Closer to the Hitchhiker's Guide to the Galaxy tradition. All they need are the words "Don't Panic" on the front cover.

If it doesn't, it has no real value.

Almost nothing on Wikipedia is true, but all of it is "verifiable" in that there's at least one published source making the claim. Even if 99.9% of all other sources say the opposite.

(The Haka is not, for example, a war chant. Which is attested to, but because it's claimed to be online, it is "verifiable".)

The number of errors is huge, the number of "official editors" who know anything about the subject is extremely small, and this has always been the case. This is why nobody reputable cites Wikipedia. Wikipedia is where you go to find out what is potentially interesting, but you ALWAYS obtain the facts from elsewhere, if you've any sense.

Your best engineers will be AuDHD, and AuDHD folk don't do well in these sorts of synthetic tests.

I can code in 20+ languages for over 12 different OS', write code that's good enough for CERN to run for particle accelerator data collection for a decade, have accelerated systems by an order of magnitude by understanding the poem space better than their previous engineers, but I am abysmal at exam style settings.

The software industry is riddled with SEs who write abysmal code but can pass a coding test easily. They're the folk that you have always needed to reject but didn't because they ticked a box.

The DOD/DOW have been working on this for a while. They discovered, early on, that the device had a propensity to cause permanent nerve damage.

It's very unlikely the US have fixed that, as they'd largely sidelined the work as a failure. However, at 36%, Trump needs something he can call a win, no matter what the long-term consequences.

The irony of this LLM AI stuff, trained on the data from the open Internet, is that it is appears to be great at finding and penetrating all sorts of vulnerabilities.
Seems to me that this will raise the costs of exposing anything to the internet and make putting stuff on the Internet a lot harder.

This is fine if you are comfortable showing that SOME chip can be run in space at SOME cost, but it skips all the details that drive whether a data center in space is will work logistically and economically. There are many of these details, including but not limited to:
- Is it possible to use convective cooling in space? If so, how does one design it so that the convective fluid doesn't just dissipate? If not, can chips be designed that successfully evacuate heat radiatively or do the chips burn themselves up?
- What is the balance between collecting energy on the sun side of the system and protecting equipment from cold on the shaded side? Does the equipment have to survive large temperature cycles? If it can't does the data center have to use thrust to rotate itself to keep the hot side hot and the cool side cool? What would that mean for design cost, for refueling?
- How does all of this equipment (solar panels, batteries, GPUs, communication equipment, thrusters, control systems) deal with the high levels of ionizing radiation in space?
- The design and duty cycle of all these pieces of equipment implies various levels of refueling, repair, etc. Assuming all that can actually be done successfully at rocket's length, is the total cost within what the market will pay?
These details whether a data center can be made operational at all and if so whether their economics make them worth doing.

What Google is doing strikes me as a pretty sensible step - put a few chips in space and see what happens.

What Elon Musk does is claim that something will be solved next year and do some pump-and-dump crap around a company he claims will be doing that thing next year.
Where is "Full Self Driving", as in you can actually let the car drive and not monitor it? Where are the fleets of "Optimus" robots? Where is Tesla's super-awesome battery tech? And the Cybertruck that can float?
Oh right, they'll all be here at {year} + 1.

Sadly, it often is faster to do that than work within the old bureaucracy and its entrenched people, rules, and regulations.

This is pretty much a textbook case of a stopped clock being right twice a day.

Notwithstanding Hegseth's absymal record prior to and within the Department of Defense/War/Being Tough, the war in Ukraine has showed how different war can be run than the way our military is currently organized.

Even if if you don't buy the army having "Transformers" type robots punching each other out (and I don't), it's clear that there is going to be a lot of "AI"-ish and highly automated stuff fighting wars. There needs to be a unit devoted to understanding FPV drones on land, sea, and air, plus all of the infrastructure for using them, defending against them, the communications to run them, training people to manage these kinds of weapons and units.

...Modern ejector seats cause spinal injuries from which full recovery is "unusual". Smashing the spines of pilots might not be the most effective way to boost morale.

$22,000 per hour. The F35 will be nearer $42,000 per hour. These are under-estimates, as they do not include price hikes, crashes that close businesses, crashing into businesses, crashing into heavily-populated sports centres, etc.

Here's the question. What is the difference in work between review and understanding the code versus writing the code in the first place?

I think the results vary a lot by scenario. In some cases, the business problem, the requirements, and the implementation are clearly separable, and exceptions can be clearly defined and partitioned a prior. These are amenable to automatic code generation and "higher level of abstraction".
In other cases, the business problem and requirements can only be untangled by asserting some sort processing logic, and then working from there. The more granular the logic that has to be specified, the closer you are to saving time by just writing the code yourself.

I suspect that the estimates of massive labor savings put almost all work into the first category. In reality, I believe quite a lot of work lands in the second category, and much of the work in the first has already been automated.

Case in point, TFA mentioned writing a calendar app for her family. There are a bajillion calendar apps for families, and nearly every piece of enterprise software now ships with calendaring and project management features of some sort.
So this was a hobby project, or her family has some extremely unique un-apped needs. Who knows, as an Amazon exec she may need to schedule their personal helicopter, private jet, and security detail and that probably is more complex than everyone else's soccer practices, doctor's appointments, and school forms. But I have a hard time seeing how there's a huge space of unmet demand for calendar apps. The biggest problem with calendar apps is all the stuff that is going on outside of them that people do not want to put in their calendar. AI does not change that.

There is a whole culture around the tech bros that enables them to believe that what is good for them is ipso facto good for society. This allows them to think of themselves not as the ruthless capitalist barons they are, but as actual helpers and patrons of society.

This is also the basic manifesto of venture capital. Profits are inherently good and "consolidating fragmented markets" (i.e. monopolism) is also inherently good because they bring "innovation" to "everyone".

It's an exaggerated kind of neoliberalism from people who spend very little time outside of a rather narrow world.

The notion that AI can solve political problems is pure balderdash. Political problems get solved when people organize and interact with each other in ways that create stability and trust and enable societal or policy changes that address the problem. You can run AI a million ways to "solve" climate change, but until something gets people to negotiate and adopt different ways of operating, all AI has done is burn compute (ironically exacerbating climate change) while generating another set of outputs that nobody pays attention to.

But these are not people who like the slow boring of political work; they'd rather fantasize about some computer doing that.

Whether we like it or not, whether we use Windows or not, Microsoft is going to insist on adding the tax whether the PC comes shipped with Windows or not. This is illegal bundling, Microsoft has essentially ignored prior rulings, and it's pretty much a certainty the courts will be fine with this.