Speaking only for myself, that is absolutely accounted into my despair when I see an AI poster. Not just the fact that we're having a wide range of cultural and regional artifacts being stamped out and homogenized by the convenience of AI posters, but the fact that I'm only able to point them out right now because the public backlash is somewhat limited to the online sphere, so most people are okay with the default result it spits out. And we're one default ChatGPT/Claude "curated diversified aesthetic selector" feature away from making my current discernment completely useless.

Of course there's an argument of "if it's so good you can't tell, is there really a problem?", but here I'm responding specifically to the argument of a negative reaction to AI posters not taking the above into account.

[0]: https://michaelnotebook.com/df/index.html

Seeing how bad frontier models were at the ARC AGI 3 benchmark (this pre-Astra), I thought it would be fun to translate aspects of pwning a program to spatial game-based mechanics and try to have some of the "reasoning" happen in the game world instead of all in code. This resulted in challenges where they have to interact with code to change parameters in the game to solve puzzles.

While players do have to read code, the game expects no prior knowledge of C programming specifically, and tries to explain concepts such as structs, byte hex encoding, null-terminated strings, endianness, etc. from scratch. It has 10 rooms and builds up to the player being made to do a buffer overflow to clear the final room. There are items the player acquires to inspect source code, see program memory, and translate raw bytes to unsigned 4-byte ints or ASCII strings.

There are still some gaps in the ramp up that could be better filled in and might require more hand-holding for complete beginners (I am thinking of adding an item that explains basic C semantics such as variables, conditionals, loops, I/O etc. when you directly hover over the code in the computer). Would love feedback from people here!

Comments URL: https://news.ycombinator.com/item?id=49712697

Points: 2

# Comments: 0

[0] https://www.goodreads.com/quotes/1192222-a-lot-of-fans-are-b...

There were really two components to it I valued - the inherent interest in the topic itself of course, but also the sense of community and recognition you got as you dove deeper and created cooler things to share with people. At some point the second overtook the first, and I internally started to value the outcome or reaction to something I would create more than the satisfying pursuit of making itself. Doing a thing wasn't valuable unless it would be acknowledged by some group of people. It's been a long process trying to get rid of that mindset and I'm still not done, and has caused a lot of anxiety and questioning my interest in the hobbies I love.

A lot of the questioning has been driven by the absolute saturation of my online feeds. The algorithm can so quickly figure out the type of stuff I like and show me so much of it that it begins to start feeling meaningless in 10 minutes of scrolling - deconstructing that value system in real time (and I guess technically helping? hooray for accelerationism). If there is so much out there, what am I supposed to create that could add to the pile in a meaningful way? It really takes a lot of effort to sever yourself from the algorithmic mindset and remind yourself that you got into it because you like doing the thing.

Not to mention the second order effects, like the ancillary feelings of guilt that show up because I start questioning why I want to avoid seeing things about topics I supposedly enjoy. Part of it is personal flaw - I find it difficult to see something cool without feeling some imperative to figure out how I could do it myself. When I don't want to feel bothered, I'm driven towards more mindless consumption that doesn't run the risk of reminding me that I should be doing something. I feel I would be much happier engaging in the hobbies I have if I was a bit less aware of everything related to them going on in every part of the world, and the algorithm is a big culprit in that. Detaching from it is not just hard because "phone feels good", but also because it means completely giving up on the value system I grew up on, where you could know if you were doing something meaningful if it got attention from other people in the communities you wanted to be a part of. "Attention" has been too commoditized at this point to use it as a fulfilling indicator of return (but unfortunately, some people have to play this game if they want to support a living off a creative endeavor - at least I can ignore that part because I just make music or do creative coding as a hobby).

I've been scraping instagram posts recently to identify AI misinformation accounts that all repost each other's carousels and get hundreds of thousands of likes in engagement. Thinking of ways to present it and your dashboard looks very helpful. Did you experiment with any other types of visualizations before deciding on a stacked area chart for the clusters?

A notable exception would be people like CEOs and managers higher up in big tech, who might be used to skilled engineers and domain experts reporting to them in unfamiliar lingo. Maybe that's why we don't hear as much on the everyday annoyances of Claude's language from that camp?

And somehow, the programming language he made (that currently bans AI contributions outright) ended up being instrumental for a tool that's in the center of the whole craze, and doing Tony Hawk 900 aerial spins with agentic AI software rewrites under a frontier lab. I'm not commenting on the effectiveness of the rewrite, but it seems fascinating that for a moment Zig found itself at the intersection of two very different camps that the software world seems to be divided into. It seems in retrospect like a bomb waiting to go off even if they didn't jump ship to Rust!

[1] https://www.youtube.com/watch?v=zLZwpH5lCD4

I think my downfall might have been questioning the efficacy of my method beyond it just "feeling right" and trying to find ways to show myself that I am learning.. maybe the correct approach is to just do whatever and hope for the best

That's a very good question. In my case, when learning in the mode I described above, my brain is often more worried about the amount of unknown stuff versus focusing on what I am working on at the moment. With that, learning starts to feel like a burden and leads to burnout (even though I enjoy solving problems). I can become closed off to new information, and disheartened at everything I don't know.

The problem I'd like to solve is being able to retain excitement about learning while working on projects without it feeling like an obligation. There are probably more base issues to address for this (eg. being less controlling of outcomes). And there's particularly a lot of doubt now that AI coding is the norm, and I feel uncomfortable with moving faster because I feel attached to all the details. That has turned the debate much larger about what anyone really needs to know if they can blitz through stuff at the speed of prompting.

> Usually homeworks get very easy which is a good sign

I agree, but having homeworks in itself is a big advantage! It means you're in a structured learning environment where someone has taken the care to create a learning path and added checkpoints with carefully designed problems designed to refresh + assess what you've absorbed until that point.

When you're learning outside the classroom (for yourself or on the job), there's this sudden loss of all bounds where you now also have to draw a line of what knowledge is reasonable to acquire and what is out of scope. For example, when working independently on personal projects, I sometimes get sucked into reading about the details of protocols, standards, RFCs etc. and have a difficult time deciding when to stop.

One answer is learn just enough to solve the problem at hand at move on, but if you're in a field like eg. security you feel an obligation to do your due diligence when designing something. And of course, with LLMs now it's possible to move past any problem without learning anything at all, so you have to be even more intentional in which things you do decide to learn.

But when I relate these activities to learning itself, it becomes difficult to gauge how much of it is necessary. If I don't constantly serialize or test what I'm learning, how do I know I'm learning anything at all? And externalizing so much starts to feel like a burden and gives me less confidence in my own intrinsic ability to just go and do things.

I was going to post an "Ask HN" on this topic to see if other people related to this. If anyone has been in this mindset before and managed to break out of it, I'd be curious to know what helped!