Director of Technology for Refund Retriever. Fancy title, but I'm just the Chief Nerd that runs code, architecture, ops, etc... Used most languages the serve up web traffic since 1999, but mostly Ruby, Javascript, and ColdFusion/CFML.
Whatever opinions I express here are mine alone, and don't reflect those of my employer.
My HN username is how you'll find me pretty much everywhere else.
I think it's the other way around. They're not talking about CEO ability, they're talking about CEO performance, which aren't the same. Money follows performance (based on outcomes, not actual talent). It's like professional sports - win a championship, get a big payday, regardless of how it happened.
The Grok ones are amusing, almost comically bad. However, whenever I've tried to pass an image creation request to any of the Opus models, it's been far worse, like first week of using Microsoft Paint bad (while ChatGPT would create social media quality images using the same prompts)
Possibly, but aren't the tech giants positioned to win a war of attrition? Then again, they're more likely to sit on that cash and wait for the opportune time to buy a frontier lab.
I also bought a pellet extruder kit, which can use shredded PLA waste, but haven't gotten around to putting it on a printer. (https://greenboy3d.de/)
I've had pretty good luck taking waste and melting it into silicone molds. Doable with a cheap toaster oven and purge waste and other scraps (though you need to break down larger waste)
For those who want to recycle filament without doing the work themselves, this person has such a project (and maybe there are others out there): https://erikaprintsshop.wordpress.com/
> I could not remember whether the machine had gotten even slower, or whether it had always been like this and I had once thought even this was fast.
In addition to the usual suspects (aging hardware, planned obsolescence, bad memories), I wonder how much CPU is burned by software trying to talk to long dead backends and the retry loops and errors that occur as a result.
A quick check (ie, letting ChatGPT do the math) shows that if all caps on taxation was eliminated, it would help, but not solve the problem completely.
Of course, but that's the thing about "informal" agreements: presumably there's no record of it. Price collusion happens way more than most realize. I was at a trade show for my employer when a competitor walked up to our booth and out loud proposed that we should all fix our prices. We didn't of course, but the brazenness of saying something like that out in the open like was what surprised me.
I frequently blur the line between ad-hoc DSL and pseudocode, and just hand it off to the LLM. I want to get the thoughts out of my head as fast as possible, using whatever structure makes sense to me. Even if you know all of the code to be written, I think this is a huge win with LLMs, where your intent is more important than syntax.
I don't know if I'm unique in this, but I find myself asking people I know about things that I know I can easily lookup faster. At this point it's more a social ritual than actual information gathering.
As my diabetes has progressed, I find myself sleeping more odd hours (it can be hard to fight off the tiredness that comes after a meal), and I can be frequently woken up my extreme lows where my body is screaming for carbs.
There's truth in this, git is probably the least of anyone's worries when it comes to understanding your code if you're pulling in any external dependencies in your codebase. For a typical web app these days, you're hundreds of thousands of lines deep in code written by someone else on day 0.
> Put two machines on a desk, each about $2,000. One is a tower with an NVIDIA RTX 5090: 32GB of the fastest consumer memory ever shipped, 1,792 GB/s. The other is a mini PC the size of a paperback, an AMD Ryzen AI Max+ 395 "Strix Halo" box with 128GB of soldered memory at roughly 256 GB/s.
Doesn't change the conclusions of the article, but each of those machines is more like $4k+
If you're referring to the markup charged by OpenRouter, you can use harnesses like OpenClaw/Hermes without it and go direct like you're saying. If you're talking about actually "routing", then you don't get that out of the box. However, the popular use of those harnesses doesn't often use the smart routing approach with a single agent. Instead, the approach is to create multiple agents, each with a role and a model tailored to that role by cost and functionality.
Correct. You do benefit from some headroom for things like launching browsers etc but refurbs or mini PCs (with at least 16gb; ideally 32gb of memory) from the likes of Minisforum or GMKTec work well enough if you're wanting to spend a little bit of money.
I have no issue paying for subscriptions, and I pay for many. However, usage-based can quickly outpace value. For example, we do a lot of background tasks that integrate with third-parties. A small change over something we have no control over can explode the number of events sent to a bug tracker, even though event 100,000 is no more valuable than event 1. Yet at a certain size, you're paying hundreds of dollars a month to track those 99,999 extraneous events. The common answer is configuration, but honestly, it's easy enough to just spin up one of the many self-hostable options like GlitchTip and call it a day.