CISSP

Failed CISSP.

That cert has a 50% fail rate. Study more and take it again. It’s the hardest non-math exam I’ve taken and is a lot more difficult than most expect. I passed it first try but it was very challenging.

But here’s a tip without breaking NDA: The most technical answers on the exam are usually wrong. And that’d because it’s not a test of technical skill, but rather what’s the best overall solution for the security and health of the company itself. In other words you must think like a member of the executive leadership team, not like someone doing sysadmin stuff.

That tip will almost always eliminate 1-2 of the answers right from the jump.

BenQ

I bought this monitor as a second one both for the work machine and my Mac Studio. It’s an unusual 3:2 aspect ratio and so far I’m really liking it; it’s nice to not have everything be so crazy wide as is the fashion these days.

I’m not watching a movie (most of the time). I’m tryna work. It’s great to have it be tall enough to see some stuff. I just wish it were 5K or better yet, 6K. Then it’d be perfect. As it is, it’s just good.

Recommended, though.

AI Hihi

A common mistake I see is that people now believe that all AI is an LLM. However, there are tons of other types of AI (not an exhaustive list): search algorithms, game-playing agents, recommender systems, robotics control, computer vision, classic expert systems, optimization, etc. Many of those precede modern LLMs by 40+ years.

When people hear “AI,” they think ChatGPT. But I had a colleague who was working on AI email spam filtering back in the 1990s. So this stuff has been around a while.

Filt Flit

Those who say AI is not useful just don’t know how to use it. Simple as that. I wrote the below filtering and search improvement tool for our internal search engine’s web interface mostly using AI. Not even an agent — just plain ChatGPT.

None of what you see above save the search box existed prior to me and the AI constructing it. It’s all local CSS and JavaScript, by the way, but it provides advanced filtering and some additional search capabilities to an application that had zero of that before. It even uses local browser storage to remember the state of whether the filter bar is hidden or not (button not pictured). In other words, this isn’t some trivial application. Though they are not the most efficient, there are over 1,000 lines of code that allows the above to function, including some UI improvements not shown in this screenshot. (The “Quote” button toggles the result between quoted search and not, and the filters and chips filter the results returned. There are also dropdowns that do the same. The tool detects file extensions and file types dynamically and counts them.)

And sure, I could’ve written all this myself. With the LLM, I did it in 4-5 hours between its code, my own, and tweaking the outputs. If I’d attempted to it all myself sans LLM assistance I think it would’ve taken me minimum 80 hours and more likely 150-ish. That said, I am not a good programmer; a decent one could’ve probably done it in 20-30 hours from scratch.

In short, this is something extremely useful to me that I probably would not have attempted without LLM assistance. The time investment simply would’ve been too large. This will be true for wide swathes of the economy. Expand in your mind on the possibilities from there.

Git Up

I like open source, but if often spawns absolutely atrocious technology. For instance, Docker, Kubernetes and Git. All of them are poorly-designed and overly complex. None of them are all that well-suited for the use cases to which they are typically applied. Only Git works better than the tech that preceded it, and that’s only because it’s just shitty instead of incredibly shitty. The bar there was so low that anyone with reasonable coding skills and project management capabilities could exceed it.

These three technologies were adopted heavily because of their complexity, rather than in spite of it. That’s what developers are attracted to and that is who — largely — runs the tech world now. When it was more my type in charge (systems people), we gravitated to simpler, better-specified tech that was constrained, well-specified, focused and fast.

Those days are long gone now, and the world is much worse for it.

AI Ayo

MIT report: AI can already replace nearly 12% of the U.S. workforce.

I agree with that assessment. No matter what you’ve read — particularly if you’re on the left and predisposed to hate tech and any form of knowledge advancement that doesn’t relate to a new, innovative gender — the newest AI models are quite good at many things. Certainly they could replace most of the juniors in my own field that I’ve worked with over the years and many of the mid-level folks too.

No one is reckoning with how quickly this fact is being weaponized by the MBA class or that it’s already demonstrably occurring. Predictably, economists are and will be in denial about this at least a decade after it is already happening.

Arms and Harms

This is true. And is not true of US tech companies, no matter what you’ve read. The Chinese tech companies are quite literally arms of the CCP; Microsoft is not an arm of the CIA.

You Mean the Trash Gets Trashed?

Why!?

This seems extremely common for users to do and I’ve wondered about it also over the years. People will store important emails in the “Deleted Items” in Outlook and files in the “Recycle Bin” on Windows and the “Trash” when it used to be called that.

I think because it’s easy to do? Even though it’s dangerous. I assume they are attempting to use these places like an archive. Stupidly, poorly, and with no thought behind it at all, but I guess that’s what is going on. At previous roles I’ve had to make announcements via email and Slack that, “We do not back up the Windows Recycle Bin or Deleted Items in Outlook. Do not store needed or important items in those locations. They will be deleted and will not be available for any recovery or restore.”

It worked better than nothing, but people still did it anyway.

Just Change

Why can’t I “preview message” from a trace anymore? Is it because I’m shitty?

No, Microsoft has broken something completely yet again. Something that used to work rather well. It’s the Mozilla-Google-Microsoft way: take something that was working fine, destroy it, then tell you that the new way is better and that you “just hate change.”

I tried to use this tool last night and just gave up and found another method. I guess we’re in the “everything just gets roundly worse forever” stage of tech now.

Junior Shift

This is a real problem.

The best AI agents are better now than a junior dev or infra person. I’d say that the highest-performing AIs now are about the equal of a dev with 2-4 years of experience and an infra person with 3-5 years of experience.

This fact is nuking those lower-level roles.

A lot easier. You basically have to recapitulate half of computer science to get a junior to do something sensible; no such problem with an AI.

I can now write PowerShell scripts a lot faster with AI help. But I already knew how to write PowerShell, so it’s a lot easier for me to prompt the AI and then I can use my prior knowledge to evaluate what the script will achieve, so I can be certain it won’t blow anything up. Juniors have no idea so AI just makes them worse and more dangerous because they both cannot prompt the AI correctly and are unable to determine what the results will do.

Having no junior to senior pipeline is going to be a huge issue in 10 or so years, though. Not sure what we’re going to do about that.