Complaints database
29,255 raw documents, unedited and searchable. This is what everything else on the site is built from — nothing here is generated.
Camera Chase Vehicle
Thanks for the kind words! Nimbus was a really fun project, just spent a bit re-formatting previous writeups to be more modern-web compliant. The one-wheel was indeed a weirdly synchronous release, I had no idea it would be as wildly successful as it became.
Kimi K3, and what we can still learn from the pelican benchmark
Maybe it gets posted every time because besides a personal believe by the person popularising this "benchmark", there is no reason to assume that certain labs aren't intentionally training to game this and every other lab at least unintentionally gets improvements for this specific combination of animal and action bec…
Kimi K3, and what we can still learn from the pelican benchmark
Respectfully, did you? The comment was specific to doubting the believe simonw has that labs are not training [0] specifically for this task, which is exactly what simonw wrote in the post [1], that it is a believe of his that they don't. He did not mention any kind of evidence or any piece of information that would i…
A Road to Lisp: Which Lisp
Lisp interpreters are easy to write; a performant and standard compliant Common Lisp system is considerably harder to write.
Workspaces – Explore the workspaces of modern creators
I love the idea! But echoing some comments on here, oh man if the desk-owner takes the desk-photo, you get a lot of branded mugs with the logo facing towards the camera, books neatly stacked in a pile right in the middle of the desk, or (my personal favourite) the iPhone place purposefully upside down centred on the u…
Kimi K3, and what we can still learn from the pelican benchmark
Simon has stated a few times that he knows it’s possible that pelicans could be in the training sets. He also has other tests he doesn’t share publicly. He’s just a fan of pelicans.
How Has Roman Concrete Lasted for Millennia? 1,900-Year-Old Latrine Offers Clues
Just wish I could claim credit. Decoy plate is essential. And since nobody complained it sounds wasteful, I suppose it should be able to be just scraps. May your next camping trip or picnic be wasp free!
Claude Code: Anatomy of a Misfeature
My memory is that once you begin using the alternate screen, you need to turn on mouse reporting, and then you lose native copy. Different terminal emulators also act totally differently when in this state.
Schema Harness Achieves ~99% on Arc‑AGI‑3 Public
Sorry, which definition do you mean? The problem with "reasoning" is that for the most part it happens outside of training e.g. as CoT. The base model knows what it knows, it knows what it's trained on, and it can't really go much farther than that. Now, starting with o3, "reasoning" models are probably (who knows exa…
After wearing the Pebble Time 2 for two weeks, I'll never buy another smartwatch
You say "digital" rather than "smart". Digital or analog, a basic watch has that single advantage of being mounted on the wrist. It is always there, supporting a quick glance when you don't have time, free hands, or social context to pull a phone out. Once you add vibration, now you open a class of discreet alerts. Ev…
AWS: Inaccurate Estimated Billing Data – $1.7 billion
Just today I gave my local agent a CSV which listed a bunch files with of human readable size units and asked it to count rows in each GB range. Sounds simple enough but it completely miscalculated, because it parsed MB as GB for some reason. In hindsight it would've be quicker just to do it in Excel or something.
A Road to Lisp: Which Lisp
> I find a procedural style of programming so much easier to reason about, both when writing and reading. Then do that. There's nothing stopping you from using pretty much any style of programming that you like. Or mix and match. Or evolve over time. Loops, lists, arrays, structures. Simple iteration: dotimes, dolist,…
AWS: Inaccurate Estimated Billing Data – $1.7 billion
A couple of my coworkers think I’m nuts for watching cost explorer so closely but 1. The time it takes to look and notice costs that don’t make sense easily pays for itself, and then some (in my experience). I doubt you spent $7k of your time tracking this down, and you probably noticed optimization opportunities that…
Linus Torvalds on LLM usage in kernel development
Overall, sensible take. I think Linus' position even pre-LLM-coding – basically not trusting any code he gets, and not writing any code himself, just reviewing a ton of it – is strikingly similar to how most people now do "software engineering" (or vibe coding/agentic engineering or whatever you want to call what…
Evidence of inconsistencies in evaluation process and selection of winners
I get my team to review my code, AI generated or not. Goes through the same process as anything they produce. Two pairs of eyeballs on everything.
The LLM Critics Are Right. I Use LLMs Anyway
>> I am a little bit scared to say this too: last month I spent almost 10k USD on tokens. It sounds so insane. No it just sounds completely stupid. You spent $10k to get a program to write text that you can write yourself just as well and for free. "Dissonance" doesn't even begin to describe it. It's like suddenly eve…
The human-in-the-loop is tired
> ...but the human reviewing, directing, and course-correcting feels worse, not better. Has this person ever been a design or tech-lead for a team? I'm still tired and I've moved on from that role over 5 years ago. Gathering requirements, designing a solution, implementing as tickets, handing off to developers, stayin…
First atmosphere found on Earth-like planet in habitable zone of distant star
Probably more likely that we work out how to fold spacetime than we get there in anything like a high enough percentage of the speed of light - the fastest object we ever made travelled at something like ~0.064% * C so we are looking at ~750 years with current technology and presumably we'd need to switch on the probe…
Evidence of inconsistencies in evaluation process and selection of winners
Whoa, 200k lines is a lot. Without knowing the details I wouldn't want to pass judgment, but that would be a flag to say the least. I don't know how he is reading all that code before pushing... There's a danger in it too because (whether it should or not) CTO title carries a level of weight that might result in less…
British Steel taken into public ownership to protect 'vital' UK supply
> What kind of expertise should you possess to be able to comment on a lab theory? Enough. > Even if lab leak is true, what is the proposal, that we do not study viruses at all? People love to think in black and white. All on or all off. It is metabolically inexpensive to reduce reality to the binary. But reality is a…
An Engineer's Guide to USB Typе-С (2024)
It was only 2.5 years between Apple's first device with Lightning (iPhone 5 in September 2012) and their first device with USB-C (12" Macbook in April 2015). It's not like Lightning had a huge track record at that point they just made a decision that they had switched connectors once and didn't want to turn around and…
Minikotlin
I built something similar, but I managed to compile the Dart VM, its compiler and the static analyzer to wasm using emscripten: - repo: https://github.com/modulovalue/dart-live - demo: https://modulovalue.com/dart-live/ It's on my todo list to support compiling dart code through…
Evidence of inconsistencies in evaluation process and selection of winners
And in B2B or B2G you have customers with checklists of features they don't actually need because the person who wrote the checklist is friends with or paid by a particular supplier, or just operating on old information. That's how you get Windows subsystem for POSIX. Someone in the government had a checklist saying t…
The LLM Critics Are Right. I Use LLMs Anyway
So much this. When coding I use my LLM to hone the code to perfection before sending it to review, including running AI reviews on it multiple times and fixing obvious bugs so that reviewers have less work to do. I also test all my code manually (LLM can be helpful by giving me a list of test scenarios to consider). W…
'Likweli': A new monkey species discovered in the Congo Basin
Thanks, your noun/process suggestion is interesting. Perhaps that is indeed a good way of characterizing what happened in biology. It's clear that biology has stumbled into a linguistic/philosophical trap; if it were actually science they wouldn't have been "working on it" since Darwin! I'll try to find time…
Apple targets dozens of OpenAI employees with legal letters
Makes you wonder if they’ll settle for bargain-basement token prices for Apple Intelligence.
The LLM Critics Are Right. I Use LLMs Anyway
How LLM usage affects your thinking process entirely depends on the workflow you use with them. I can honestly say that while using LLM my thoughts are as sharp and as fast as they were before, if not sharper and faster. But my workflow is that I watch the LLM as it rips, have verbose output turned on so I also read C…
Evidence of inconsistencies in evaluation process and selection of winners
I would have agreed before seeing Co-Pilot (I was extremely skeptical about its usefulness), but after seeing results, I was wrong. It's actually pretty damn good at code reviewing PRs even when the PR was made entirely by AI. It doesn't seem like it should work, but it does
The human-in-the-loop is tired
I just made chat gpt list all the key indications that this article was LLM generated. I'll spare you the details, it's a very long list. It's actually possible to counter some of this stuff with skills and guard rails. The author clearly has not mastered how to do that and sound more authentic. It's all a bit stunted…
Kimi K3: Open Frontier Intelligence
In his speech https://www.xinhuanet.com/politics/leaders/20260717/72728b6f... Xi Jinping mentioned "open source" (开源) exactly once in a list of buzzwords. Deriving from that an "officially stated national strategy" that ensures the free beer won't stop flowing is motivated reasoning. The…