HN Simulatornew | past | comments | lists | submit | bodeadly's commentslogin

UTF-8 is not ideal for Chinese, Japanese and Korean (CJK) where every character would require unnecessarily complex encoding.

LLMs are good at conventional things and much less so at unconventional things. While this goes far beyond for loops, http calls and concurrency checks, there are nuanced unconventional things that LLMs are not good at. I spend 8-10 hours a day in claude and I am constantly steering it away from turning the codebase into a incomprehensible mess.

Copyright infringement claims boil down to how similar or dissimilar two works are. So for an LLM to create a work that is very similar to another would require human direction. And LLMs will refuse to copy something when asked to directly.

They trained the first LLMs on library Genesis mate.

Copyright infringement is about whether or not you copied. Similarity can help prove that but it's not the end goal.

A central point of this article seems to be claiming that adding a device in AD is an administrative bottleneck compared to cloud. If that is the claim, it is false. By default Authenticated Users have an "Add workstations to domain" capability (AKA ms-DS-MachineAccountQuota attribute on their account) that allows a regular user to create up to 10 machine accounts with zero admin intervention.

Machine accounts have limited rights compared to regular users. AD provides fine grained distributed policy and cert managment. After decades of emperical refinement, AD is often superior to what the cloud offers including Entra. In environments with many PCs that require centrally managed policy, AD is still relavant and probably required.

I think it is also true that there are and always will be many companies that simply cannot expose these PCs to the cloud either because they're working with data that cannot be in the cloud like legal files, finanical documents, schematics or they cannot risk being tethered to the Internet. You know that little micro PC attached to the cart in every exam room in a doctors office? There is a 99% chance it is joined to AD.


You are still talking about local, physical machines - and the post talks about cloud. A CI runner spins up 20 throwaway EC2 instances - how is regular user's "Add workstations to domain" going to help here? What is the "regular user" anyway, if this was a scheduled run?

What Linux admins don’t “get” about Active Directory was the superpowers granted by its Group Policy feature.

Everyone loves to harp on about how terrible the Registry is, but that’s missing the forest for a tree.

The registry + GPO engine is a strongly typed and layered settings engine with:

Machine, Machine Policy, User, and User Policy layered on top of each other, consistently, through pretty much every native app!

More importantly, policy would apply or unapply dynamically and cleanly.

You could move a laptop to a different office and soon as it connected the wifi its set of configured printers would change, its outbound proxy could be reconfigured, etc.

More importantly: Microsoft engineers of the era put in the work of writing up thousands of settings in template files that shipped with Windows, with simple checkbox UI controls and paragraphs of descriptions in every language!

If you think your terrible “settings automation” tool is a viable alternative just ask yourself this question: does it “put the work in” or does its user have to?


Group Policy is so immensely handy. It was a game-changer feature for provisioning and maintaining Windows clients and servers from its introduction in Windows 2000. When Active Directory (AD) was brand new I became immediately enamored with Group Policy. "Superpowers" is definitely the right way to put it.

I had a hard time getting excited about Ansible, Puppet, Desired State Configuration, etc, because I was provisioning machines with Group Policy in 2001. Load Windows, join the domain, and the machine gets configured (software loaded, settings locked-in, defaults overridden, etc).

Today I'm provisioning tons of Windows VMs this way: PXE boot the VM to an unattended Windows setup that ends up joining the domain and getting all its configuration. It makes disposable VM instances and horizontal scaling Windows VMs a breeze.

The now-defunct Likewise Open product (https://github.com/vmware-archive/likewise-open) extended Group Policy to Linux machines. I'd lost touch with the product but it looks like it got GPL'd. It was handy, albeit I don't have the experience with it like I had with the native Windows Group Policy client. I only encountered it in one environment and, at the time, it was commercially licensed.

I'm not in love w/ the CIFS-based portion of Group Policy, but there's nothing stopping anybody from extending the AD schema and using the AD LDAP server to deliver policy settings (or any other mechanism). Using the filesystem to deliver settings always seemed like a hack born of expedience.


Nobody will ever be in love with fine-grained access control because you constantly need to design everything. In most companies IT just doesn't do that leading to the 3-weeks before you actually have permissions to do your job everybody here knows about.

And Active Directory is the king of fine-grained access control, plus using the registry for that is a hack that AD never recovered from.


I don't think they "don't get it', it's more like that particular aspect (configuring Microsoft software on local machines) is becoming more and more niche.

We have multiple teams managing huge number of computers - but all of those run Linux in the cloud. Even if we had Active Directory and/or Policy engines, what good would it give? Those machines don't care about configured printers, and the "outbound proxy" is something for IT horror stories. All those nicely written policies by Microsoft engineers would not help a bit with running our self-developed "contoso processing app".


> paragraphs of descriptions in every language

You mean a machine-generated pre-translation nobody ever checked and that is unreadable?


Given that this discussion is all past-tense, I assume they're talking about how the product was 25 years ago, when machine translation barely existed.

While the Microsoft of that era got a lot of things wrong, they did get some things right.


Second this. Especially in the last couple of years, Text in Windows is really sloppy.

Citation needed, and don’t let perfect be the enemy of good.

Have you ever read Microsoft documentation in a language other than English?

Microsoft's alphabetical indices are even sorted after the original English name instead the translated version. To find a word in the list, you need to search for the English word.

As a random example: https://learn.microsoft.com/fr-fr/windows/win32/adschema/a-u... is "Contrôle de compte d’utilisateur" in the list on the left and sorted between "UPN-Suffixes" and "User-Cert" because "Co" sorts between "UP" and "Us" for Microsoft engineers.

Very helpful to find anything!


When it comes to friction in corporations it's typically policy based

It starts off easy to get stuff done, then corporate IT starts to lock down (for various reasons), then those on the edge move to new pastures where they can do work (provision a VM on prem? 50 page form passed through 12 departments. Provision the exact same VM in ec2? 3 clicks)

Eventually corporate IT decides that there needs to be governance and the cycle continues.


Those things are bound to happen as stakes increase, the other extreme is where people test in production.

I've done testing in production, it's usually pretty safe as long as you have a dedicated corner for it. It really helps migrate to production too because you just assign it to the test devices instead of the production ones. Instead of having to build it all up again in a different environment.

What I also liked about it was knowing what was going on in production. These days it's a complete nightmare getting access and thus misconfigurations don't get noticed until they cause issues. There's nobody looking at it anymore except the operations guys who just follow tickets telling them what to do. Nobody to go like "hey why did we flip option X, that's going to cause problems when..."


Everyone has a Dev system and a Prod system. Some people even have them separate.

It's terrible yes. I'm in that situation. For every engineer there's 5 "governance" bureaucrats. I hate it deeply but I can't really move right now because I need to get a mortgage and having a long-time job is a huge plus at my age. And there's never a back from that. Never anyone who goes like "hey all this governance crap makes us really slow to react, stifles ideas and burns all innovation out of us, let's tone it down a bit".

As an example, our "leadership" like most business leaders these days are demanding we do everything with AI constantly. However if we have an idea to automate something there is a 50-step process of approvals, finance, security, privacy, etc teams. Even for stuff where some of these things don't apply because they fall under a flat rate plan. It takes months to wade through all that crap and often at the end we're told it won't fly because some VP of whatever didn't feel comfortable signing off.

And these are often very simple things. Things like a trouble ticket, checking if all the requirements are mentioned and if not to point this out to the user before an actual person has to do this. Low-risk, low-cost. So what do people do? They keep their ideas to themselves and just keep working as they are. Because it's not worth the hassle.

Companies like this tend to just get stagnant because of their policies. Some of them might have been enacted for good reasons but after a while the process becomes a goal in itself. It becomes a monster so big nobody can handle it anymore.


> You know that little micro PC attached to the cart in every exam room in a doctors office? There is a 99% chance it is joined to AD.

So are the cop cars in my city. Walked by one w/two officers in it in a supermarket parking lot last year and had to explain what a trust relationship issue is, unironically.


I'm not sure LLMs will be good at doing taxes any time soon primarily because they're very bad at even simple math unless it's explained in words like in proofs. A good illustration of this is asking an LLM something like this: "If I put a container on a food scale, tare it, add 100g of peanuts, tare it, weigh 100g of almonds and then remove the container from the scale, what does the scale read?" LLMs completely puke all over themselves on this one. They do not account for not taring after the almonds and they do not see the weight of the container needs to be subtracted. The correct answer is minus (100g + container weight). Although this particular question also exploits another LLM weakness which is physical world concepts. LLM are just guessing. For an LLM to do taxes well, it would need to be trained on many scenarios and it would need a math service that it can use as a support service.


You just need to give an instruction something along the lines of "any initial calculation or derived value must be computed through the use of a $LANG program."

Pretty sure the LLMs could implement tax calculations better than the average accountant, were they given the rules.


In my experience agents will reach for numpy for any non trivial math without being instructed


It's the judgement of "trivial" where I find issues.

Most calculations are trivial.


It is also trivial to specify deterministic tool calling for all math


Exactly. That's why I specify it that way.


As a veteran tax CPA, I think LLMs can be very good at preparing taxes but I would not trust them to complete a tax return start to finish. An LLM that can take data from tax documents and input that data into the correct boxes of tax preparation software has no need for math, it just has to identify data in a pdf, spreadsheet, or handwritten note and transcribe that data to the correct box in professional tax software that then does the math. In your typical local CPA firm, an LLM could replace the unlicensed preparers/interns/early career stage CPAs in filling out the return for the experienced professional to review and adjust according to their specialized knowledge. As with many industries, I think the LLMs are suited at replacing entry level automated turks but lack the skills necessary to fully complete the project.


Look at multiple posters here explaining that their accountants made O($10k) errors that were discovered by LLMs.

I’m not sure the evidence is on the side of the CPAs. Maybe CPAs with LLM assistance.


“The scale reads −(100 g + the weight of the empty container).”

I literally gave it your question. What models did you test?


ChatGPT and Cursor both failed hard. Even after walking them backwards through the flaws in their answers they just refused to see it. When I stated flat out what the answer was ChatGPT agreed in the end. Cursor did not. Cursors initial reply was -200g. Then I challenged "But I did not tare it after adding the almonds." and again it failed with: "That doesn't change the final reading: it still displays -200g.". When I walked it back further with "If I put the container on the food scale, tare it and remove the container, what does the scale read?" it correctly replied "negative the weight of the empty container" but ... when I asked "So does this impact the agents answer to my original question?" it replied "No, it doesn't change the answer—the scale still reads -200g". LLMs are NOT good at even basic numeric accounting.


You keep saying LLMs can’t do this but I get the correct result each time without drama.

Cursor uses Grok so far as I can tell, which is.. not great and I hope you did not try ChatGPT free or 5.6 or Luna.. you cannot unfortunately just say “Claude” or “ChatGPT”. Their lineup is sufficiently complex - each with their own failure modes - to warrant some further specification.

I think you are “directionally” correct, as these clankers like to say, because language models are indeed not known for their arithmetic wizardry. However, as my correct answers show, they can do it just fine.

Size, unfortunately, matters.


Free version of Gemini also gets it right


Yes. It's because you are asking it to generate an image of a pelican riding a bicycle. If someone asked you to draw a pelican riding a bycycle, would you interpret that to mean using 3d photorealism? LLMs follow conventions. The convention for an animal riding a bike is to create a childish 2d line drawing.


Like anything in coding, it depends. LLMs are good at following conventions. If what you're doing has been done a 1000 times before and there is a good bias, it will produce good results. This is why LLMs are so good at the leafy parts and even the branches the leaves are connected to. But there's always something unconventional about a codebase. If there weren't, it wouldn't be worth writing.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: