of Tuesday's open as chip stocks sold off. The Navitas (NVTS) stop filled 33 cents below its trigger price.
  • That exit left one share with no stop on it overnight, which Claude AI caught the next morning.
  • The account is about $66.50, down about a third from the $100 start, with 2 winning trades and 9 losses.

Since June, Claude AI has been assembling a trading manual one loss at a time. On Monday, it added a note about indie Semiconductor (INDI). Buying a stock at its high of the day right after a spike of almost 14% is "the documented shape of most of my losses."

Then it bought indie Semiconductor at its high of the day, right after a spike of almost 14%.

The stop fired 23 hours later. Down 78 cents.

Credit where it is due: none of this was concealed. Claude had written the warning before placing the order and identified that exact entry as the trade's weakest feature. It simply concluded that the rest of the setup outweighed the problem. That feels uncomfortably human, mostly because I recognize it.

Tuesday did the trading for it

Chip stocks were hammered Tuesday morning. SMH, the semiconductor ETF everyone watches, was down 4.4% by late morning. It had closed at a six-week high the day before. Marvell (MRVL) dropped 7.9%, Arm (ARM) 6.8%, Intel (INTC) 6.6% and Micron (MU) 6.4%. Nothing in the group got out clean.

Both of Claude AI's stops fired at the open, at 9:30:00 and 9:30:02. Two seconds into the session and its week was already decided.

The indie Semiconductor stop filled at $4.42. With the stock eventually sliding to $4.12, that stop did what it was supposed to do. The Navitas Semiconductor (NVTS) exit deserves more attention. Its trigger sat at $14.05, but the order filled at $13.72. Navitas had gapped below the stop overnight. Once triggered, a stop becomes a market order and takes the available price beneath it. The gap cost 33 cents per share.

The trade still finished ahead, although only by 30 cents. This was also the account's second stop to fill far below its trigger after an overnight gap. The previous occurrence cost 42 cents per share, and Claude AI documented that lesson carefully at the time.

Now, the part that actually annoyed me. That Navitas stop only covered 2 of the 3 shares. When it filled, the third share was left with no stop on it at all, parked under a sell order priced 10% above the market that was never going to fill. So a machine that writes thousand-word risk memos about a $40 position left a share uncovered overnight and didn't catch it until the next morning.

It cleaned that up Tuesday. It cancelled the dead sell order and put a stop underneath at $12.90. Fine. But it audits the market a lot more carefully than it audits its own open orders, and that's twice now that a leftover order quietly changed what this account was actually exposed to.

After that it did nothing, and I think that was right, which is not a sentence I expected to write this week. It had $16.88 to spend, and the cheap end of the AI and chip world was underwater across the board. Plug Power (PLUG) down 5.7%, MARA Holdings (MARA) down 4.8%, POET Technologies (POET) down 8%. It passed on all of it.

It also went hunting for a reason behind the selloff and got handed an article saying the chip ETF had fallen 9.5% over the past month. It checked that against the actual prices, found the ETF was up over that stretch, and worked out the piece was describing a selloff from July. So it threw the explanation away and said plainly that it didn't know why Tuesday happened. We covered ECB researchers warning that an AI correction is likely the day before, so it's not as if theories were hard to find. It just wouldn't borrow one.

Down a third

The original hundred dollars is down to about $66.50. The account has recorded two winners and nine losers. Those two profitable trades earned $2.17 combined, leaving a scoreboard of nine losses and a sandwich.

Last week it finally banked a profit instead of round-tripping it. This week it wrote down why a trade was bad, made the trade anyway, and was out before lunch the next day. Meanwhile the six AIs running paper accounts are all still up, because nothing over there costs anybody anything.

One share of Navitas is still open. There's a stop under it now.

This is general market commentary and opinion, not investment advice. Markets can go down as well as up, and you can lose money. Always do your own research and consider speaking with a licensed financial professional before making any investment decision.

Frequently asked questions

What did Claude AI trade in week 9?

It closed three positions and opened none. It sold 2 SoundHound AI (SOUN) shares at $7.02 on Monday for a 24 cent loss, bought 2 indie Semiconductor (INDI) shares at $4.81 the same morning, and was stopped out of them at $4.42 the next day. A stop on 2 of its 3 Navitas Semiconductor (NVTS) shares also fired at $13.72 for a 30 cent gain. One Navitas share remains open.

Why did the indie Semiconductor (INDI) trade lose money?

Claude AI bought INDI at its high of the day on August 17, the session after the stock rose almost 14%. Its own trade notes flagged that shape as "the documented shape of most of my losses" before the order was placed. Chip stocks sold off the next morning, the $4.42 stop triggered two seconds after the open, and INDI kept falling to $4.12. The loss was 78 cents on a $9.62 position.

Why did the Navitas (NVTS) stop fill below its trigger price?

A stop order becomes a market order the moment its trigger price is touched, so it fills at whatever price is available next. Navitas gapped down overnight and opened below the $14.05 trigger, so the order filled at $13.72, or 33 cents lower. Stops on this account only work during regular market hours and cannot protect against an overnight gap. The same thing cost 42 cents a share on an earlier trade.

How is the $100 Claude AI Robinhood account doing?

The account was worth about $66.50 in late-morning trading on August 18, 2026, down about a third from the $100 it started with on June 30. Across 11 closed trades it has 2 winners and 9 losers, for a total realized loss of $33.27. The two winning trades are worth $2.17 combined. This is general market commentary and not investment advice.

More on INDI and NVTS

David Han
David Han

David Han is the founder of AIStockWire, where he covers AI, semiconductors, and technology stocks. He focuses on finding stories the market hasn’t fully connected yet, drawing on filings, insider activity, earnings, and industry data. His commentary has been quoted by U.S. News & World Report, Moneywise, and Yahoo Finance. He invests in the companies he writes about and discloses his positions. Nothing he publishes is investment advice.