tompark

I've been in too many conversations where this topic comes up, and it's very disheartening to me. Gamers insist there are plenty of great narrative games, and every example they give is basically a branching story with bunch of flags that gate which branches can be taken. If I give the Holodeck as a counter-example, well that's just too pie-in-the-sky.

These conversations remind me a lot of Paul Graham's Blub Paradox: "Blub is good enough for him, because he thinks in Blub." Current SotA narrative games are good enough for most gamers, because all they've played are branching story games.

I first attended CGDC in 1994. It was two years after his Dragon Speech (which I knew nothing about), but the highlight of the conference was his talk about the challenges of story-based games. The part I remember is how he modeled stories as branching trees with fan-out, foldback, tree-of-death, etc (he covers this in the "Architecture" chapter of this book "The Art of Interactive Design"). I didn't really follow Crawford's work on Erasmatron or later, but by the late 90's it sounded like his story model had changed from a tree to a graph network structure, like a finite state machine. While it was an improvement, I was a bit skeptical that this model was enough. Nevertheless, I spent a lot of time thinking about the problem. You see, he'd already infected me with his dream of interactive storytelling.

By the time I moved to California and took a job at EA/Maxis on The Sims 2.0 team, I had decided that true interactive storytelling (as I saw it) was not possible until game AI was sophisticated enough to enable autonomous NPC chatbots. So I put that dream aside while I pursued a career in software development.

Here we are, over a quarter of a century later, and that AI technology is here now. For those of us who have been waiting for this moment, it is almost miraculous. It might be the end of an era for Chris Crawford, but it is just the beginning of the AI-based interactive storytelling era.

This sounds like that Yogi Berra-ism, "No one comes here anymore, it's too crowded."

I suppose well over half of the people who read HN are too young to remember the dotcom craze. Everyone had every scrap of money tied up in tech stocks. IMO, the hype over AI is relatively small compared to other hype cycles. The goofiest part was the endless predictions about the singularity, and "you don't know what exponential growth looks like, man!". I mean, it can still happen, but for a while that's what AI was all about.

There's a larger point here that *debugging methodology* in general is a key skill that many programmers seem to overlook or take for granted, and even think they are good at it while making wasteful mistakes -- real Dunning-Kruger material here.

* I can't tell you the number of times I've seen developers get a hunch and waste too much time going down a rabbit hole, instead of devising a test to attempt to disprove the hypothesis before plowing into a huge code review and seeing ghosts in every shadow of the codebase.

* Print statements for debugging aren't necessarily always the wrong tool, but the mindset should be that they are code and can have bugs too. Good debug logging should be part of the code telemetry infrastructure, so you can toggle it on/off. If it's going to be written and deleted, then use a debugger instead.

Actually this post is getting long so I'll stop now, but I could go on for several paragraphs.

Oh wait, I promised a debugger story. One time I started a new job, they assigned me a bug that other developers had spent a lot of time on but never resolved. It was a random crash, in a game, that rarely happened in less than 12 hours of continuous runtime, typically after 18 hours. The crash occurred even when the game was idle with no player input. The crash stack trace was very different in many instances, so the failure was happening some time/distance after its cause. My first pass was to devise tests to eliminate subsystems: try to repro the crash after disabling all graphics/animation, another test that disabled all audio/input, etc. That led to narrowing it down to a problem with the music. Telemetry showed no memory leaks, it was not running out of heap. The music playback was in a thread, double-buffered. I added a bunch of conditional breakpoints that were based on assert-like conditionals. One of those caught the interrupt happening in between two statements that should have been in a critical section. Instead of using a mutex around the whole thing, I just switched the double buffer logic to a queue, which resolved the problem.

Linux is fine, it's similar enough to macOS (and Windows, for that matter). For my tastes, all the OSes require some tweaking and compromise. I don't understand the fascination with tiling window mgmt, and disable it everywhere (was surprised it got added to macOS Sequoia). I love having different wallpapers on virtual desktops. Etc etc, there are all sorts of different setup options are available on one platform but not another.

It turns out that my stumbling block is the trackpad and keyboard -- I know this issue is highly specific to each individual, so you may not care about this. I haven't tried them on a Framework laptop but they're both horrible IMO on the Thinkpad. (Other people tell me that I'm crazy and they love the keyboard, so YMMV.) The trackpad feels like plastic crap and responsiveness is subpar. Causing a right-click by pressing slightly right of center is annoying. The page up/down keys in the inverted-T pad get in the way. Even on Mac, I didn't like the era of full size left-right arrows, and bought a couple spare old Mac keyboards that don't have those keys. Thankfully Apple fixed that in more recent designs.

Since I code all day, I'm constantly using the key sequences for copy/paste and search/replace. Maybe with time you can switch your muscle memory for these things, but I think it's simple to do these things on a Mac keyboard and not quite as easy (e.g. ctrl/shift-insert) on Linux/Windows.

However, the issues that you'd run into are most likely going to be different than the ones I did. You're probably just going to have to try it and find out if you like it or not.

I beefed up the 'kill sticky' one <https://news.ycombinator.com/item?id=32998091> but that's not very interesting. Here are some other ones:

* Re-enable zoom on iPhone (for pages that disable it)

    javascript:document.querySelector('meta%5Bname=viewport%5D').setAttribute('content','width=device-width,initial-scale=1.0,maximum-scale=10.0,user-scalable=1');

    javascript:(window.openPageWithoutScripts=async%20function()%7Bconst%20resp=await%20fetch(window.location.href);const%20text=await%20resp.text();const%20doc=new%20DOMParser().parseFromString(text,'text%2Fhtml');doc.querySelectorAll('script').forEach(script=%3Escript.remove());const%20w=window.open();w.document.head.innerHTML=doc.head.innerHTML;w.document.body.innerHTML=doc.body.innerHTML;%7D)(); 

    javascript:(function()%7Bvar%20a=window.open('about:blank').document;a.write('%3C!DOCTYPE%20html%3E%3Chtml%3E%3Chead%3E%3Ctitle%3ESource%20of%20'+location.href+'%3C/title%3E%3Cmeta%20name=%22viewport%22%20content=%22width=device-width%22%20/%3E%3C/head%3E%3Cbody%3E%3C/body%3E%3C/html%3E');a.close();var%20b=a.body.appendChild(a.createElement('pre'));b.style.overflow='auto';b.style.whiteSpace='pre-wrap';b.appendChild(a.createTextNode(document.documentElement.innerHTML))%7D)();

    javascript:(function()%7Bdocument.oncontextmenu=null;window.oncontextmenu=null;%7D)();

    javascript:(function()%7B;document.onselectstart=null;document.onselectchange=null;document.ondragstart=null;document.onmousedown=null;window.ontouchstart=null;window.ontouchend=null;%7D)();

Most devs who like AI-coding seem to think AI code-completion is more efficient than chatting with a LLM. Yes, it's true that code-completion is *faster*. But I think chatting with a LLM is more effective.

I'm not going to go into specifics about it now, but maybe if you give it some thought you'll realize the difference. When you're coding you don't always convey your intent to the AI, so it's important to add context with comments. Most coders are too lazy to do that.

Chatting can be just as bad, because many people have horrific prompting style. But I think it's more natural in chat to provide context and explanation, as well as corrections and "oh that's not what i meant" and "in your second point, what exactly do you mean by 'coverage'?", etc. The chat interaction allows you to really hone the code iteratively where both you and the AI have the meaning nailed down.

Uhhh, so, for example, if you were kinda antisocial before, then gave notice and started acting friendlier... this means you were being inauthentic??? Hehehe

Isn't it just the opposite? If you actually feel differently, isn't it more authentic to act the way you feel?

While we're dishing out random advice, here's mine: Don't live in fear of what others think, especially right after giving notice.

This past week I just happened to spend 5 consecutive nights in a capsule hotel in Yokohama, and 2 nights in a goshiwon[1] in Seoul.

In theory I liked the idea of capsules but the experience was not what I hoped. The capsule hotel was highly rated (on Booking.com) and would have been fine for 1-2 nights, but def not good for a longer stay.

- The capsule structures were metal with plastic panels, and would creak whenever anyone got in or out. It was easy to accidentally bump the wall and make a big noise.

- Climbing down the ladder was awkward.

- You could hear anyone snoring in the whole room.

- The aircon in the outer room was great, but as soon as you shut the screen across its opening, the capsule got stuffy despite its ventilation fan.

- Once at 1am, a guy intentionally pounded the wall of his capsule a few times, which reverberated through the whole structure, triggering shouting (in angry Japanese) from other capsules, and in response he just pounded the wall several more times.

The goshiwon cost about 1.7x more but was at least 4x better. A twin bed in a tiny room, with a tiny desk and a tiny window. In comparison it was an amazing value for the price.

Maybe I'd stay in a capsule again if I just needed a place to crash for the night after midnight, but otherwise never again.

[1] https://en.wikipedia.org/wiki/Gosiwon

also: https://www.youtube.com/watch?v=6MtdI5LKfgU

Having it delivered is risky in several ways. If they try to force you to get it delivered, I think it's better to wait until after the preorder period.

And everyone on this thread is basically saying the same thing.

My point about the way most dev shops currently do business is that there's a grey area where there's not a bright clear line. As a manager who hired these dev shops, and as someone with a boss who wanted me to hire these dev shops, I was forced to deal with this issue every time.

> If one person does interviews, while someone else does the work, that is what we call a lie.

If you see my sibling comment (https://news.ycombinator.com/item?id=41649280), I explain how that lie is actually quite typical.

> splitting profits 50/50 just for being part of a sales effort? Too good to be true.

Agreed. But sales commissions can be higher than 50%, so you never know.

1. Most dev shops work in a similar way:

During my career I have worked as a middle manager at multiple companies where I was on the other side of the table. That is, I outsourced projects to dev shops. All of them would bring a technical lead to the initial interview, and all of them would pass the work to one or more junior devs. Of course, they'd say that the tech lead was overseeing the work, but in some cases it was only marginally true.

I came to expect that this bait-and-switch is how larger oursourcing shops work, so later I tended to hire individual contractors where it wasn't a problem. However, since that meant more of my time was spent overseeing the contractors, this approach didn't always sit well with my own boss who wanted higher volume output. So often I was forced to play this game with outsourcing dev shops.

Are those dev shops being unethical? If they got the job done, and they told the client that the junior employees handled the work and the tech lead simply approved the final submission, then I think most employers would be ok with it.

2. Working 2 full-time jobs is ethical IF the employers know that's what you're doing:

If an employee is so proficient that they can do their work at two different full-time jobs, then most employers shouldn't have much reason to complain. But generally we're talking about lying to the employer and since employment contract almost always include an exclusivity clause then it's no minor verbal lie. There's no way this is more ethical than what the dev shops do.

I haven't used Altair but I wanted something higher level than D3 (which I've used but it feels fairly low level) so I went through the learning curve for Vega.js, and it could handle every usecase that I tried, some of which were pretty complex.

I get what you mean about old buildings in SF (living in a drafty victorian is not as pleasant as a modern flat), but I love that particular building. I always take the stairs instead of the elevator, striding two steps at a time, whether going up or down, they're not slippery. The bathrooms are indeed funky but pleasant enough, and even ventilated with an open window. The oak tables by the huge multistory windows on both library floors are a wonderful work environment, with decent outlets and wifi.

https://techstartups.com/2024/03/27/netflix-acquires-ai-powe...

In fact, it seems like they rearranged the exposition from the books just to lay some groundwork for the later seasons. I don't know if this series is supposed to go three seasons, but if the book 3 fairy tales don't come until season 3, it's cool that they set that up in season 1.

Overall I agree, it's pretty good. I may need to watch the 8 episodes again, because I was distracted by comparing it against the book and the Tencent/Chinese adaptation. Some scenes which I could tell were supposed to be emotional or shocking weren't as impactful as they could be, because I knew what was coming.

The biggest change was the Oxford Five, which gave character continuity from 3BP to the Crisis Era, and I think it worked ok. Unfortunately that aspect was also detracting, giving a "small world" sense that they all knew each other. The books sorta had that same issue but to a lesser extent (e.g. Da Shi was involved with all sorts of things that I would have expected was beyond his role as an investigator, even ending up on security detail for Luo Ji, ignoring his radiation poisoning from the first book).

I've been impressed with recent adaptations, like 3BP and Dune Part 2. Villeneuve changed a lot of little details but the story was still intact and impactful. Similarly 3BP/Netflix kept the cool ideas from the book while simplifying parts and dropping some complicating subplots -- e.g. ETO factions, Ye Wenjie cutting the rope at Red Coast Base.

One thing I wish they hadn't cut out was Ding Yi's analogy of the cigarette/pipe filter during his discussion with Wang about the sophons. It was helpful to me to imagine the unfolding of higher dimensions. In the Netfix show, they just seemed to expect everyone to understand that the unfolded particle would look like a mirrored surface.

A note about Tencent adaptation: The CG and acting by the non-Chinese actors were kind of subpar, but the nano-filament scene was awesome and just as good as the Netflix version. I was looking forward to the Netflix version to see if they could top it, but it was pretty much the same.

Speculation: I wonder if Raj will turn out to be the Zhang Beihai character, continuing on to the Battle of Darkness.

Last week when the tomguide review came out, he mentioned about a "Red Wedding" moment midway in the season -- but I know the real "red wedding" is in book 2, and nothing compares to that. What was he talking about? Jack's death?

Am I wrong?

Every time I fail like this, it's because I tried to do too much too fast and it wasn't sustainable. If it's not sustainable, it's inevitable that you'll quit.

For me it's been a mistake to focus on the goal (i.e. don't focus on "lose 30 pounds"). Instead, establish exercise as a daily habit, regardless of the goal. Set aside the time to do it every day at the same time of day, a sustainable amount of time (not too much!). Ramp up slowly, don't injure yourself, don't get impatient, just do a little every day. Mentally commit to doing it for the rest of your life. Don't think of it as a thing you're doing for some other purpose, just think of it as part of your life that you must do every day.

It took a few years of weekend hikes before I was willing to make a commitment to walk a whole 1 hour every day -- and the only reason I did that was because I was going to do a multi-day high-altitude hike (EBC trek) and I knew from experience that if I didn't train for it then I'd probably not complete it. It was a decade before I moved on to other forms of exercise -- it's a very long process!

Someone else mentioned you should cook all your food. I agree, cooking your own meals is an amazing habit with some key benefits (not just nutrition). It provides instant gratification and also once it's established as a habit then you can use it as an exemplar for how to establish other daily habits.

Edit: By "exemplar" I mean that you need proof to yourself that a small gradual habit will eventually lead to the benefits you want from it. It's not easy to see the worth of delayed gratification that doesn't come to fruition within a few months. It requires "faith". Having gone through the process once will give you the proof you need to have faith that you can do it again.

- 3D Mesh Processing and Character Animation, R.Mukundan, 2022 Springer

- Character Animation With Direct3D, Carl Granberg, 2009 Course Technology

- Advanced Animation with DirectX, Jim Adams, 2003 Premier Press

- Real-time 3D Character Animation with Visual C++, Nik Lever, 2002 Focal Press

Textbooks

- Physics-Based Animation, Erleben/Sporring/Henriksen/Dohlmann, 2005 Charles River

- Computer Animation: Algorithms and Techniques, Rick Parent, 2002 Academic Press

There's lots of open source too

* If you use FileVault and don't allow your Apple ID to control it, make sure to establish a specific place where you store the FileVault recovery key. You'll access it rarely (if ever), so if you don't have a standard place where you store secrets like that then you'll forget what you did with it.

* Rename the drive to something unique so you'll recognize this computer by its drive name. This is helpful if you login to other LAN Macs via SMB.

* For your username, consider using a handle instead of your actual name. Also, macOS by default will set the computer name using your Apple ID name ("Joe Schmoe's Macbook Pro") which you probably want to change.

* Thankfully recent macOSes allow you to shut off the startup boot chime instead of screwing around with SystemAudioVolume/BootAudio

Other system settings to adjust:

   Battery/Energy Saver
   Lock Screen (screen saver, display off, require password)
   Hot Corners (when I leave my desk, I turn off the screen by flinging the cursor into a corner)
   Firewall 
   AirDrop/Handoff (airplay receiver off, if you don't need it)
   Dock (remove all apps, move to side, auto hide, etc)
   Mission Control (I use multiple desktops with subject-based wallpapers)
   Change the screenshots location (Cmd+Shift+5)
   iCloud (turn off for apps that you don't want to sync)
   Keyboard (Key nav on, turn off spelling if you don't need, add keyboards if you use multiple)
   Trackpad (force click off, data detectors off)
   Finder (sidebar, toolbar, quick actions - e.g. automator scripts)

* Web browsers - adjust: new window/tab appearance, download location, autofill (off), search engine suggestions, enable the inspect/dev menu

* Little Snitch - migrate your prev ruleset and custom settings

* Sublime - my setup is complicated due to custom markdown settings and package conflicts

* On most apps I've purchased, the single-user license which allows the key on multiple computers for a single user (e.g. Little Snitch, Sublime, Acorn, etc). Best to keep your license keys organized just like 2FA/FileVault recovery keys.

* After apps are installed, you can adjust keyboard shortcuts (e.g. override the Quit item to Shift-Cmd-Q)

Dev:

* Migrate `~/.ssh` keys/config, `~/.gitconfig`, .profile/.shrc

* I have my own way of installing certain cli tools (e.g. Golang, Nodejs, Flutter, etc) where I mimic a package manager (like Homebrew) but install at `~/cli`. If you do something similar, don't forget to `chmod 700 cli`.

* I use NixOS on servers but didn't like the drive partitioning madness on Mac, so I still use Homebrew for other things.

* Ruby/Cocoapods: 1) rbenv via Homebrew, 2) ruby via rbenv, 3) Cocoapods via gem.

* Python: via Mamba (though I hate the name miniforge3, and wish they'd standardized on mambaforge)

Finally, before wiping your old Mac (to sell or hand to Apple for service), have a checklist:

* Redundant backups? (Some ppl argue against a disk image via Disk Utility, but I like being able to copy/mount the dmg image. I don't like Time Machine.)

* If you log out/in with your Apple ID account on a different Mac (e.g. the old Mac), will the new Mac ask you for 2FA approval?

* Can you log into your Google account(s) on the new Mac?

* Did you migrate any stored logins from the old Mac (e.g. in keychain or browser)?

* Can you ssh into your servers? Connect to AWS?

* Can you push to Github repos?

HTH

His story timeline:

* 2016: Grind for 4 years, got burnt out and took a job.

* End of 2021: Got fired, started as a solopreneur again.

* H1 2022: Launched 7 projects in 6 months.

* H2 2022: Fell in love with one of them, went all-in for 6 more months.

* Early 2023: Switched back to 2021 approach, 8 projects in 8 months.

* Fall 2023: "quit all my faceless AI projects and focused on products I’d use"

* Sept to mid-Dec: "ShipFast alone made $168,100 in total revenue"

I found it interesting because he oscillated btwn the "ship a lot of stuff fast" and "care about a product that I'd use myself" approach. Sounds like he did two full cycles of that until he hit a winner. (Not that you can't ship fast on smtg you care about, but if you launch a dozen products it's hard to treat each one like your baby.)

I think this story is more honest than most. This was not a simple narrative. The reality was chaotic, had ups and downs, and had no coherent narrative. He gives his lessons at the end but what I took away was that anyone trying to do this has to go through the process themselves to learn their own lessons.

The fact that his hit product was SaaS boilerplate isn't a negative to me. (At least it's not courseware on "How to be a solopreneur".) He pitches four solopreneur-oriented products at the end of the article. Maybe you don't like that, but I think it's smart, because why did he write this blog post and who is supposed to read it? I mean, we all take away something from everything we experience, and you can interpret it very differently from me.

This solution isn't for everyone, but I made my pc case out of makerbeam rails, hanging the GPUs vertically, attached to the motherboard with PCIe x16 riser cables.

This photo doesn't show the riser cables but it does show how the GPU attaches to the rails: https://www.instagram.com/p/BMcSU_ig-Ij/