Mid term play, mid conviction/risk

$TWLO @ $144.83, $GTLB @ $71.64, $TEM @ $69

Recently I’ve been thinking about Phase 2 of the AI tailwinds. 2023/24 was AI Phase 1, and it was dominated by infrastructure players. Examples of the key winners were semis, fabless chip companies, power and energy providers, servers and hardware providers etc.

Now, Deepseek was a pivotal moment - it reflected the ongoing commoditisation of LLMs. Essentially, the cost of LLMs should increasingly trend downwards due to competition and technology advances, and the end state becomes a winner take all LLM that is likely free for use (also the reason why current players are ready to throw so much capex to build the best LLM (or AGI eventually) now - bc it’s a winner take all market).

With commoditisation of LLMs, it’s negative for the upstream as cost and demand of compute decreases. What benefits? Probably companies that actually build on top of LLMs to add value for their customers. To relate to this, think of Windows OS being made free, and the apps built on it are the ones that create and accrue value. This is my current mental framework for AI endgame.

I’ve been looking into SaaS companies like $TWLO @ $144.83, $GTLB @ $71.64. Beyond that, I think healthcare and oncology will really benefit from using AI to create value in medical diagnosis.

For oncology, I’m in $TEM. Currently it’s $69, stock has clearly ran up lately, so waiting for a pullback is probably best. But we are probably still relatively early in the phase 2 trend, and besides I think healthcare is one of the areas where AI can really create a lot of meaningful value to humanity; so I really like the stock.

#spaceupdate

$RKLB +26%, $LUNR +44% in ~2 months since Nov 24.

Added to $RKLB @ $28.61, and added $ASTS @ $28.35 on T-mobile + Starlink launch yesterday

I continue to be very bullish on space. Public key leaders are only ~$10bn or less in mcap, and if space is going to be a key ambition of the Trump and future admin - very likely given Elon’s role in govt - then it’s not far-fetched to think govt investment gets directed to this industry.

We are also seeing real value to humanity per se being developed, in satellite connectivity and asteroid defence.

The risk is high though as the sector is very nascent, so current leaders may not shape up to be future leaders. But overall space will continue to gain importance imo, so I’m keeping a close eye.

#exchangesupdate

Closing this trade out at $COIN @$280 (-13.5%), $HOOD @$57 (+61%).

I’ve turned bearish on COIN since Jan, and increasingly bullish on HOOD for a lot of reasons:

- Most US based customers’ first exposure to crypto is actly through HOOD and not COIN. Cant find the data now but I’m pretty sure I’ve seen it before.

- I think market wrongly perceives coin to be THE crypto play, when in actuality there’s not much difference between hood and coin due to both of them having to comply with regulations — which slows down listing speed and competitiveness. (Eg for $trump coin, COIN only listed it 1 day before HOOD). Though I admit, US-listed exchanges will never be able to compete with lawless binance in listing speed; no point really talking about that in this HOOD vs COIN comparison.

- HOOD’s UI/UX is simply much better. COIN’s terrible, and lots of customers complain about their fraud detection false flagging on Twitter.

- If animal spirits are back, I’m sure it won’t be just limited to crypto. HOOD’s TAM is simply much bigger given access to trad equity market + options.

- Valuation wise, HOOD is also cheaper and has a stronger growth story vs COIN due to the larger TAM. For more look into their recent investor day for their future plans.

- COIN’s only “moat” vs HOOD is their L2 - Base - and in creating an entire ecosystem that they control and direct exchange users towards, hence why they are heavily subsidising use of it now. But L1/L2s are super competitive, so not very bullish on Base being a stand out vs Solana.

The bear case for HOOD however is that their customers (aka degenerates) lose money at such an unprecedented pace, that HOOD basically has to consistently invest in marketing to attract customers; pressures margins. Given the run up and my expectations of future volatility in 25/26, R/R seems off to continue holding these exchange businesses. I think $IBKR @$222 should outperform both HOOD and COIN from here.

Feel free to share this channel to anyone that’s interested!

It’s not financial advice, just meant to be a highlighter on interesting stuff, and a diary for myself. No entry price or price targets given out here too. Posts are unedited/undeleted to maintain transparency.

I try to share my entry point, and very very VERY occasionally my exit point to keep myself accountable. Sometimes I don’t share when I exit, so you should not expect this, and should take profit/cut losses whenever you yourself deem fit.

Short mid-term play, low-mid conviction, mid risk

Started building exposure to China in the past week, in the easiest way possible; long $BABA @$108 and $1810.HK @$43 HKD

Main thesis is in building out a hedge to American Supremacy, given DeepSeek & Trump’s America First policy; $BABA probably functions as well if not better than any index bets because it’s such a diversified business, plus it’s a direct AI beneficiary.

On Xiaomi, stock has run up a lot in the past year but I still like the company’s “uniqloish” position (cheap but quality) in a country facing a potential recession, while Beijing tries to resuscitate consumption. Co is also branching into EVs and related, so the growth story is there too.

Everything in China is cheap as hell on a valuation basis, but the risk is investors don’t trust that they can actually get their money out of China - a very legitimate concern. But R/R now favours a long now (unlike previous years) given the recent technological and political catalysts for a potential inflection in sentiments.

PS: It’s hard to beat the CCP, but let’s try (I like pain)

Elon’s retweet makes me think it’s either 1) robotics related, or 2) Starlink + iPhone related. Leaning towards the latter; if so $ASTS should react positively.

Or could just be a huge nothingburger too.

No hard evidence, but I have a strong feeling that the best humanoids/robotics companies (probably the next big trend besides space) will come out of China rather than US, given their historical hardware/systems innovation. E.g. see Unitree’s $14k USD humanoid

I think $1810.HK Xiaomi is a strong contender - given their expertise with home and car ecosystem appliances + their venture into AI with their poaching of Deepseek’s top prodigy. Nothing officially on this front yet, but it’s what I will be keeping an eye out for.

Do ping me if yall know any interesting public robotics companies!

A little mental framework that might be helpful for others.

Basically how hedge funds think about their PnL, amongst other things, boils down to 2 concepts; slugging ratio and batting average.

Slugging ratio = average nominal $ amount that you win vs average nominal $ amount that you lose

Batting average = your winning ratio - how often you win considering all your trades

Basically, slugging ratio x batting average = PnL

This means that while you have to get a majority of your trades directionally right (though the best PMs only have a batting average of 56%), more importantly, you have to have the guts to absolutely size up when you have conviction (maximize wins), or to have the humility and discipline to size down when you have low conviction (minimize losses) for a high slugging ratio.

What separates the top 10% of PMs from the rest is probably their ability to manage sizing and execution, and not whether they get the stock price directionally right or wrong initially. With positive slugging, you can be a winning investor even with a 50% win rate, or even at <50%.

It’s hard to buy higher when you are right or sell lower when you are wrong, but to win at this game this mental barrier must be completely removed (it’s hard). Some totally reasonable examples of sizing up can be adding after a fantastic print, adding as the market comes to see your non-consensus view, adding as reality outperforms your initial expectations.

One of the things that helped me was to calculate my batting average and slugging ratio on a stock basis. It helped me to figure out where I needed to improve.

If your slugging sucks, improve your execution (are you taking the appropriate amount of risk for the level of conviction you have?).

If your batting average sucks, improve your research/lower your trades turnover (take only higher conviction trades).

Good read if you’re interested:
https://visioninvesting.substack.com/p/batting-average-versus-slugging-average

Short term, low-mid conviction, mid risk

On China, added some semis exposure and AI tech this morning

$1347.HK (Huahong Semi) @$36.06 HKD, $603501.SH (Will Semi) @$153.33 RMB, $2158.HK (Yidu Tech) @$8.28 (Similar to Tempus AI of China)

Semis are a good bet here (tho slightly late) as the gatekeeping of chips from the U.S. will naturally lead to domestic solutions — DeepSeek is a great example. In this day and age, it’s very hard to keep technology within closed walls imo.

Also, I think China will likely beat US in creating AI usecases relevant to the consumer given how open and integrated their apps are + willingness of consumer to surrender private data + US priority on finding AGI first.

Eyes on $BABA Q3 2025 earnings tonight. High stakes event

I think bad print gets ignored for the bigger story and thus gets bought up quickly, while good print will be a pleasant surprise - so I’m long into the print and will buy post print if it falls. Gut says up post print. Focus is on e-commerce and cloud/AI commentary. E-commerce deceleration prob priced in, while cloud/AI “true” upside unknown to outsiders and thus can be hyped up by mgmt.

Only concern I have is that in China, news/earnings almost always get leaked early to those on the ground (sometimes even by mgmt themselves), so the stock’s price action before the earnings (negative here) may be indicative of the print — though this could also just be general risk off into print after a huge run up so not really a reliable signal.

We’ll see, godspeed