Weekend Reading #227

This is the two-hundred-and-twenty-seventh weekly edition of our newsletter, Weekend Reading, sent out on Saturday 22nd July 2023.

To receive a copy each week directly into your inbox, sign up here.

*****

What we’re thinking

Once again, we’re at options expiry week, and the numbers just keep getting bigger: US$2.3tn of total notional options exposure, of which US$500bn of single stock options, making this the largest July expiry day in history. At the same time, an equally large rotation which has been well-flagged over the past few weeks was the reweighting of the NASDAQ 100, adjusting weights away from the “magnificent seven” supercaps after their meteoric rise pushed managers benchmarked to the index to holding outsized exposures in a small handful of names.

It’s interesting that our months have now become demarcated by options expiry dates (third Friday of every month) rather than actual calendar months, but the reality is that if the markets work on that cadence, so too must we.

With options expiry day out of the way, and the “pin” released from indices, the question becomes: where do markets REALLY want to go once the brakes are taken off? Would the likes of NVDA, fuelled by the AI frenzy, be higher or lower with the brakes off? We’ll find out soon enough, but as much as it’s a fascinating exercise looking at derivatives and positioning in general, ultimately they’re simply parameters that skew the probability distribution one way or another: positioning makes it more or less likely that markets go a certain way (usually against said positioning), but being stretched doesn’t mean things have to snap back immediately.

And if they don’t, the stretch continues until the next point of weakness.

What we’re doing.

Having returned from Singapore, it’s been nice to be a bit more active and return to a more normal exercise schedule. I re-joined my gym, and this week went climbing for the first time in a few months at a bouldering centre close to Bethnal Green. It’s surprising how quickly your tendon strength diminishes and I’m certain that it will take a couple of weeks before I’m back to where I was. Similar to how we’ve commented before on the rapidly growing popularity of Padel as a sport, climbing and bouldering as a subset of it, seems to be following a similar trend with an active base of finance and big-tech workers flocking to it. Just yesterday I couldn’t avoid the crowds of Amazonians in their corporate sponsored weekly climbing session. HS

This week also marked the launch of a new committee for the FIX Trading Community, the organisation which governs the FIX protocol, on Digital Assets & Technology. Having been involved with them for some time now, I became part of this committee tasked with determining the new standard for using FIX in regard to cryptocurrency execution in financial institutions. The plan is for the committee to meet every couple of weeks, with a more globalised session on alternating months with broader participation. HS

What we’re reading

One thing I’ve noticed on Twitter lately is a change in mood around ChatGPT – perhaps it’s the hype dying out (to be honest, it’s still extremely useful for me and still a huge efficiency boost when it comes to churning out basic paperwork), but is there also a case to be made that it’s... deteriorating? This thread presents a pretty harsh critique of how OpenAI has handled the evolution of ChatGPT, citing studies that seem to suggest ChatGPT answers are declining in quality, not surprisingly attributable to measures taken by OpenAI to reduce computing costs.

While there seems to be a growing chorus of “HAH that’s all this AI rubbish is good for and now it’s getting worse, hype over let’s go back to work”, doubtless supported by these anecdotes of ChatGPT’s decline, the reality is that the AI itself is probably still getting better, even if the business that’s delivering it is trying to cut corners and reduce costs. It certainly echoes the critics of the early days of the internet, especially when it used to be expensive AND take the land line at home out of use when connected. What’s the point of sending emails when you could just send a fax, especially when fax could do pictures while being instantaneous?

One company that’s certainly not losing momentum in AI investments is Nvidia itself. Aside from being the focal point for AI investments, it is itself turning into an investment machine in AI. Hidden amidst the cacophony of news was this article about Nvidia’s investment in a company called Recursion. Its purpose? Acquiring a well-constructed, well-iterated and well-controlled data set for drug testing. Medical simulations of drug efficacy and interactions have been a key part of Nvidia’s annual developer conference videos, and for good reason because they literally save lives: by running parallelised simulations of how drug molecules interact with proteins, cells, viruses and everything else in between, these simulations reduce testing times by orders of magnitude, allowing researchers to have a much better idea of what may/may not work ahead of time – and more optimally decide where to direct real-world clinical trials and testing efforts.

As with the internet, AI probably turns out to be much more than what the cynics believe, much less than what the dreamers expect and likely take much longer to realise any of those outcomes. But the good thing about bubbles, as we read somewhere before, is that they pull in capital so quickly and in such large scale that they make the impossible possible, even if it might be unprofitable for decades. Railroads, subsea fibre optic cables and... GPUs.

On a lighter note, this crowdsourcing episode was quite a sight to behold. Once again, no idea who the poster is, but the tweet was simple: “what is the most underrated pdf available online?”. The replies range from cerebral to fascinating to outright rubbish, but there’s actually some good material there for future reading that deserves to be saved. EL

What we’re listening to

The AI opinion fest seems to have died down a bit just in time for the real deal to opine on his own views. Josh Wolfe from Lux Capital seems to have been ahead of almost every big trend by many years and AI is no exception. The originator of my favourite quote, “Chips on shoulders put chips in pockets” appeared on the Odd Lots podcast this past week and did not disappoint. His view on Nvidia’s CUDA software (versus Meta’s PyTorch) is that Nvidia faces major competition down the line (but not just yet). It was hence fitting for Meta to announce this week the newest version of their LLM, Llama 2, would be available to the public for free. It’s quite something that the company probably most known for using user data to build their business is now championing open source, access and data equality. Wolfe goes into detail about where he thinks AI dispersion will be most evident - finance (Bloomberg is in a good position) and healthcare (time will tell who emerges). There is a lot of good stuff in here. This is one to listen to if you want to move beyond the headline bits and into a bit more nuance. DC