marty @Abs_Marty

SSDs vs HDDs... the data centre storage tech battle. A hard disk drive (HDD) is electromechanical: it spins magnetic platters continuously at a fixed speed (commonly 7,200 RPM in enterprise drives) and moves a physical read and write head across the surface to locate data. Power is only consumed to keep the platters spinning, whether or not any data is being read (plus additional power for head movement during access). A solid-state drive (SSD) has no moving parts; it stores data in NAND flash cells and only draws meaningful power during an actual read or write operation, dropping to near-zero when idle. The debate here comes from the mechanical versus the electronic distinction that determines power usage: •SATA SSDs draw around 0.04 watts idle and about 3 watts active •NVMe SSDs can peak around 11.5 watts active under heavy load but are also near-zero at idle •Enterprise HDDs draw roughly 4 to 7 watts idle and 9 to 11 watts active... more energy more consistently The tradeoff for this energy? Cost per gigabyte. •HDDs cost roughly $0.02 to $0.04 per GB •SSDs run around $0.10 to $0.20 per GB There is no clearcut winner to the storage battle. Data centres run a tiered strategy rather than decide on one or the other. When performance-per-watt is necessary (latency-sensitive, high-IOPS data): SSDs.
When cost-per-GB matters (archival, cold data): HDDs.

A "comatose" or "zombie" server is one that is racked, powered, cooled, and connected to the network, but has done no useful computing work for at least six months. It draws idle power the entire time. In 2008, they were an estimated 30% of total servers. (McKinsey) In 2012, they were an estimated 30%. (Uptime) In 2015, they were an estimated 30%. (Stanford & Anthesis) In 2026... they are an estimated 30%. (Energy Digital) Why?? There's a pretty significant asymmetric risk in decommissioning a server. Server owners cannot allow their IT to go offline - so much of their sales pitch will be based in reliability. The whole system is built around reliability. This means that the possibility of removing a server that is still running, and causing an outage, is just not worth the risk. Unused capacity accumulating costs over time becomes a small price to pay in order to maintain reliability. It's all about reliability. What this means? (Remember that 30% of servers is not 30% of energy usage (cooling, air systems, etc)) Two decades of improvements in the industry - virtualisation, cloud migration, and IT asset management - have not changed the fact that ~10% of the energy use that people complain about goes to nothing at all...

I was thinking back to how I’d go about my a levels with Claude; could be seriously detrimental to learning. I then thought how I was recently studying for my driving theory test and built a Claude artefact to ask me the exact questions that would come up. Took about five minutes to come up with a perfect test replicant. Once you dive into these tools and let go of anachronistic beliefs about what makes learning or working ‘noble’, you can go very far.

Scottish countryside has to be one of the most beautiful aesthetics in nature. Nice mix of engineering ingenuity and highland cows in lots of the views you get too.

@danielnewmanUV ** Meaningfully short on compute and energy

@DAcemogluMIT But but if private corporations don’t move ai forward dangerously quickly we might… lose the ai race?

@user712588888 @unusual_whales Wouldn’t be able to compete in perpetuity if any form of disincentive were put on robotics at the point

The biggest tech Ceos won’t really care about electoral politics as much as they will about those in power in the deep state - nvidia export controls for example will matter so much more than a Newsom versus Vance spat Nuclear founders care only about the nrc (different given it’s a bipartisan issue) therefore have no incentive to align. Makes the Valar atomics overt support interesting for them.

@inyouratmspher3 So worth knowing that working alongside university (if you’re in London this is a must) is so basically a necessity nowadays. Not even just internships but during the year. Equity research is a great option and could probably be done efficiently and pro bono to mutual benefit.

It starts a lot earlier than this too. If you're planning to build or run data centres over the next few years, there are only a handful of ways to actually get Nvidia chips, and most of them go through nvidia's partner network. A lot of those partners basically live inside Nvidia's ecosystem/ work for nvidia: their whole service supports the nvidia stack. Once you pick one, you're locked into the whole stack, and barriers to change become too high to bother.

I'll end this with a conspiratorial take that I thought of whilst writing the piece. We're seeing massive amounts of churn amongst the U.S. military hierarchy. I think that, internally, the military understands it might not be ready to fight a state-of-the-art war with China in a world where Taiwan has become such a crucial geopolitical node. I don't think it's unrealistic to say that, in a world where semiconductors are an ultimate commodity, the U.S. is fully focused on how it will retain connection to Taiwan. I can imagine that, in much the same way that the U.S. looked at its strategic hierarchy, it looked at its ability to defend sovereign land in the Middle East (i.e. Israel) (we can consider Ukraine in this and say that it was a wake up call, especially given Ukraine has now developed a military beyond America abilities to defend itself) as a proxy for Taiwan and thought that churn and iteration would be required. Remember that in the East Pacific, America does not have the same litany of military bases that it does in the Middle East, and Guam doesn't exactly secure it's presence in the region (wrong continent) in the same way that Qatar or the UAE does. This is to say that America did not need to start the war with Iran. It was the result of inevitably failed diplomacy. I wonder if this administration (Hegseth is a less than conventional secretary of defence whom I struggle to see is doing the job further than propagandising) has handed complete strategic sovereignty over to the Pentagon, who, better than anybody, would know what America would have to do in order to prepare to take on their greatest adversary. If that means a (and I hate to use the word because it belittles so much damage and horror into futility) cosmetic war in the Middle East that drives funding and technological iteration, then so be it says the Pentagon. I'll probably write more on my thoughts on this in the future.

As we move into the 'next phase' of the war in the Middle East, the uniquely over-prepared UAE we saw may now no longer exist. Something that very closesly aligns with the question, 'is America at all prepared to prepare its allies?' What the UAE currently has for air defence is a layered network stitched together from five countries: THAAD and Patriot PAC-3 from the US, Barak-8 from Israel, Pantsir-S1 from Russia, Cheongung II from South Korea, and the domestic SkyKnight (a good product but entirely short-range drone focused). The major stocks of ballistic missile thus far have derived from American and regional American ally purchases; America is running out of missiles... According to an analysis by the Center for Strategic and International Studies, the number of American Patriot interceptors fell from 2,330 at the start of the conflict to fewer than 850 by the end of July. US THAAD stockpiles, meanwhile, dwindled to about 250 from 452. At current projections, following Operation Epic Fury, it will take 8 to 14 years to meaningfully replenish American stockpiles. The UAE needs shit now. Remember, deterrent quality is a function of two factors: the disposition of munitions on hand and the capacity to produce more (Magazine Depth Affordability is a Choice, Alex Oliver). America's deterrence quality has waned - better to be frank (see below). The modern cost exchange has proved too steep and America's dependence on legacy munitions too expensive. Commercial is not always better, but it is decidedly cheaper. In cruise missiles, Convenants, Anthems, and Barracudas will all be priced at about a fifth of the cost of legacy Tomahawks; in ballistics, it's the same as Castellion's Blackbird GL will cost under $500k compared to the Lockheed PRSM Increment 1's $2 - 3m. Does this mean that we will see Anduril churning out missiles for the UAE? I don't know. What I do know is that the UAE will be examining how best to keep itself safe, American or not. Ultimately, why would it care if the missiles that kept the country safe were American or Russian. Why should it care? America is short supplies and so is Russia. Aside from the incredible opportunity this presents neutral defence markets (e.g. South Korea), this means one thing... Independence has to be the next step in the country's journey. For too long it has been a consumer that has lived on trade and foreign labour. It now has a (wealthy) population that has experienced and survived a war. Lean into this. Lean away from trade. America's war time economy has been driving the country forward since the end of the Second World War. I think that given the geopolitical positioning of the UAE, it is entirely necessary that manufacturing independence is built. Use Sharjah as the hub. I will talk much more about Operation 300bn across my posting, this is only an introduction. I think it's going to be a fascinating cog in the country's resurgence (defence is interesting to me, but infra and the robotics supply chain are even more so). Essentially, this post acts as a call to action. I think it'll be worth keeping your eyes on the idea of supply chain independence in the country, quietly so as not to disrupt the American relationship, but with assuredness: it will happen, in the face of war the incentives are too high. How quickly and in what manner will be interesting, though. Will defence startups see the opportunity ahead of them? Will the country push for geopolitical relationships to strengthen defence prospects (South Korean Hanwha will be working with EDGE to build an integrated air defence network)? 2027 will be fascinating as a game theory case study: America will insight further conflict, the UAE will have to protect itself, America may not be able to help as extensively. How does the UAE react?

Messy personal post to really begin this account... Life is about reaching the ultimate end (happiness): for me that's all about glory and adventure. We all know that everything in life is a means to an ultimate end, and since the turn of August I have been complementing a final means (penultimate end). I think for the SF founder type, whether they admit it or not and in whatever definition they choose, achieving glory comes immediately before happiness. To me, glory and adventure must come together. Why do something big if it's been done before. This thinking has lived in my head as Pierrot's excellent video essay title of, 'there is no glory at home' - the defining factor in me deciding to uproot my life post university, leave London, and move to Aberdeen of all places to work for a deep tech startup. Doing difficult, stupid things becomes easy when you romanticise them to yourself... I really believe in glory, and in a very Roman sense. I think it's a concept that can consume a life (for good or for bad). Specifically, I believe in gloria: that one should strive to do something impactful in their life, in the name of public service. More generally it is just associated with doing something big, something industry or generation or world defining. It is glorious to become the President, or build the next great company, or, in a much nicer sense, to improve people's lives (not saying the first two can't be the third...). What's interesting to me is how, in the modern day, the notion of the reluctant hero - the person who has a chance at glory thrust upon them without looking for it - is extinct. One now must search for glory. Anybody doing anything great is searching for glory. To me, that means one thing: the most glory will go to those who put the most on the line. To save a country, you must put yourself in harms way. It's a game of risk. One must operate at the absolute pinnacle of their 'excellent standards' (Pierrot, please watch) for as long as possible, doing the most difficult thing they can, knowing that it could amount to nothing. The idea of the reluctant hero is romantic, yes, but what could be more romantic than putting it all on the line for a chance at glory. It so often goes wrong. For every Cincinnatus there is a Boris Johnson who believes himself to be Cincinnatus (fun reference for those who remember). But for every Travis Kalanick (Atoms will be incredible) for whom it all unravels, there is a Patrick Collison for whom it could not go more right. Glory is all about putting it on the line, knowing it could go wrong, and doing it anyway. Hand in hand with glory is adventure. I believe that adventure must be unique, it must be unknown. One cannot go on a journey that has already been traversed, or else they will be doing it with a roadmap. When guidance is provided (and we have access to guidance in abundance), you are not exploring the King Solomon's Mines or the Misty Mountains, you are hiking a walked trail. That's fine, but why not go down the path untrodden. Why not sail into uncharted waters and do something Christopher Nolan will adapt into a blockbuster hit (Odysseus didn't look for adventure but when the time came he did rise to it; let the analogy live). It's so rare to see genuine adventure these days. I struggle to believe that what I'm doing now is adventurous. I don't see why moving to a new city and starting a new job is particularly unique. Maybe brave, maybe exciting, but not adventurous. This is to say that adventure today is preserved for the few. Only some will have the chance to do something genuinely unique, to voyage into the unknown. I am choosing to think that it is afforded to those who search for it, again it requires intention. I want to believe that after years of self-improvement and preparation, the opportunity arises. When that time comes one (I) will have to be willing to risk it all, knowing that the years spent searching, may come to nothing. What better way to spend a life, though. The common thread here is risk. Pierrot speaks about glory and adventure as risking mortal destruction (I only really mean a sense of personal failure). When the goal is too small, or the path trodden, one cannot put themselves in the face of genuine personal destruction. That begs the question, though: can one really fail, can they really face destruction, if they have truly searched for glory and adventure. This is where I think there might be nothing more romantic. You just can't go wrong (you can, don't be a dick) if you're truly pursuing glory and adventure. 'There is no glory at home' is a great call to both. Worth thinking about when you do something difficult.

Listening to Mark Carney talk about the US-Canada relationship in his forward guidance a few weeks ago. He sets out issues so clearly and so simply; engages with modern mediums of communication (why hasn’t Trump started a YouTube channel? Mamdani isn’t doing simple sit-down longer form content like this); and seems like such a principled person. I imagine the standards he’s demanding from his ministers are insatiable. Young Canadians should be clamouring to try and work for him. I also reckon he’s revelling in the Trump-centric world we live in. He can operate with such decisiveness and his ‘values-based realism’ seems so intellectual relatively. My god how lucky are Canada.

@a16z Power is furthest upstream, then the black space, then grey space. Can someone own all three systematically? Somebody could dominate by connecting all three. Space?

@a16z Vintage gpu reselling side hustle

Wondering why would Trump have even tried to have won in 2020? Aside from affording him time, what would winning then have done that winning in 2024 couldn't have done tenfold?

American oligarchs benefit most from Greenland; we aren't going to see a decline in grocery prices. When Trump's done spearheading autocratic rule in America, what happens to technofeudalists? With Vance it stays the same (obvious) but Rubio? Newsom? Osoff? AOC?

And how many do you think there are today in this region?

Best guesses, based on water and air economization, where is the best climate on Earth for a data centre?