corytheboyd

FWIW sounds like a great use case for some rules files. I’ve only worked with Cursor and Roo but they both support it.

This of course only works for the “stop recommending X” part of your problem, but maybe something like the project here helps fill in with up-to-date documentation too?

Both Cursor and Roo also support URL context additions, which downloads the page and converts it to a machine readable format to include in context. I throw documentation links into my tasks with that all the time, which works out because I know that I am going to be sanity checking generated code against documentation anyway.

I use browser bookmarks organized into folders like /dev/blog /dev/tools etc. so that I can browse by category later when I think “what was that tool again?” A flat list of bookmarks drives me insane, all I want to do is delete it.

I tried Obsidian for a while but I’m too prone to bike shedding with it, so I gave up. I DM myself in Slack with bits of info to keep around for the near term, because it’s already a simple feed ordered by time.

The point isn’t about saving yourself from the layoff, it’s about having a network to help you find a new job, because yeah, the layoffs are disconnected, heartless, and clumsy. If you’re always building at least a couple relationships, some of those people inevitably quit on their own, branch out, and after years of this, you end up with a decently sized network to help you out (and you can help them too when the time comes)

> I'm using Ruby for a startup because it's 10x more productive for building a webapp versus a compiled language

I’ve been a Ruby/Rails dev for about 14 years. Just started working on Java projects at work and… holy shit, it slows down the entire organization and they don’t even seem to know how much of it is because of the terrible immovable object that is the mountain of Java code they made. Before this experience I had always assumed that rigid types alone would be a boon… oh how wrong I was. The rigidity just leads to people writing 10x more garbage to get simple things done, which causes a 10x worse maintainability problem. The rate at which I hear “oh lol just another NPE” is astounding, how is this constant occurrence not interpreted as a complete failure of the “type” system????

Maybe a crack team of Java Experts wouldn’t do this, but the reality of most software companies is that people of ALL skill levels are writing code. Poorly written Ruby is a whole lot easier to figure out (as an expert of Ruby) than 10x the amount of code in Java. I hate this.

Huh yeah I guess I never thought of it as not a rhythm game. At higher difficulty it’s really more of a DANCING game, if you think about what you’re doing with your body. I think the observation about the scoring system backs this up too, because it’s based on direction and speed of motion, not timing. Of course you need to have some timing mastery to hit the blocks, but the scoring is not based on it.

It’s been so long I don’t remember if they explain the scoring system fully in a tutorial… but it becomes very intuitive if you _flow_ the direction the blocks want you to go in, with liberal follow through. Also, this makes way more sense at higher difficulties, because the lower ones are so slow that flow and follow through just isn’t intuitive.

I’m not at all interested in high score min/max, but I do confidently play most “dancey” expert+ (as in, not the ridiculous Camellia stuff), and this is how I think about the game.

Whoa this looks really cool! I love how you made something physically 3D to stand out in a world full of 2D knobs and sliders… but it still has 2D sliders because those work the best for dialing things in with precision.

I’ve tried it, but it didn’t seem to have enough structure log support, so I gave up. I just want a way to say “this is a jsonl stream” and have it do the right thing, which is parse out the json paths of each line, and let me filter by those, or add columns for those to the viewer table. It’s not trivial to do this, but that’s exactly why it needs to be done.

Damn, if you took out the “Kubernetes” part, and made it generalized, it looks like you built something that I have wanted to see for a long time. I think log explorers work best as a GUI, and that they need deep integration with structured logs. Basically I just want the DataDog log explorer but locally, and able to simply intake from some files. Some have tried, but they are always too simple, not parsing out properties of structured logs and building good filtering on top of that. I think setting up Grafana/Loki/Whatever is way too heavy for such a simple ask.

Anyway all that is anecdotal, what you made here is really cool!

Yeah, that feels more right, and feels like a problem that is only going to get worse.

How do weekly 1:1 meetings with a manager not catch this very quickly? Okay, maybe the original suave interviewer comes back for those… Still feels like a good EM would pick up on discrepancies between work done and how the suave person talks about it.

It depresses me, but you’re probably right about in-office work being the only guarantee against this type of scam. I wish we could just have nice things.

> yet fake people are getting hired left and right.

Hate to be that person, but what are you reading that makes you think this is true?

Agree that the article is pretty dumb though, especially the OSINT and Crypto “don’t trust, verify” comments. Feels like content marketing that didn’t really hit.

> […] push these tool-use mandates without actually quantifying the impact LLM tools have on productivity

The way I have seen this “measured” is by asking (demanding) people to pull some “time saved” number out of their ass. That number is then taken as fact, without question. So the measurable productivity gains are all based on coerced people making things up, omitting instances of the LLM tools slowing things down. It’s a house of cards, and it’s going to fall after a few more months of empty promises without results. “You claimed 300% more productivity, but delivered the same amount of work, and took on this massive AI bill. What the fuck are you doing?”

LLMs are cool, but they aren’t magic. This shit is exhausting, just skip to the part where you fire a bunch a software engineers, because that’s clearly what you actually want to do :/

OBS can always get you there, but you will be doing the scene work on your own.

I guess the point is that VSCode won because it collected critical mass (despite people practically chanting Microsoft’s “embrace, extend, extinguish” game plan). If a sort of exodus were to happen, perhaps that same critical mass would stoke the flames of mass plugin catch-up.

I have no skin in the game, I’m just a guy in the internet back alley throwing down a couple bucks on a dice game.

Shoot yeah I did haha

> bring the ads back sooner or later, even if you pay for the service.

This has already happened for subscription TV services. Your previously ad-free subscription now has ads, but you can get rid of them again by upgrading! It’s fucking gross. It’s also of course just going to work, and become the new normal.

Eh MS wasn’t going to just let VSCode derivatives soak up all the AI gold rush money, these companies knew the risks. I wonder what it’s going to mean for projects like Zig, a migration of VSCode refugees could crank things up to 11 pretty quick.

I think they are talking about products like Cursor.

I think your intuition about this all is right. Maybe we’re holding it wrong, maybe it’s plateaued and won’t get better, or maybe it will get massively better. Whatever happens, I think it’s the right call to hold a sober opinion— LLMs are just another dumb, expendable tool.

If a tool does not consistently produce results, you HAVE to take that at face value. You can’t just remove the numbers bringing down the average and say you have reached 95% success. When you see polar opposite experiences from so many people, the only reasonable takeaway is “so it’s unpredictable, very hard to use, or both.”

I’ve grown weary of people repeating absolutisms they hear online, too.

As is usually the truth in practice, it’s a mess, which is why I’ve seen combinations of upfront planning and code spiking work the best.

An upfront plan ensures you can at least talk about it with words, and maybe you’ll find obvious flaws or great insights when you share your plan with others. Please, for the love of god, don’t ruin it with word vomit. Don’t clutter it with long descriptions of what a load balancer is. Get to the point. Be honest about weaknesses, defend strengths.

Because enterprise corporate code is a minefield of trash, you just have to suck it up and go figure out where the mines are. I’ve heard so many complaints “but this isn’t right! It’s bad code! How am I supposed to design around BAD code!” I’ll tell you how, you find the bad parts, and deal with them like a professional. It’s annoying and slow and awful, but it needs doing, and you know it.

By not doing the planning, you run the risk of building a whole thing, only to be told “well, this is nice, but you could have just done X in half the time.” By not doing the coding, you risk blowing up your timeline on some obvious unknown that could have been found in five minutes.

I agree, they are very much exaggerating both time spent writing code, as well as the amount of time LLMs shave off. LLM coding very much does NOT take “near zero time,” I would argue that sometimes it can take the same amount of time or longer, compared against simply knowing what you are doing, with tools you know how to use, referring to documentation you know how to interpret, understanding the systems around, the business around, what team A and C need in two quarters so we better keep it in mind… etc.

I’ll do whatever shit the industry wants me to do, I don’t particularly care if it’s dumb. I mean, it doesn’t FEEL great to work on dumb things, but at the end of the day, I’m around to help implement whatever the paycheck writer wants to see. I genuinely don’t mean that negatively either, I feel like I’m just describing… employment?

Software just isn’t a core part of my identity. I like building it, I like most of the other people who write it, and I like most/some of the people paying me to build it. When I’m done for the day, I very much stop thinking about it (not counting shower thoughts and whatnot on deeper problems)

So what if I end up fixing slop code from AI hype in a couple years? I have been cleaning up slop code from other people for 15 years. I am painfully aware of slop I left for others to deal with too (sorry).

So yeah anyway, your comment resonated. Hype is annoying, but if it sticks around and becomes dominant, my point is, whatever, okay, new thing to learn.

I’m sure there’s a flashy way to say it, but yours reminds me of this one:

> Only ask for permission if you want to be told “no”

Guessing the Go experience is largely from faktory https://github.com/contribsys/faktory

But I dunno you can just go ask him about it, assuming he still does those Friday morning “happy hour” calls :)

I never thought about it until now, but it’s insane that the companies who offer both LLM products and cloud compute services are double dipping— they get the LLM product to sell, as well as the elevated load egress (and compute, etc.) money. When you look at it that way, where’s the incentive to even care about inefficient LLM scraping, leaving it terrible makes you money from your other empire, cloud storage egress costs.

Facebook gave you access to friends of friends, which was huge when EVERYONE was on it. That coupled with the Events system was pretty awesome. I don’t know if WhatsApp does this TBH, but it’s something you don’t get with typical messengers.

I think the whole custom sort DSL is swinging the pendulum back a bit too hard. People just want a deterministic sort, with no injected ads or hate-bait, from exactly the people they follow, with the ability to apply basic filters on top of that. You know, the old default, the one that was designed to be useful, not to extract maximum ad dollars from people’s eyeballs with vacuum tubes.

> Hammers are NOT designed with a specific purpose, they are just big heavy metal things with a handle for leverage.

Well they’re most certainly not the right tool to fasten screws with :P

I know my example is very contrived, I wasn’t trying very hard, just went with the first thing that came to mind.

> LLMs are much more versatile and capable of performing a wide range of tasks. Yet, at the same time, their capabilities are ill defined.

That’s my point, I want to skip to the part where we know what LLMs are good for, what they are bad for, and just consider them another tool at our disposal. We’re still in the phase of throwing shit at the wall to see what sticks, and it is exhausting more often than not.