Honestly, the " Ukrainian president Zelenskyy was "put into place by the CIA under Obama" despite being elected president in 2019 during Trump's first term." thing is another one I find totally baffling to try to understand from a 'MAGA' perspective.

The claim that Zelenskyy is some sort of CIA puppet is, of course, false; but even if it were true that would essentially just make the situation in Ukraine a proxy war of the sort that the guys sneaking stinger missiles into Afghanistan back during the Soviet era would have given a kidney for: Massive Russian losses in men and materiel, a more or less complete humiliation for claims that post-soviet Russian military hardware is anywhere near as credible as advertised; and Europe goes from losing interest in NATO to forming an eager line and banging on the door. All that for a relative pittance in military aid, including a lot of stuff we were just waiting to age out in storage.

From a 'MAGA' perspective that seems like an absurdly good deal.

Is anyone else as puzzled as I am about how this is supposed to be serving American greatness?

That seems like a fairly niche application. Sure, all kinds of hiring outcomes are possible if daddy is on the board or whatnot and this wouldn't be markedly less adequate than other plausible-enough excuses; but, barring some sort of strong reason for them to try to shove you into the job, 'AP course' is already well known to mean "not college level experience but your high school had easier ways of passing" which is pretty marginally prestigious.

This instance is particularly dramatic; but the incentives make avoiding this class of problem exceptionally difficult.

Is it in anyone's interests to have Heathrow go down and flights delayed all over the place? No. (more specifically 'no' among all airlines and national transport types; there's obviously a free-floating group of random extremists and foreign policy opponents of the UK who are in favor; but nobody expected them to pay for infrastructure anyway)

Is it in anyone's interests to toss their own cash into the pot when the power supply to Heathrow hasn't failed in years and, anyway, the backup generators are probably fine, no need to check what those whiners in facilities are saying? Also no.

It's perennially difficult to find someone to do preventative spending for fairly low risk situations; sure, sell me on paying two or 3 times more for electrical substations when the one we have has worked just fine for years; and it's even harder when the benefits of reliability are distributed among a fairly large number of entities whose day-to-day incentives are to lower costs to the degree possible: Having to cut Heathrow out of their plans suddenly and without notice likely cost the various airlines a small to midsize fortune; but, on any average day, every cent of 'cost recovery' that goes to Heathrow rather than to them is margin stripped away; and it's not like BA wants to get stuck paying more for the 'B' in their name when their competitors are paying less; so there's no obvious party on the hook.

He's definitely not an industry optimist; but the efficiency thing is arguably part of Coreweave's problem.

They don't get paid per unit of LLM performance; they get paid per unit of compute time(and currently lose money on it). Efficiency is actively bad for them unless it drives up adoption by greater than the amount saved; or if it makes certain workloads amenable to moving off the specialized larger clusters that they operate and onto smaller systems(the big GPU clusters are fairly involved HPC infiniband stuff in addition to high thermal density; single-chassis SXM systems are on the thermally dense side but much less exciting; and PCIe can scale down right to janky colo if you want).

Given their ongoing losses and their debt load their only real hope is either that OpenAI and friends can continue to bring in dumptrucks of VC cash to pay for open ended training exercises(ideally CUDA-dependent, to avoid direct price pressure from competing chips that have immature tooling but significantly smaller margins; or someone coming up with a very valuable and very computationally expensive, ideally infeasible to run on smaller scale systems, LLM workload that will significantly increase compute spend and willingness to pay. People just getting more 'decent-ish' for less H100 wall time will not help them.

If the raging bloodbath that is Coreweave's(OpenAI's new pet compute vendor, now that the MS honeymoon has soured a bit) S-1 filing is anything to go by it might genuinely be that expensive, possibly yet another being sold below cost. I assume that some of their high-hype models are at least experiments in trying to break even; but for a technology that is supposedly unleashing a tsunami of efficiency there is shockingly little money in 'AI', once expenses are factored in. A good day to be Nvidia; and enough wheeling and dealing that anyone who knows how to be a transaction cost or gamble with other people's money has a decent shot; but it's genuinely impressive how far underwater all the nominal leaders are on their glorious new innovation.

Obviously a lower filing fee is a savings regardless of who you need to prepare the filing; but I'd be curious both about what percentage of the cost of obtaining a patent is fees (vs. the time required from patent lawyers and people suitably knowledgeable about the invention) as well as what the effects of going with or without lawyers are on chances of successful application, number and breadth of claims, and the like.

Is the situation like that of criminal law; where going in without a lawyer is typically such a bad idea that failing to ensure that a defendant has the option of one counts as a due process violation? Are lawyers fairly optional; but cheaper(at least on an efficiency basis) than sending the subject matter experts whose work you are patenting in to work on the patent? Are lawyer more of a speciality choice when you are trying to push the bounds of how many claims you can wring out of a given patent?

"living off the land" attack strategies are extremely popular. Way lower risk of signature-based detections than bringing in your own custom attack tools.

This is the sort of primitive, instinctual, behavior you see in organisms not cognitively advanced enough to require prior authorization. Pitiful really. As we all know, it's the sophisticated demands of billing and coding communication that encouraged the evolution of behaviorally modern humanity; absent that theirs is clearly a dead end path.

If it's an ESP32 specifically you don't presume to own the whole device. Most of the RF stuff is a vendor blob talking to undocumented hardware.

Not necessarily as different from your average server as one might like(even on linux systems with mainline compatible hardware it's totally unremarkable for there to be several DMA-capable peripherals, often ones with external network connectivity, running big chunks of blackbox firmware); but much of the wifi and BT behavior is basically off doing its own thing, with the limited separation just being down to sharing some resources costing less(vs. the older approach of just slapping an entire separate wifi/BT implementation on its own chip and having it presented to the microcontroller over a serial interface or the like; often as an extended Hayes AT thing.)

The stuff about spoofing 'trusted' hosts sounds a bit like hysteria from the wonderful world of BT security being something of a faith-based afterthought compared to networking flavors where adversarial environments were properly considered from day 1; but, since you get very little control over the RF side on the ESP32, it's probably good that someone is having a proper empirical look at what it is actually doing for anyone who knows how to ask.

Isn't the entire niche for 'fintech', especially now that basically any consumer-facing bank has a website and an app that are lousy but adequate and the cellphone guys are all in on payment processing, essentially pretending that structuring your operations to avoid banking regulations is actually a technological innovation rather than a legal strategy?

Obviously there's incidentally a lot of tech involved in pushing banking beyond handwritten and manually verified checks during bankers' hours(and the banking industry knows it; it's not for nothing that MICR codes are mid-50s; but the people actually just doing tech for financial purposes, rather than preening at 'fintech' startups just do business process automation, mostly quietly, and the people who talk 'fintech' tend to be just playing regulatory arbitrage.

I suspect that a major incentive is to retain control over what vendors do rather than what vaguely techy or price sensitive users do.

Compare the trajectory of CDs and DVDs: the means by which many MP3 players, especially the higher capacity ones, were actually filled often didn't necessarily reflect perfect adherence to copyright law; but, because ripping a CD raised no DMCA issues, even relatively risk-averse and family friendly consumer tech outfits shipped with CD ripping out of the box and Apple moved eleventy-zillion ipods and everyone else shifted even more MP3 players but made less money at it. DVDs, by contrast, had CSS cracked wide open relatively quickly and never fixed; but DVD ripping remains a DMCA issue and nobody particularly mainstream touched a rip-based replacement for DVD players. A few niche outfits did what they could and got sued into the ground despite selling fairly expensive niche A/V gear that actively pandered to Hollywood with things like requiring periodic re-insertion to keep a DVD in the library or using raw disk images and only decrypting at playback time rather than storing cleartext video.

None of that stops rips from being trivial to obtain, or dodgy physical copies showing up for a pittance at flea markets; but it gave the film guys a de-facto veto over what features would be acceptable in the consumer playback market that lasted basically as long as the relevance of disks did; with future streaming activity likely to provide even more granular control; since re-authentication now happens almost constantly; rather than once like with DVDs, or fairly occasionally with BD+ obfuscation updates.

It's not clear that the film guys made good decisions; we know from history that the VCR saw hilariously strident objections before making them a zillion dollars by opening a major new market; but they care more about getting to make decisions than about their quality; and that's what even pitifully weak DRM gives them.

I'm...deeply puzzled...on why any license or data would be needed to 'operate firefox'.

This isn't a thin client where I'm asking Mozilla to process stuff for me and return the results; it's a local binary doing stuff in response to local inputs and the remote hosts I point it at. No licensing involved with me processing my own data; and any chatter with remote hosts would be subject to their terns. Unless they are in fact tapping the line why are they talking as though using firefox makes Mozilla Inc. a party to the data transfer?

My layman's understanding was that the history of direct human vs. computer chess veered away from competition and toward training purposes or engines obeying uci_elo limitations quite a while ago; with the last human victories of any real interest being some draws in the early 2000s.

Has this just survived as a perfunctory interviewer question; or did humans vs. bots actually stay interesting longer than that?

So it's a win-win for Google! Next week; a proposal from 4-6 of the largest 'cloud' and software vendors to formalize an oligopolists for safety consortium; but don't worry; it's still competitive because you can maintain certification by sharecropping on any one of them at the price they choose!