Same, I cringe when I read this structure.
2,781 karma · joined January 10, 2013
Here is my guess for the puzzle: creative work is subjective and full of scaffolding. AI can easily generate this subjective scaffolding to a "good enough" level so it can get used without much scrutiny. This is very attractive for a creative to use on a day to day basis.
But, given the amount of content that wasn't created by the creative, the creative feels both a rejection of the work as foreign and a feeling of being replaced.
The path is less stark in more objective fields because the quality is objective, so harder to just accept a merely plausible solution, and the scaffolding is just scaffolding so who cares if it does the job.
Henry George, and David Riccardo before him, figured that as productivity and thus wealth increases the value accrues to the land owners, not capital not labor.
This is because Land is the fundamental bottleneck of human activity, the core finite resource. And as everything else gets more productive, the land itself becomes more valuable.
So, yes, tax Land, and redistribute as a dividend to all citizens. After all, no one can be credited for building that Land.
It should give you an idea of how hard it is to do a SOTA model from scratch!
If you relax the SOTA aspect, Karpathy's nanochat has you covered: https://github.com/karpathy/nanochat
In the game we're building we generate, compile and run code (C#) in real time to let the player "train and command" its monster in creative ways. So, I've thought about this.
You need both a popular language and one that has a ton of built-in verifying tools.
The author correctly highlights the former, but dismisses the latter as being targeted to humans. I think it is even more important for LLMs!
These coding agents are excellent at generating plausible solutions, but they have no guarantees whatsoever. So you need to pair them with a verifying system. This can be unit tests, integration tests, static / type checks, formal methods, etc. The point is that if you don't have these "verifier" systems you are creating an open loop and your code will quickly devolve to nonsense [0].
In my view, the best existing languages for vibe coding are: - Rust: reasonably popular, very powerful and strict type system, excellent compiler error messages. If it compiles you can be confident that a whole class of errors won't exist in your program. Best for "serious" programs, but probably requires more back and forths with the coding agent. - TypeScript: extremely popular, powerful type system, ubiquitous. Best for rapid iteration. - Luau: acceptably popular, but typed and embeddable. Best as a real-time scripting sandbox for LLMs (like our use case).
I think there is space for a "Vibe-Oriented Programming" language (VOP as the author says), but I think it will require the dust to settle a bit on the LLM capabilities to understand how much can we sacrifice from the language's lack of popularity (since its new!) and the verifiability that we should endow it with. My bet is that something like AssemblyScript would be the way to go, ie, something very, very similar to an existing, typed popular language (TS) but with extra features that serve the VOP needs.
Another aspect to consider besides verifiability is being able to incrementally analyze code. For structured outputs, we can generate guaranteed structures thanks to grammar-based sampling. There are papers studying how to use LSPs to guide LLM outputs at the token level [1] . We can imagine analyzers that also provide context as needed based on what the LLM is doing, for example there was this recent project that could trace all upstream and downstream information flow in a program thanks to Rust's ownership features [2].
Finally, the importance of a LLM-coding friendly sandbox will only increase: we already are seeing Anthropic move towards using LLMs to generate script as a way to make tool calls rather than calling tools directly. And we know that verifiable outputs are easier to hillclimb. So coding will get increasingly better and probably mediate everything these agents do. I think this is why Anthropic bought Bun.
[0] very much in the spirit of the LLM-Modulo framework: https://arxiv.org/pdf/2402.01817 [1] https://proceedings.neurips.cc/paper_files/paper/2023/file/6... [2] https://cel.cs.brown.edu/paper/modular-information-flow-owne...
This has nothing to do with American Hustle culture and just with that professor's judgment.
The professor noticed it (presumably via seeing poor "show your work") and gave zero points on the question to everyone. And once you went to complain about your grade, she would ask you to explain the answer there in her office and work through the problem live.
I thought it was a clever and graceful way to deal with it.
Each of these phenomena have a name: there’s Jevons Paradox, which means, “We’ll spend more on what gets more productive”, and there’s the Baumol Effect, which means, “We’ll spend more on what doesn’t get more productive.”
```
I don't think that's exactly right. Jevons says "we consume more on what gets more productive" and Baumol says "the unit cost increases for that which is less productive".
The typical example for Baumol is the orchestra (or live music) which is today much more expensive than in the 1800s. I don't think we spend more in aggregate than we did in the 1800s!
Edit as I continue reading: ```
Other goods and services, where AI has relatively less impact, will become more expensive - and we’ll consume more of them anyway. ```
This definitely NOT the case. Basically the author is saying we will consume more of everything, which is not true! We famously stopped using horses and all the relevant industries.
The unit cost for horses, however, did increase!
What the author should be stating is that the new production bottlenecks will command a higher price and probably play a bigger role in the economy, but not everything gets to be a new bottleneck.
I wish it had the Wikipedia style of writing as a default, as in, much more matter-of-fact writing (even if not everything is a fact).
I think part of the problem is that people overwhelmingly vote for this style with up votes and revealed preferences.
Maybe there should be a more meticulous feedback / prompt system where I can highlight a paragraph or sentence and ask annotate my feedback so that it doesn't go for that style.
I already spend too much time reading LLM outputs on my own interactions. And I get sick of their style because of it. So when I read it during leisure time, it just triggers a gut rejection.
Especially because they are so formulaic / template-y.
Now... Was this article LLM written?
This part triggered all my LLM flags: ``` Adding a bicycle chain isn’t just decoration—it shows understanding of mechanical relationships. The wheel spokes, the adjusted proportions—these are signs of vision-driven refinement working as intended. ```
The only other player I've seen is Mechanical Orchard
This sums up my experience with US Healthcare. They bill expecting you to autopay, and either have no incentive to bill correctly or they outright are trying to scam but the result is that every hospital bill is sus.
This also makes insurance a lot less inherently valuable: you are paying for someone to do this untangling shitshow on top of the actual insurance. As if the hospitals just put the billing burden on the client.
There has to be a penalty for sending wrong bills, or they should pay me for my time wasted.
Finally, the prices are so inflated that often the price without insurance in Europe is the same as the copay/coinsurance in the US.
Its a fucking catastrophe.
Also, I've heard from others that the Qwen models are a bit too overfit to the benchmarks and that their real-life usage is not as impressive as they would appear on the benchmarks.