What I traded: memory and semiconductor equipment

June’s research followed AI spending into memory, packaging, equipment and networking. Some of that work turned into actual trades: MU, a leveraged SK hynix ETF and KLAC were profitable, while AMAT had an unrealised gain at month-end. A researched company was not automatically a purchase; the actions below are those recorded in the monthly review.

  • MU: I repeatedly entered and exited short-term trades. They were the largest source of realised gains in the review, but involved frequent turnover.
  • 07709, the 2× long SK hynix product: I made two rounds of trades and captured positive momentum. KLAC also contributed through a short-term trade. AMAT remained open at month-end, so its gain was still on paper.
  • The month-end stock positions included Ping An, AMAT and TSM. The remaining short puts included Laopu Gold and NVDA; those obligations needed to be considered alongside the shares already held.

What I traded: losses, IPOs and protection

  • Ping An: I continued holding the shares; their unrealised decline was the largest single drag. MRVL: I traded the shares and several put positions, with one put loss outweighing much of the premium earned elsewhere.
  • NVDA: I cut an earlier put position shortly after entry as the risk deteriorated, while a different put remained open at month-end. Laopu Gold also had a loss on an earlier put trade; the remaining contracts were separate positions.
  • IPOs: 圣邦股份 was profitable and 大金重工 lost money. The combined IPO result was close to flat before separately recorded financing costs. The month also included SQQQ, UVIX, SOXS and RKLB option trades, mostly with small positive contributions.
A better research map did not automatically produce a better trading process.

My reflections · edited from the review

The industry framework became more useful, but I expressed it through too many small decisions. I repeatedly traded similar exposures while also managing long-term holdings, options, IPOs and hedges. The activity made it harder to see which process was earning its place.

MU and the SK hynix trades could be successful while the overall month remained modest. Ping An’s price decline and MRVL’s option loss mattered more than the number of winning trades. I also needed to distinguish AMAT’s unrealised gain from profits already realised, and judge Ping An on its long-term thesis rather than on a short-term trade’s timetable.

Costs were another warning. Repeated entries consumed part of the return, so finding more opportunities was not necessarily the most useful next step. The report’s central lesson was that my research was developing faster than my execution discipline.

My AI research partner’s suggestions

The review proposed the following priorities for the next month. They describe intended improvements, not a claim that I had already followed them consistently.

  • Label each trade before entry: long-term holding, option income, event trade, IPO or hedge. Record the reason, holding period and evidence that would invalidate the idea; review each category separately.
  • Combine shares, leveraged ETFs and options that depend on the same industry into one risk assessment. Check all short puts under simultaneous assignment rather than relying on the broker’s available buying power.
  • Review repeated entries as one event, including round-trip fees, slippage and financing costs. Ask whether a new entry adds a new reason or just repeats the previous trade.
  • For each hedge, identify the position it protects, the trigger and the exit condition. Judge its contribution to the combined position instead of treating a small hedge profit as a reason to trade more.

The reported result

The estimated return was +1.45% using approximate Modified Dietz. This accounts for the timing of external cash flows, but is not an exact daily time-weighted return. Realised trades and month-end valuation changes both contribute to the monthly result.