That part is confusing because it's not like they provide an example of how default GPT-5.6 output compares with GPT-5.5 both with default output and prompted for brevity. Whenever I use such prompts, it's usually because I want the model to give me the gist in a few sentences. I'd be stunned if GPT-5.6 was that concise by default. I would think that could "break" a lot of things for developers who didn't know to make prompt changes after upgrading to 5.6. What if you were expecting GPT to be as wordy as it usually is? Then suddenly your output is not wordy enough?
Smells like OpenAI trying its best to stave off financial armageddon for another few months. Then again, I'm not sure why they chose to waste so much output computation on verbal diarrhea all this time up to now.
There's a lot I don't like about Nintendo, but the one thing I admire about them is they understand that fancy cinematic graphics aren't what make a great game.
Organic Maps a great app in many ways, but I still don't get how people can actually use it every day and say it replaces Google Maps when its search feature totally stinks. I know it's a hard problem, but this is the number one thing that needs to somehow be fixed. I can't tell if I'm just too dumb or if FOSS/degoogle fanboys are just pretending. I just know I've tried to use it exclusively many times and always had to give in to Google Maps because the search totally failed.
> No! Don't install that lodash thing without explicit approval from IT. Oh, you want a license for Charles Proxy? Gee, I dunno... we've got a budget to maintain.
Employers in 2023:
> No! You can't use ChatGPT at work – it's a security risk.
Employers in 2024:
> Okay, you can use Github Copilot I guess, but you'll have to endure boring corporate training on what you're allowed to do with it.
Employers with dollar signs in their eyes in 2025:
> We attended a seminar about vibe coding. Why aren't you dumbasses keeping up with the times? Use Claude Code for everything! Don't write any of your own code anymore. We don't even really care if you use yolo mode. Just review code and push 10x more features! Use unlimited tokens! Money printer go brrrrr.
Employers in 2026:
> You mean giving one or two companies full autonomous access to our workstations while stupifying our engineers wasn't a sound business plan?
I've definitely realized this phenomenon after a few occasions of erroneously trying to rely purely on instructions to get an LLM to do a thing or take on a role, especially without persistent cloud-based sessions that have internal checklists and other opaque guidance. They're essentially poor at self-managing, but can do really well when they are limited in scope/context and are worked into a sort of state machine that guarantees they perform certain tasks predictably. They won't always do those tasks the exact way you expect them to, but at least they actually do them, and because of that they are more likely to have the correct prior context to perform the next task better. Because they are so prone to selectively ignoring directions, that can quickly send them down an incorrect path that compounds on itself.
It's also a kind of funny thing how HN has the attitude of "never implement your own encrypted anything" but then demand their apps build in e2e encryption. It may be one abstraction higher, but it's still fundamentally the same problem. With the unfortunate exception of web browsers, if I'm going to use something that performs encryption, then I want encryption to be the only job it has.
I'm not sure I will ever upgrade Immich again. When I upgraded to the next minor version (I forget which off the top of my head), the data migration corrupted my database such that no images would be served. Fortunately I had the old database backed up, so I restored it, rolled back to the older version of Immich, and things were back to normal. I like Immich, but this is not good for software that's beyond the first major version, and also handles archiving people's personal data.
> People want to make it seem like you need to always use the latest and greatest frontier models to be taken seriously as a developer.
Except you kinda do. Try getting a job today without mentioning Claude experience. In another year it'll probably be something else. Saying you like to use Copilot today makes one seem elderly.
Not saying you need frontier models on a technical basis, but for career PR you probably do.
Leadership at Google should face prison time for this sort of practice. We wouldn't accept it in the physical space, so why do we accept it cyber space?
What the author says about ego goes both ways. People often reject arguments because of ego. Arguments can imply that they way someone has been doing something is suboptimal or even flat out wrong, or at least that's how they may be perceived. Even if something you're arguing for can improve the situation, the other parties may refuse to give it a chance because they need to protect their egos.
At some point, people have to introduce ideas into a broader consciousness, even if they clash with other ideas. How else will anything actually get done? Putting forth an argument doesn't necessarily have to come from the ego. Even if one does come from the ego, that doesn't mean the idea itself is bad.
I've mostly stopped trying to argue or debate on any topic because the probability of being chronically misunderstood usually outweighs any benefit that would come from successfully persuading the other person. I'm never convinced that I'm 100% right on anything, and life is too short to spend it arguing with those who do; which describes a lot of people.
The other reason I rarely argue anymore is that, if I am correct on something, reality usually proves that I was. That doesn't mean everyone else is gonna say "Ravenstine was actually right", because they never do, but at least I get the satisfaction of having been able to trust myself.
I love this! I'm kind of sad that I'm likely beyond the point where I can ever have kids, but what you describe are absolutely the kind of things I'd want to provide them if I'm lucky.
One recommendation I have is a basic 3D printer and OpenSCAD installed on the family laptop. I can see that opening up a lot of added interactivity with other things like the Legos, robotics, etc.
All these things are apparently valued at trillions of dollars these days. Where's the trillions, or hundreds of billions worth in improved quality of life? What has gotten better other than the ability to produce more crap?
Plus I've found "legit" to be a moving goal post. One day a show is on one platform, the next it's on another, or it becomes unavailable except for [insert random foreign country here]. Even HD is a ripoff sometimes when half the episode comes in all compressed looking. They'll blame my bandwidth except I have no problem streaming an episode without adaptive compression over Bittorrent.
People can say what they want about piracy, but it continues to be what I consider a necessity against culturally important media being further tainted by rent seekers looking to make another buck in any way they can.
I'm sure most people wouldn't outright say that, but I have no doubt that many corporate leadership types think that way. After all, the C-suite has higher incidences of NPD and ASPD so it shouldn't be surprising. Along with being out of touch, some of them probably think their peons should have the means to do the same.
Whenever companies make statements like this and then people act surprised when they backtrack, I can't help but think of a bit of my favorite dialogue from Star Trek Enterprise.
HARRIS: We had an arrangement!
KRELL: You did what I wanted. I don't need you anymore.
HARRIS: You agreed that both our governments would benefit if the two of us worked together.
> there don’t appear to be downsides for overdoing it
There may or may not be a downside depending on what one considers a downside.
In one of my other comments I just made in this thread, I mentioned my experience taking relatively large doses of 20g a day. While I found it has cognitive benefits, it did interfere with my sleep, though not catastrophically. If a person happens to enjoy sleep, then it's probably best they stick to 5 or 10 grams. On the other hand, if you need to pull an all-nighter, the sleep interference (as well as the better recovery the following day) may not be seen as a downside but beneficial.
But yeah, from a toxicological perspective, creatine does seem very safe even at those doses.
Funny because I'm not sure I quite notice it immediately when I stop taking it, presumably since the body retains a certain amount of it. But I definitely notice a difference when taking it the first day after not having had it a little while, especially when taking 20+ grams a day. It gives me so much mental energy and alertness that I won't get a whole lot of sleep, but the next day I end up feeling just fine. If I remember correctly, there have been at least 2 studies that suggested high doses of creatine reduce the side effects of suboptimal sleep, and that definitely seems to be the case in my experience.
That part is confusing because it's not like they provide an example of how default GPT-5.6 output compares with GPT-5.5 both with default output and prompted for brevity. Whenever I use such prompts, it's usually because I want the model to give me the gist in a few sentences. I'd be stunned if GPT-5.6 was that concise by default. I would think that could "break" a lot of things for developers who didn't know to make prompt changes after upgrading to 5.6. What if you were expecting GPT to be as wordy as it usually is? Then suddenly your output is not wordy enough?
Smells like OpenAI trying its best to stave off financial armageddon for another few months. Then again, I'm not sure why they chose to waste so much output computation on verbal diarrhea all this time up to now.