Having tried most (all?) of the commercially available + open source options, and even tangential competitors like CC, Conductor, Antimetal, etc. I haven't found anything that's close to the experience of Cursor. The harness they've built is incredible.
I'd even go so far as to say that any competitors that are direct (windsurf, kiro, etc.) aren't even in the same universe. Cursor is just so much better, faster, has better features (plan and debug mode), and squeezes much better results/code out of the same models. They absolutely have some secrete sauce that the other options just don't have.
Cursor is my favorite of the VS forks. Agree that it delivers better plans than others. I prefer using Claude in Cursor over CC CLI when I am heads down going through bugs. I am disappointed in how "little value" in token use Cursor provides compared to others.
Not quite first party, but composer 2 is far superior to grok for coding. Unless you're eluding to them using SpaceX infra to train their own model vs. using grok
For programming, the way I would build a curriculum is to force students to actually learn how to program and code first. This is simple by requiring them to write code inside the classroom by hand for all exams.
I would make this the focus for 90% of the first 2 years of their degree.
I would then have them spend 75% of their last 2 years learning how to use and program with AI. Aside from knowing how things actually work, there's no more important skill now than mastering AI.
First part is basically my degree. You can cheat as much you want, but when you have to write algorithms by hand on paper in the exam it doesn’t matter - you will have to learn. Then depending what you want to do for masters it’s basically machine learning, which mostly mathematics.
I don’t know why you’d need to teach anyone to code with LMM’s though. If you are a CS major (or any reasonably intelligent person) and can’t figure it out on your own, well, workd needs bakers and carpenters still you know?
What I am learning is mathematics, which what all CS including AI and machine learning, really is. ”Mastering AI” for me means building your own models and AI applications and undestanding linear algebra, multivariable calculus and the propability theory behind it all.
The cynicism or even lack of interest is because it's extremely underwhelming.
If you ask 100 people in 1969 what humans would be doing in space in 57 years I can guarantee not a single person would guess that we've done nothing of substance. And that the most exciting thing we have done that involved human space travel is simply flying around the moon, people wouldn't believe it
Humans going around the moon will be amazing every single time for the next 10,000 years it happens for anyone who isn't already a miserable person. Going for a swim in the ocean is an amazing experience every single time and I can do it every day. It's still a great feeling. Going to the moon is so much more extraordinary in the literal sense of the word. The fact that any collection of creatures is able to do it is remarkable.
It really sucks that you responded to my exhaustion with the exact thing I'm exhausted by. And you aren't even saying anything new. Please don't do this again.
MacOS is a bug filled nightmare, and it's still light years better than Windows. I haven't used Ubuntu extensively since early 2019, but it still wasn't comparable to OSX at the time.
Apple and specifically MacOS is significantly worse than it has ever been, but again, still far better than the alternatives.
I don't understand why. People spend money on other things besides housing. Because people spend money on multiple things, it doesn't really make that much sense to say that our index of inflation should track be one thing. I mean, if the price of food and healthcare tripled, I think you would probably say that the inflation metrics should go up.
Ofc, focusing on just one thing is very convenient for people who want to tell a particular story. (inflation is so bad! look at housing! there's so much deflation! look at food and TVs!)
I think it's because housing is the biggest expenditure for my family. Like I said, you should build your own index, not using the CPI or other people's index. Similarly, change in life can increase expenditures, too, e.g. getting a child prompts the family to buy a house instead of staying in a condo.
For my family, housing is easily the primary expenditure -- around 6,000 CAD while (food + vehicle) amount to less than 2,500 CAD monthly. For a similar family in the same area with on vehicle, I estimate that housing probably takes at least half of their expenditure.
Yes, that would be stupid, which is why it doesn't work that way. The basket is weighted according to how much people spend on each item. Eggs are not weighted the same as rent.
The CPI does have a problem with not updating the basket as frequently as it could, which means it doesn't catch substitution effects and tends to overstate inflation.
It'll never happen because it shines a light on uncomfortable facts that would risk far too much cognitive dissonance across the political spectrum. Please keep the discourse to identity politics, culture wars, the Epstein files, and large-scale, unprovoked acts of international warfare; those will all be much easier for us to talk about as a nation than what we should do about housing prices.
Not even close, not when all things are considered. $50/hour is 100k/year, which is still considered a decent salary. 24k/year in 2000-2002 was definitely not considered a decent salary. $12/hour for sw engineers was evil. I hung up on that recruiter and cursed for a while, cold-called my way to a transitional $20/hr job, and then finally landed somewhere at $55/hr which is when things started to feel normal again. $55/hr back then is not the same as $230/hr now.
AGI is here. 90%+ of white collar work _can_ be done by an LLM. We are simply missing a tested orchestration layer. Speaking broadly about knowledge work here, there is almost nothing that a human is better at than Opus 4.6. Especially if you're a typical office worker whose job is done primarily on a computer, if that's all AGI is, then yeah, it's here.
Opus is the very best and I still throw away most of what it produces. If I did not carefully vet its work I would degrade my code bases so quickly.
To accurately measure the value of AI you must include the negative in your sum.
That "simple orchestration layer" (paraphrased) is what I consider the AGI.
But yeah, I suspect LLM:s may actually get close enough. "Just" add more reasoning loops and corresponding compute.
It is objectively grotesquely wasteful (a human brain operates on 12 to 25 watts and would vastly outperform something like that), but it would still be cataclysmic.
If we can get AI down to this power requirement then it's over for humans. Just think of how many copies of itself thinking at the levels of the smartest humans it could run at once. Also where all the hardware could hide itself and keep itself powered around the world.
Yeah, but a human brain without the human attached to it is pretty useless. In the US, it averages out to around 2 kW per person for residential energy usage, or 9 kW if you include transportation and other primary energy usage too.
I suspected it wasn't just battery farms, but more like what you see in less mass market scifi where the humans are used for more than just batteries... they'd also be some storage and processing for the system (and no longer humans).
However at that point I don't see the value of retaining the human form. It's for a story obviously, but a not-human computational device can still be made out of carbon processing units rather than silicon or semiconductors generally.
I ran a quick experiment with Claude and Perplexity, both free versions. I input some retirement info (portfolios balances etc), my age, my desired retirement age etc. Simple stuff that a financial planner would have no issue with. Perplexity was very very good on the surface. Rarely made an obvious blunder or error, and was fast. Claude was much slower and despite me inputting my exact birthdate, kept messing up my age by as much as 18 months. This obviously screws up retirement planning. I also asked some questions about how RMDs would affect my taxes, and asked for some strategies. Perplexity was convinced that I should do a Roth conversion to max up to the 22% bracket, while Claude thought that the tax savings would be minimal.
Mind you, I used the EXACT same prompts. I don't know which model Perplexity was using since the free version has multiple it chooses from (including Claude 3.0).
AGI is when it can do all intellectual work that can be done by humans. It can improve its own intelligence and create a feedback loop because it is as smart as the humans who created it.
No, that is ASI. No human can do all intellectual work themselves. You have millions of different human models based on roughly the same architecture to do that.
When you have a single model that can do all you require, you are looking at something that can run billions of copies of itself and cause an intelligence explosion or an apocalypse.
"Artificial general intelligence (AGI) is a type of artificial intelligence that matches or surpasses human capabilities across virtually all cognitive tasks."
This is a statement that I've always found to be circular and poorly defined for the other reasons I've listed. Any technology that even gets close isn't AGI like I said, it's ASI for the reasons of duplication and time to train.
It is also a line of thinking that will bite us in the ass if humans aren't as general of thinkers as we make ourselves out to be.
This has always been my personal definition of AGI. But the market and industry doesn't agree. So I've backed off on that and have more or less settled on "can do most of the knowledge work that a human can do"
Why the super-high bar? What's unsatisfying is that aren't the 'dumbest' humans still a general intelligence that we're nearly past, depending how you squint and measure?
It feels like an arbitrary bar to perhaps make sure we aren't putting AIs over humans, which they are most certainly in the superhuman category on a rapidly growing number of tasks.
API Opus 4.6 will tell you it's still 2025, admit it's wrong then revert back to being convinced it's 2025 as it nears it's context limit.
I'll go so far as to say LLM agents are AGI-lite but saying we "just need the orchestration layer" is like saying ok we have a couple neurons, now we just need the rest of the human.
> there is almost nothing that a human is better at than Opus 4.6.
Lolwut. I keep having to correct Claude at trivial code organization tasks. The code it writes is correct; it’s just ham-fisted and violates DRY in unholy ways.
I’m very pro AI coding and use it all day long, but I also wouldn’t say “the code it writes is correct”. It will produce all kinds of bugs, vulnerabilities, performance problems, memory leaks, etc unless carefully guided.
The market has clearly passed it by. I was a huge Heroku fan. It even inspired my first startup in 2014 (basically a healthcare tech version of Heroku). At the time, I thought it was the future, and found messing around in AWS, etc., too time-consuming and unnecessary. That was when Rails was all the rage.
I'd even go so far as to say that any competitors that are direct (windsurf, kiro, etc.) aren't even in the same universe. Cursor is just so much better, faster, has better features (plan and debug mode), and squeezes much better results/code out of the same models. They absolutely have some secrete sauce that the other options just don't have.
reply