nyrulez

Bold claims that writing code was never the bottleneck. It may not be the only bottleneck but we conveniently move goal posts now that there is a more convenient mechanism and our profession is under threat.

Things haven't changed much in terms of truly new ideas since electricity was invented. Everything else is just applications on top of that. Make the electrons flow in a different way and you get a different outcome.

"But then I look at what these tools actually manage to do, and am disillusioned: these tools can be worse than useless, making us net-negative productive." It starts from this premise right in the first paragraph. And goes on to illustrate an example that proves their point ("Let's pick one of the best possible examples of AI-generated code changes.").

anyways, it doesn't matter that much :) we could be both right.

I am not sure what you're implying. The first sentence makes no sense. LLMs aren't giving you non-deterministic code. The code is shown to you and have complete control over how it looks and operates. Not understanding the mechanics of how the code is generated by the LLM doesn't make the output non-deterministic.

If you choose to accept bad code, that's on you. But I am not seeing that in practice, especially if you learn how to give quality prompts with proper rules. You have to get good at prompts - there is no escaping that. Now programmers do suck at communicating sometimes and that might be an issue. But in my experience, it can write far higher quality code than most programmers if used correctly.

"Mastering" is a strong term and is even misleading, especially when talking about tools that give you leverage. I mean if someone masters running, does it mean you never use a car? There are thousands and thousands of instances in everyday programming where AI is going to be 10x-100x faster than any human, especially at the function level and even file/script level.

I can give you a concrete example since things sometimes can be so philosophical. The other day I needed a LIS code (Longest Increasing subsequence) with some very specific constraints. It would've honestly taken me a few hours to get it right as it's been a while I coded that kind of thing. I was able to generate the solution with o3 in around 10 minutes, with some back and forth. It wasn't one shot, but took me 2-3 iteration cycles. I was able to get highly performant code that worked for a very specific constraint. It used Fenwick trees (https://en.wikipedia.org/wiki/Fenwick_tree) which I honestly hadn't programmed myself before. It felt like a science fiction moment to me as the code certainly wasn't trivial. In fact I am pretty sure most senior programmers would fail at this task, let alone be fast at it.

As a professional programmer, I deal with 20 examples every day where using a quality LLM saved me significant time, sometimes hours per task. I still do manual surgery a bunch of times everyday but I see no need to write most functions anymore or do multi-file refactors myself. In a few weeks, you get very good at applying Cursor and all its various features intelligently, like an amazing pair programmer who has different strengths than you. I'll go so far as to say I wouldn't hire an engineer who isn't very adept at utilizing the latest LLMs. The difference is just so stark - it really is like science fiction.

Cursor is popular for a reason. Lot of incredible programmers still get incredible value out of it, it isn't just for vibe coding. Implying that Cursor can be a net negative to programmers based on an example is a lot of fear mongering.

I mean the headline "Net-Negative Cursor" is a pretty far reaching conclusion. The article does try to generalize on the implications from a code snippet for AI powered programming. The headline isn't "The example on Cursor's website is incorrect".

Amazed at all the negative rhetoric around coding with LLMs on HN lately. The coding world is deeply split about their utility.

I think part of that comes from the difficulty of working with probabilistic tools that needs plenty of prompting to get things right, especially for more complex things. To me, it's a training issue for programmers, not a fundamental flaw in the approach. They have different strengths and it can take a few weeks of working closely to get to a level where it starts feeling natural. I personally can't imagine going back to the pre LLM era of coding for me and my team.

Damn. Didn't expect so much bitterness this morning. What a strange set of "demands" for a pretty decent piece of kit.

Landing page is nice but I am confused why this would be better than generating a prompt from the LLM itself? Latest models are extremely good in taking the seed of a problem and enhancing the prompt to the nth degree until you're satisfied.

Asking a programmer to ditch their editor is just wild. A lot of this thread seems to be full of non-programmers trying to figure out why Cursor exists lmao

Calling Cursor a vibe coding app is just wild. The journalists and influencers are so out of touch with the software engineering world.

The point isn't to compare random AI art with random human art. The overarching sentiment lately has been that AI art feels bad and has this "Fake" quality to it.

This survey is refuting that argument. AI art can be used in media just like human art and people can't really tell (or care if they can't tell the difference).

It's about stuff we don't know yet. From today's lens, the essay seems absurd. But I think it's hinging on continued discoveries that improve one or all of learning algorithms, compute efficiency, compute cost and applying algorithms to real world problems better.

5 years ago, I wouldn't have believed any of what exists today. I saw internal demos that showed 2nd or 3rd grade reading comprehension in 2017 and statements were made about how in the next decade, we will probably reach college level comprehension. We have come so far beyond that in less than half the time. Technology isn't about scaling incrementally and continuing on the same path using the same principles we know today. It's about disruption that felt impossible before - that feels like a constant to me now. Seeing everything I've seen in the last 20 years, it's going to continue to happen. We just can't see it yet.

I missed the "yeah, we're done here" part. Don't think Waymo is going to pause their safety research program now.

tbh this is the main thing I was looking for on the webpage. It's hard for me to jump ship from my existing workflows unless I see a professional and highly sophisticated app example. I am sure it works for simple cases but I am concerned about limitations that one hits as one does more complex things

Weird lack of examples. I was curious but I am not going to download before I have some idea of what I am getting into.

Curious to know why you think it's vaporware. Are the latest LLMs like 3.5 Sonnet bad at original programming based on your experience? It hasn't been the case for me when using it for real world projects lately.

This is cool. There is a lot of hesitation to move significant assets from my Charles Schwab brokerage account. That's going to be my main hurdle. I need to understand the safety, trust, trading mechanics, flexibility that I have with Charles Schwab, all of those things become very valuable when talking about high amounts for retirement purposes. I don't really know what you can do there to attract such clients? Will anyone move $5m from their top tier brokerage to Double yet?

I love the science of evolution in that there's no need for an underlying model for such an invention to actually verify the science and the possibility of it, just that such an invention is possible. Basically anything and everything is possible with evolution and that doesn't really feel like science to me.

I am still glad they exist. If this thinking was very prevalent among AI pioneers (which it probably was for companies like Meta and Google, that's why failed to innovate here), we wouldn't have gotten revolutionary products like ChatGPT.

For me the fundamental question remains if it's a matter of the right input data (beyond language) or is there something about our brains that cannot be replicated by artificial neural architectures for it to discover new knowledge and invent new things for humankind.

I mean are we as humans planning ahead of the new few words? I certainly am not. But what matters is a deeper understanding of the context and the language model itself, which can then produce sensible spontaneous output. We as humans have the advantage of having a non language world model as well as abstract concepts but all of human language is a pretty strong proxy for it.

The spontaneity of it isn't the issue, it's what's driving the spontaneity that matters. For e.g. 1M context window is going to have a wildly more relevant output than a 1K context window.

Curious if they survey the drivers too, the people who are affected by this. I see no mention of this in the article. I'd think they should have a major say.

Do you know the amount she was consuming and the brand? Total Dosage, dilution and the per unit strength of the ACV brand can really matter, like anything else we might consume.

The paper does do a good job of describing this pretty important detail: "Groups 1, 2 and 3 consumed 5, 10 and 15 mL, respectively, of ACV (containing 5% of acetic acid) diluted in 250 mL of water daily".

I can almost guarantee if that amount went up to 30-50ML and without any kind of dilution, effects can be pretty adverse.

I've been consuming it by diluting a tablespoon of ACV heavily with some warm water that I can barely taste it. It's really helped fix my acid reflux while all the prescription drugs from my PCP didn't seem to do anything.

I thought this article was about copyright. Turned out it was about them having growing pains. Though the jury is still out if a truly general purpose and truthful LLM can exist one day. And how far can this tech be pushed.

What does it mean to be critical of LLMs? In that they are overhyped or that they are not intelligence?

One Medical is the best value in health services today. I wish this was the default service model for US at scale.

If those were frauds, there would be ton of lawsuits by now given their history.

It may or may not be immoral for participants, but illegality is a much higher bar.

You also mention there's a massive body of evidence in this regard. Where is it? I'm genuinely curious as I've wanted to dig into it.

There is no data that I've seen that shows blood clots for vaccinated vs covid vs unvaccinated. Everyone keeps saying there's full transparency while not linking a single source. It's quite paradoxical.

If all this data is so well connected, how come my heamtologist has no clue about this? He is a pretty senior doctor.

If data is being collected, how come my doctor didn't report this potentially adverse event?

If people are being careful about this, why did the doctor outright reject any connection between vaccination and clots, even if they happened next to each other. How is that scientific if you reject observations that don't meet your theories.

I'll believe it when I see an actual study or data on this topic. The lack of transparency around vaccinated cohorts is very evident.

If you disagree strongly, I'd love for you to link me to anything. Just saying "trust me" isn't very scientific.

I think you provide a good link that's a basic start.

Here is my central issue: I still can't find basic cohort level information about blood clot incidences at the cohort level: covid/vaccinated/not vaccinated. I think this is pretty basic stuff. Since you seem to be knee or chest deep into this, any idea why this information is so hard to find? It's a pretty obvious question that I think someone like you would be curious too I assume?

I remember the study you provided. It was from the early days. What happened since then? Millions of people have been vaccinated repeatedly in the US itself, so where is the follow up for clots and other issues?

My cousin who works with a healthcare provider mentioned casually to me that health insurance claims around blood clots have gone up rapidly in the last 2 years. I am all for believing its purely covid related, but why not just dispel any myth and release the cohort information as I mentioned above. It's pretty basic stuff as a data scientist myself.

There is a bit of "trust me bro" vibe going on around any kind of large scale cohort analysis for vaccines that makes me very uncomfortable.

If I am completely misguided about my last sentence, would love to see concrete numbers at a population scale (from a credible source), not just a study among some participants. It would be make me pretty happy to see real population level vaccine cohort data, which so far seems to have been evasive for me (any my doctors).

----

Independently, I am going to share something else. My doc was VERY uncomfortable even implying that vaccine could be a possible culprit behind my issues, even though there was such a clear temporal link and I'm young/healthy otherwise. I guess there is a lot of pressure on docs to not be considered Anti Vax or even have their credibility be tarnished. But I think you can see how that can bias data and reporting one way or the other.