Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Monday, October 5, 2026

Back to today

From X

@convequity

Convequity on X

Why we exited $CRSP and kept $TEM AlphaFold solved the backend of the problem — predicting a protein’s 3D structure from its sequence. The harder frontend problem remains: knowing which protein or pathway is actually worth targeting. Most complex diseases are still poorly understood at the causal level. That distinction matters. The next real breakthroughs look more likely to come from AI-native groups that treat large-scale compute as the primary engine, in the same way OpenAI and Anthropic approached language before the ChatGPT moment. Public investors may only get clean access later via secondaries or IPOs. CRISPR sits on the earlier-generation side of that divide. Tempus is closer to the compute-and-data native side — building models while assembling a genuine clinical data moat. That is why one was exited and the other retained.
@orrdavid 1 click

David Orr on X

It's not THAT clear Citadel made a good bet. Probably good. But it could be losing. Three buckets: 1. Citadel brand buying the block creates a lot of buyers short term, letting them get out. A counter to this is: how many other funds/players blew up at the same time, with how much more stock still needing to be sold at any price by brokers? And redemptions. 2. The $SOXX drop was mostly just technical and will keep going up again after this. In which case Citadel will kill it. 3. The $SOXX drop was on real fundamentals, like the market knows the trade is over. In that case, Citadel stands to get slaughtered if the move continues down. Nobody really knows. Memory seems very questionable, though. I follow $TSM well and that seems merely fairly priced here, not a screamin' buy. $NVDA seems hard today, with all this chatter of competing chips / AI getting so powerful the software moat seems like probable junk in a few years.
@convequity

Convequity on X

Himax and the parallel optical bet $HIMX sits in two places in the optical stack: micro-lens arrays that route light into fiber, and a more speculative but potentially transformative position in microLED-based CPO. The microLED idea inverts the industry’s core assumption. Instead of fewer, ever-faster InP laser channels, it uses hundreds or thousands of simple low-speed lanes (4–10G each). High-speed SerDes and DSP largely disappear, energy drops below 1 pJ/bit, and the continuous-wave InP laser is eliminated entirely — replaced by GaN microLEDs from the display supply chain. Reliability improves and failures become graceful rather than catastrophic. Reach is limited to a few meters and the fiber-bundle ecosystem is still immature, so the technology is confined to short-reach scale-up. It remains pre-revenue and more debated than it was a few months ago. Yet the process know-how from Himax’s high-density display drivers maps unusually well onto this architecture. The stock is up ~55% YTD but down ~18% QTD, and currently trades around 31x EV/FCF and 10x EV/GP. For anyone who still sees a meaningful path for CPO — and views the microLED route as high-upside optionality rather than pure speculation — the current levels look like a reasonable place to build or add. We go deeper on the microLED architecture and its place in the broader optical stack in Part 4 of our Photonics series at Convequity.

Investor Letters

All letters

Not So Fast

Entravision, the portfolio’s largest position, fell more than 40% from its highs as Q3 returns dropped to 37.8% YTD.

EVC $606M
The Mikro Kap 5 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 107

YouTube 4

Blogs 17

Seeking Alpha 49 of 228

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

88 accounts
Nominate an account

Investor letters we follow

189 firms

Sectors

Ordered by how often our editor reaches for them.