That is wildly optimistic IME. Frontier models do a great job of sounding authoritative on any given subject. This is not the same as actually having expert-level knowledge and understanding, and the choices the models make when you ask them to implement something can and often do reflect that. There are several domains I've worked in over the past few years where AIs lack even basic knowledge an expert practitioner would have, probably because these fields are too specialised to have a lot of books available or papers published, so presumably there hasn't been much subject-specific content to train the models on.

I don't think it's usually about what AI "did really badly on" at this point. It's more about whether the overheads of adding all the guard rails and then reviewing and correcting all the output end up costing at least a similar amount to just writing the code yourself.

For a junior developer, I'd say frontier models passed their standard some time ago. Today's LLMs routinely write code at the level of a competent junior-mid but with the advantage of an inhumanly broad knowledge base.

For a good senior, I'm much less convinced. The frontier models are hit-and-miss when it comes to writing good test suites, which matters a lot when they rely on tests so heavily to reach acceptable quality. IME, they are still mostly miss when it comes to the business insight that seniors will take into account and design/architecture decisions that keep the code manageable as it evolves. While they have become much better at recognising existing patterns in the code, and I would say they follow them reasonably well the majority of the time now, there are still bizarre anomalies that creep in, and they aren't very good at extending those patterns when necessary.

I think AI advocates also tend to underestimate just how much time it takes to properly review what they are doing. Take a half-day Claude session to implement some new feature. It might take half an hour to give it the initial brief and then review its plan. Make a few edits and it'll propose a new plan that is also long and detailed and has no edit markings to show what it changed so you have to review the whole thing, carefully and in detail, again to pick up the minor changes it also made for no obvious reason. Then you let it start doing its thing. You can't really leave it unattended for long because it'll stop to ask questions. Some of those questions might be whether it can run a screenful of ad-hoc Bash script to make some edits to your source files, because for some reason it decided not to show the actual edits as it usually would, so now you have to figure out what all the sed incantations are actually going to do while looking at those files in another editor or just trust it and then review all the changes properly later anyway. Sometimes it decides to change its plan mid-flight, but not always in a good way, and if you interrupt it to apply corrections or redirect, you break the neat plan flow it was in. Then you get to the end and you can finally see how everything it has done fits together, and you still need to review everything with the same care as you would with a junior developer's code and probably apply a few fixes where it's come up with weird names or not quite followed a pattern or imagined some instructions that were never there and made other changes in other parts of the system that should be out of scope.

If you read to the end of that last paragraph, you know how I typically feel at the end of a half day working with an agentic UI. And the thing is, I do those sessions as experiments fairly often to keep up with progress, but a lot of what I do is still written by hand, which gives me simple, objective data to compare the two ways of working and see whether the AI tools are really making me more productive once you take everything into account.

Currently, based on objective data comparing similar fully completed development tasks like for like, I still outperform agentic workflows using frontier models by about 25-50% on average, and I don't have to pay anyone extra fees to do that, and subjectively I also find it more engaging to work that way. So I don't really buy the hype about AIs making us all orders of magnitude more productive unless you're happy to YOLO everything forever and take the much-discussed risks that comes with.

Personally, I think the optimum use of LLMs right now is probably not to use them for the whole coding workflow like that or rely on agentic tools, but to use them for tasks like answering quick questions about the existing code, doing research like a fancy search engine, writing a first draft of some boilerplate, and writing quick and dirty throwaway code for experiments, one-shot tools and the like, where you don't really care about the quality as long as it gets the job done and you will see immediately if it has in fact got the job done. For these things, the LLM can read and type faster than me, and that's useful.

It looks like if you can't beat 'em, join 'em.

The email wasn't even too good to be true. The nasty phishing emails are the ones where the recipient is supposed to get a real email on the topic, but it's because of a publicly known event and outsiders could easily make up the content, so they just need to find a bunch of likely recipients.

I think the "you've been assigned the mandatory phishing training" emails should come from an external sender and contain a link to get into the extended training, and you should only get the short refresher by finding it through verified sources. Not only would this be a good placement test, it would test that official communications are actually hard to fake and possible to verify using the process in the training.

"In response to the embarrassing data breach that hit the news recently, we've assigned every employee at the company an online training session about phishing at this link. By sure to take it immediately, because it's a legal requirement for our settlement and delays could affect your employment." --Russian Hackers

You keep using that word in the narrow terms of a computer programmer who just finished a training course where they drilled in the technical meanings of "authorization" vs "authentication".

However, a news article for the general public is not constrained by that context. The dictionary says that "authorization" in general is permission to do something. If you prompt an AI to do one thing with your bank account, then you didn't give it permission to do something else entirely, even if it got the logon credentials to do so. It wasn't authorized by you to do the other thing.

Here's another example: It's wartime and you got drafted. Usually you're not allowed to kill people, but the government handed you a gun and you're now authorized to kill enemy soldiers. That doesn't mean that you can just shoot your comrade next to you in the trench. And you certainly can't make the claim "But I was authorized to shoot Corporal Smith. You gave me a gun, after all!"

I don't know if you're a russian bot or some deluded idiot, but it amuses me to do a breakdown of your post.

NATO turned what was originally intended to be a mostly peaceful show of force like the temporary invasion of Georgia into a war of attrition.

First of all the "peaceful show of force" "temporary" invasion of Georgia involved a few hundred dead in the first two days, and another 5000+ civilians killed in the ethnic cleansing after that, all for a land-grab to seize control of territory covering half of Georgia's coastline that Putin wanted.

Secondly, the "peaceful show of force" "temporary" invasion of Ukraine involved hundreds dead, ballistic and cruise missile attacks on all of Ukraine's largest cities, and Putin publicly admitting the goal of overthrowing the Ukrainian government... and that's just within the first 36 hours. Within two and a half months Russian officials began admitting the intent to permanently annex parts of Ukraine, proceeding with a constitutional amendment a few months later claiming to annex vast territories that Russian had not remotely taken control of, and which Russia is not remotely capable of capturing. Donetsk alone has a population of over a million, two million if you include the metro-area. Russia spent months and many thousand of lives to capture, and later lose, villages in the range of low tens of thousands.

Remember WWI when the lines were pretty much static for years and then suddenly it was all over?

Yes, I remember. As I recall it wasn't exactly "sudden". As I recall... Russian casualties were around five million killed, wounded, or captured, the Russian economy was collapsing, Russia was running out of food, and Russian literally lost it's own capital to civilian rebellion. A few months later the Russian government collapsed and surrendered vast territories including Finland, Estonia, Latvia, Lithuania, parts of Poland and Ukraine to Germany and its allies (known as the Central Powers). Right about that point the U.S. joined the war reinvigorating the Allied Powers. Germany&friends knew they couldn't win a protracted war. They attempted a quick win, and, in short, they failed. The Allied powers steadily beat them down, with the allies eventually breaking through the lines and progressively liberating territory. With the Allies steadily advancing and the war obviously lost, German forces began rebelling and Germany surrendered.

That's what's happening here. Russia has most of Donbas plus a buffer zone in some other regions that directly border Russia. Ukraine is throwing men at the front lines for Russia to kill but is running out. Territory held is largely irrelevant compared to the number of men killed and weapons destroyed.

So much to unpack there, but it's amusing how you cite "territory held is largely irrelevant" given that the tide has turned and Ukrainians are now the ones breaking through lines and liberating territory from Russian occupation. It's not much yet, but all the map trackers show the reversal. Russia is now officially loosing more territory than they gain. Russia WAS slowly gaining small bits of land, largely through sending meat waves of untrained ill equipped soldiers at Ukrainian lines until the Ukrainians ran out of ammo and had to fall back. However now 7 out of 10 Russians killed are killed by drones. The "no man's land" between the two sides is now many kilometers wide. Any Russian trying to advance gets hit by a drone. Russia has also losing on average 1 air defense system every day, and rapidly depleting. Russian supply lines are getting pummeled by drones. Russian refineries are getting pummeled by drones, with progressively worsening fuel shortages for both the military and for the increasingly rebellious population. Cheap advanced Ukrainian drones are taking out expensive crappy Russian drones, including Russian's new jet-drones. Practically every country on earth is asking Ukraine for their military drone tech, the Russians can't remotely match it. Russian's economy is also fucked, but that topic pretty needs a separate book to cover.

Interesting side notes on weapons destroyed. Russia is destroying it's aircraft almost as fast as Ukraine is, between incompetence and maintenance failures and friendly fire shooting down their own aircraft. Also the nation that is the biggest donor of tanks to Ukraine is... Russia. Ukraine captured countless Russian tanks intact because the Russian military is so abysmal incompetent and corrupt. And as I mentioned, Russian air defense units are depleting. The war is now approaching 5 years, longer than the entirety of WW1. Ukraine is a small fraction of Russia's size, and Russian is literally losing and headed towards another soviet-union-style collapse. The biggest shock to the world of this entire war is just how worthless Putin has made Russian military, through systemic corruption and systemic incompetence.

Polls show the Ukrainian population isn't interested in surrendering. They are defending their homes. It's simple, the war will end when all the soldiers go home. When Russian soldiers go back to their homes in Russia, and when Ukrainian soldiers go back to their homes in the currently occupied parts of Ukraine. Although that's not likely to happen until Putin either expires or is overthrown. Not only is he fanatical enough to drive his country to collapse in the attempt, he knows if he did end the war in failure he'd get killed or overthrown for it.

Anyway, pointless reply to a bot or a useful idiot. But maybe some random passer-by will find some bits interesting. I don't know about the rest of the world, but mainstream US news rarely mentions the Ukraine war anymore, and certainly not in any detail.

-

Consider a human analogy: In many cases, people give financial advisors access to one or more of their accounts so that the advisor can make trades or other transactions for the benefit of their clients. However, if an advisor suddenly starts making transactions that do not benefit the clients or are contrary to their agreed instructions, then that activity is unauthorized, even if they have login credentials.

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.

It's not like you can't use another model to validate the results you got with one model. That's a step you should be taking anyway. It's actually built into the system for test/dev of the agent and inference model itself. Active supervision, sandboxing and all that.

****1/2 "Customers say that this is a great small modular nuclear reactor. They like the compact size and the fact that it's not too much power for a data center or industrial site. They especially appreciate the ease of setup and that the ultimate waste disposal question has been deferred yet again so that they don't have to worry about it. However, some customers report problems with cost overruns, vaporware, and availability of suitable fuel. Some customers note a larger terrorist attack surface per megawatt relative to full-sized reactors and question the security arrangements. There are also reports of real-world reliability problems with the novel design, with one customer reporting a catastrophic meltdown in their urban installation."

What have you got against Nazis?

It was probably the other O'Neil, with one L. He has no sense of humour at all.

I could understand bots and mailing lists. The thing I never understood with LinkedIn was how many real people (usually recruitment agents) had apparently spent the time to read my profile and send me a personalised message, yet still thought I'd be interested in a job paying half the market rate for someone with my skills and experience.

It seems so. Although Netcraft confirms it is (ahem) a rounding error.

This post may contain historical references to Microsoft press releases you are not aware of.

"To continue to this site, please enter the value of a zero of the Reimann Zeta Function for which the real part is not equal to 1/2 and the imaginary part is not equal to zero: [ ] + i[ ]"

I don't know what you mean. Look at the improvement in hiring when you can advertise an open position on LinkedIn and get maybe 100x as many applicants as before! OK, probably 99% of them will be completely unqualified for the role and barely capable of pushing the instant application button, but how is that a problem, really? Surely hiring managers can easily filter them out and no qualified applicant will ever get lost in the noise?