* Robotics Hello World: Objective is to implement ACT model to train my arm robot on simple pick-and-place tasks. Leaning heavily on HuggingFace's LeRobot library, but stopping short of using their model implementation and training loop. https://github.com/avilay/learn-robotics
* Designing a new programming language: This is when I want to escape the annoyances of coding in Python and start daydreaming about a new language :-) https://github.com/avilay/kulfi
I’m going to annoy people who actually wield this language well. I feel like there’s less of a clear point and more of an aesthetic. Like not even metaphorical or allegorical. It’s just the overall feel of wandering and pointlessness that creates a sense of calm.
I'm not quite sure I got it either, but I guess this is probably the main gist:
> by the time most people wind up here, they’ve got plenty of regrets. Nobody gets it all right. You’re born, and then you go through life making the choices that you think are the best given the information you have at the time, and you don’t always have all the information to make the right choices. Do yourself a favour and forgive yourself for any failings on your part, you’re only human after all
I'm reminded of the lyrics from Pink Floyd's Time:
"And then one day you find
Ten years have got behind you
no one told you when to run
you missed the starting gun"
The longer you live, the more tiny little mistakes you make. Things that at the time you could have done better, if you'd known, if you'd been a bit more careful. And these weigh on you, emotionally, pretty consistently.
And while it's pretty absurd, in the story, such tiny mistakes having such outsized consequences, the story reminds us that such severe consequences are well within the realm of possibility. People do lose limbs off of little, careless mistakes. Doubly so with all the incredibly concentrated sources of energy we have in the modern world - power tools, automobiles, explosives.
Would one really lose ten years trying to pick out a single Netflix show? No. But could one wake up one day and realize that they'd accomplished nothing of note for a decade, that all their free time was dumped into Netflix shows that weren't even that good?
So, what do you do with all that? Memento Mori, I guess.
I don't know if that is true. For large swaths of the population, raising a child is their biggest accomplishment. Justifiably so.
Other accomplishments of note might be: a single conversation that helped someone change. Little acts that made the world a tiny bit better. Having brought happiness to other people. I like to think that is the meaning of not wasting your time – not just measuring your life's worth with a science/capitalism lens.
I guess I interpreted "of note" differently. Raising a child is not noteworthy, it's just normal. Millions of people do it. Nobody will be remembered for raising a child, except by the child himself.
When I'm gone, I'll leave nothing "of note" behind. I haven't won any great prizes or set any records. I haven't authored any papers. I haven't invented anything that changed an industry. I haven't cured any diseases. My name won't be on any buildings or monuments. I haven't really left any kind of discernable mark on the world or civilization or even my home town.
Raising a child is noteworthy. Same for the millions of people who do it now, and the millions of people who did it before them. Just because millions of people do it doesn't diminish the significance of it. Framing it as such, is like saying there's no point running a race if you don't come first. A personal victory is still a victory. If it matters to you, then that's all that matters. All of those things like authoring papers, winning great prizes, setting records, will all be eventually be lost to time.
It's an allegory for AI hysteria and WFH depression. Generally, anything to put a wet blanket on the nice things that have happened to tech workers in the past few years. To put salt in the wound, it's done in a style that used to delight HN.
It reminded me of those cringe videos CGPGrey put out for COVID.
The point about papercuts adding up so resonates with me! I loved Zed initially and did find it more responsive than VS Code, loved the Zed Agent autocomplete, etc. However, I eventually and reluctantly went back to VS Code. The papercut that finally did it for me was [this open bug](https://github.com/zed-industries/zed/issues/36516) because of which I was not able to step into a packaged library's code when I was debugging my own code, this was in Python.
The two don’t have to be mutually exclusive. You can let the agent code and you review it, or vice versa. No different from being a team lead where you don’t write all the code, or even review each and every line of code, but you have a very firm grasp of the code base.
on the contrary, reviewing LLM code and human code is very different. LLMs don't learn. if a human makes a mistake i can teach them to avoid that mistake in the future. if an LLM makes a mistake, all i can do is fix it over and over again. the dynamics are fundamentally different. some people may prefer to work with a machine, but i don't, i prefer to work with humans.
for me this is similar to the difference of using FOSS vs closed source software. if there is a problem on my linux machine, i can potentially fix it, on windows or mac i just can't.
both closed source software and working with LLMs make me feel helpless. whereas using FOSS or working with humans is empowering.
i get that not everyone feels that way, and that's fine. for my part i'll just stay away from LLM generated code.
In theory, vibe coding and understanding don't have to be mutually exclusive, but in practice, I think that the people who have the discipline to actually maintain their understanding of a codebase are few. I've code reviewed things from people who claim they are reviewing what comes out of the LLM carefully, and talked to them about the code, and while they think they understand the code, they simply don't, which becomes abundantly clear when I try to explain the problems I find in the code.
Totally agree with this! Being "kindly honest" is way better than being "brutally honest". Being honest and direct is important of course. I have often found that delivering constructive criticism in the so-called sandwich manner often obfuscates the message, so delivering it directly is much better. However, being kind to the receiver of the feedback by having empathy for them and supporting them as they process that feedback will help land that message far more effectively than being "brutal" about it.
This hit a nerve as I am in the middle of developing a webapp for myself using NiceGUI.
I find CSS, especially its layout framework, pretty confusing and sometimes downright intimidating to work with. `inline`, `block`, `flex`, `grid` seem reasonable when you read about them. But when using it, especially within frameworks when flexboxes are nested within grids which are nested within flexboxes and so on, it becomes hard to reason about. And then you throw in media-queries in the mix and it becomes even more dense.
Lots of nostalgia in this thread. However, I recently started playing around with IBM's quantum computer cloud service and it has been by far the best experience compared to other players. Who knows, that might be their come back story!
Surprised not to see ghost in here. They have a paid managed version at https://ghost.org and it is fairly easy to set up with any of the public cloud providers. I have set mine up with Digital Ocean (https://ghost.org/docs/install/digitalocean/).
Cool, thanks for the response. Yes, I do find that the PyTorch tutorials on distributed training are a work-in-progress.
I was thinking of starting with a basic implementation of the original paper by Jeff Dean, et. al. on synchronized data parallelism, implement basic model parallelism, explain why async parallelism works, do a simple implementation of HOGWILD!, and finally do "hello world" training using existing distributed training systems like Horovod, Distributed PyTorch, RayLib, Microsoft DeepSpeed, etc.
"Hello world" examples already exist for all of those. Reproducing them is not very interesting. If you're willing to dive a little deeper, try to implement SyncBatchnorm: explain design choices, measure the performance impact, describe any bugs you had in your implementation. Such a case study would be very interesting to read, and would probably get you noticed.
Does anybody else think this is in response to Apple's privacy crackdown on third party tracking data? If you buy on Buy, then you are still first party for Google and they can still get your signals.
reply