Have people been dropping cursor usage for Claude code? I have dropped to using cursor as just an ide with auto complete. Curious if others are doing this too.
Cursor’s autocomplete is SuperMaven (which they acquired).
From the site :
“Supermaven uses Babble, a proprietary language model specifically optimized for inline code completion. Our in-house models and serving infrastructure allow us to provide the fastest completions and the longest context window of any copilot.”
LLMs are literally auto-complete models. I just so happens that when your auto-complete model gets big enough, and you poke it in the right way, it accidentally pretends to be intelligent. And it turns out, that pretending to be intelligent is almost as useful as actually being intelligent.
Claude Code makes me feel like I'm dispatching a legit engineer to go get something done. But they come back in a minute instead of a week. Most of the time the solution gets the job done. Sometimes it introduces too much complexity, sometimes it's totally wrong, but it gets the job done. Cursor meanwhile just feels like shortening the (copy editor/paste chat/copy chat/paste editor) loop.
For $200/month you can get equivalent value to a team of engineers. Plan accordingly! The stack is no longer safe for employment. You need to move up to manager or move down to metal.
It’s a good question. I think a better benchmark than the current options is “go make $X dollars this quarter.” Right now the models fail this miserably. Claude can’t even run a vending machine inside Anthropic HQ. So there is still some kind of strategic activity that comes naturally to humans that LLMs struggle with. I know the big conundrum is “scaling solves this in the next N years” but my bet is that N > ~20 in this case.
Last week stripped out all CSS from a fairly substantial project and replaced with Tailwind equivs, it got all but a few cases right
That was gemini-cli, I could see some mistakes on trial run so created a GEMINI.md with system prompt and project description (about 50 lines) which clarified some tricky source layout situations
Second run it was fine, ran for about an hour or so -- I had attempted to do it manually a while back but it started to look like it would take a week or two
Thanks for the insight. I have seen similar uses at work, where people do a bit of an enhanced codemod to migrate code from using one deprecated thing (library, function, syntax) to another. And while a codemod has to be more exactly programmed. AI gives you the ability to cover spots in the code that may not 1 to 1 fit with what the pattern you had in mind.
For the trivial cases that's fine (just using LLM does same)
But this particular project is not like a standard site and the CSS is in small fragments across 100s files and uses constants for some things like color values in places too
In that Loopple example you can see the conversion uses the Tailwind arbitrary value notation, the -[], so background-color:#afa8af gets converted to bg-[#afa8af], but I wanted nearest pure tailwind class bg-zinc-400, the agent seems to work out color distance fine so does all that in one-shot too
That's good to know it is better at translating the code from using one style to another! It is one of the gold use cases for AI agent coding at the moment. I've seen that at work as well.
Yes, and cities with lots of trees are way more livable due to this. Planners in our town seem to hate trees with a passion, thank god we‘re moving away from this concrete desert.
Trees in cities are expensive to maintain, which is why they're often on the chopping block when budgets get tight. This is especially true in places like Las Vegas where there is little natural tree cover due to the climate. You have to have a staff of arborists to keep the trees alive in such a harsh environment.
Indeed, the town I‘m living in was 80% destroyed during WWII and that still shows in its finances. It’s amazing how long major disaster affects a region. Big drug issues, highest cancer rate in the country etc
The problem is that to get these effects you need large canopies of trees, and to get that to happen the trees have to take the space of something else. For street trees it takes away land from parking or traffic lanes; for properties it occupies both horizontal and vertical square footage since the sky above the tree needs to be clear. These are unpopular with some political affiliations and interest groups.
Even without any evaporative effect, the air cooling of leaves (at least bringing them to the surrounding air temperature) happens more easily than that of concrete pavement due to height and larger surface area. The concrete can easily get heated much hotter than the air at even 10-20ft.
Wrt. water consumption - Mediterranean species like say olive trees are kind of optimized for low water consumption, by for example having leaves covered with wax-like stuff decreasing evaporation.
A lot of normal consumers pay $20 a month for ChatGPT. I think most software gets bid down in price bc the marginal costs are zero. Where it’s not (llm token generation) prices don’t plummet and consumers build a different expectation.
this might be a good thing if viewed from the opposite perspective: people with kids/elderly parents usually can't afford to pay as much per person as people traveling alone for fun/corporate travel.
The oil example is very compelling for import substitution. And the covid example is interesting in showing the savings rate only went up as an offset of gov spending.
I'd love to see a follow up on (a) is it important for the US to increase domestic savings and (b) what are the best policies to do so, and why are they the best?
I imagine blanket tariffs might actually increase the savings rate because they increase the cost of importing all goods when the domestic alternatives are either inferior or more expensive. But I'm curious if they are the best way to achieve the savings goal.
The only policy that will increase savings rate is a stable or depreciating currency. People are incentivized to use an inflationary currency so they can maintain value.
savings are a thermodynamical impossibility. real wealth decays (livestock will die, the roof over your head will leak, the bushel of corn will rot, ...). savings must be invested for it to have future value.
Money was traditionally a way to store value, and not all things decay at a rate that matters. Roman roads still exist. The Parthenon still exists. Roman coins still exist. In any inflationary currency value erodes, and it also encourages the production of less durable goods as time pressure encourages speed of production and not durability.
It's circular. Sure it's pegged to metal, but won't tell you how much corn or land or homes you can buy with it. What should one oz of gold, hoarded in 2025, be able to buy you in 2050? Many factors will determine that. Theres no such thing as fixed value, unless the definition is self referential.
The only thing to do is turn present day savings in capital, it's the only claim one can have on wealth in the future.
Whatever people buy. If you are looking at the dollar value of gold you have to look at what a dollar would buy that year. That is the value. You will find that a similar amount of gold buys the same amount of things throughout time regardless of the dollar price.
Another way to say that is that the dollar price of gold is correlated with the cumulative inflation of the dollar0 over time.
In 1900 a 20 dollar coin contained 0.9675 ounces of gold. An ounce of gold was legally defined as $20.67. No free floating gold price. A dollar coin contained 1.672 grams of gold.
1900 $1 is equal to 2025 $172
So a single dollar today will buy 1/172 of what it would in 1900. That is inflation. Not an inflation adjusted curve. Just the drastic devaluation of the dollar.
A single dollar today will buy 1/172 of what it would in 1900, as long as what you're buying is gold. If you are buying anything else, though, your number is not relevant. And that means that your number is useless, because it's only the number if you're buying gold, and gold is almost never what we actually want to buy.
The curve suggests that there would also be periods of inflation and deflation, for instance of a market basket of goods and services, under a gold standard.
In a perpetually inflationary environment it functions that way. Stocks become a universal savings account. Everyone pouring money in raises values whether or not the company being bought has any real value.
(a) There's an argument that people should save more for retirement, but I haven't heard anything more than that about why domestic savings as a whole has to increase. If anything, this is quite a good place to naturally run a deficit! Good rule of law and investment opportunities, as well as future earnings from migrants.
(b) Targeting the fiscal deficit usually works well, especially because it's particularly yawning right now. Forced savings (sing-style CPF) work ok too though, although only Singaporeans wouldn't consider that a tax.
Both forced savings and taxes are legally mandated by the government, but that does not mean that forced savings are taxes. Implementation details matter.
Your money in your own CPF account accumulates interest (at decent/attractive interest rates that generally exceed inflation rates), and is then paid out tax-free to you after retirement.
Additionally CPF funds are managed separately from the government's consolidated revenue. They are administered by the CPF Board and are not used for government expenses in its yearly budget.
Sure sure, my point is stuff like British national insurance and American social security are considered taxes, even though much of the money in expectation goes right back to you in retirement / health spending.
In Singapore, the flows are similar, even though the accounts are broken out individually and the top-ups are explicitly done.
In my view, the key is the government telling you what to spend your money on that gives it the shade of taxation. Whether they do so with labeled accounts or not seems more of an implementation detail.
> Whether they do so with labeled accounts or not seems more of an implementation detail.
I think this is an implementation detail of huge significance.
In the CPF system, your current contributions pay for your own future retirement. In the Social Security system, your current contributions pay for current retirees' retirement.
The CPF system is sustainable, because it truly is a savings scheme. The U.S. Social Security system is not, because it relies on having a tax base that never diminishes.
> Having access to goods and service is what gives you a better living standard
I think the good faith critique is access to imports can be taken away by the other country if they want. eg. rare earth metals. So being too heavily reliant on imports without the capacity to produce domestically is less long run access
I mean China has cut off rare earths from time to time, and from time to time we don't see crippling shortages but a rather quick supply response.
If you're concerned about short / medium term timeframes, I've yet to see a broad analysis that showed stockpiling (can even do it privately!) being insufficient.
Yes. Too much of a good thing, or a good thing at the wrong place or instant of time, ends up not being a good thing.
I would recommend against having sex in a subway station at rush hour, or drinking French Cognac during a job interview, although both are good things.
We can and should discuss how much trade deficit, and the nature of it, but in essence, it is still a good thing if you don't owe to other countries money in a currency you don't control to have this deficit.
Except the US has a realistic protectionist policy it can use: defence production. It's an industry which is diverse, naturally demands locality, but can also provide an export market.
And very much was a core US growth export till very recently.
The open source ranking library is really interesting. It's using a type of merge sort where the comparator function is an llm comparing (but doing batches >2 for fewer calls).
Reducing problems to document ranking is effectively a type of test-time search - also very interesting!
I wonder if this approach could be combined with GRPO to create more efficient chain of thought search...
The article introducing the library has something about how pairwise comparisons are most reliable (i.e. for each pair of items you ask an LLM which they prefer) but computationally expensive. Doing a single LLM call (rank these items in order) is much less reliable. So they do something in between that gives enough pairwise comparisons to have a more reliable list.
> Texas is the only state in the lower 48 that has no major connections to neighboring power grids. That means growing energy demand in Texas must be met by new power generation in Texas.
> "Texas didn't [build solar farms and battery storage] for an energy transition reason at all," says Rhodes, "We just made it easy to build things here. And so people started building things here."
It's hard for me to reconcile the criticism of Texas's market structure with the fact that they seem to add more (clean) generation capacity than any other state.
Is it as simple as: Their state is growing very fast (both people and electricity per person due to datacenters) leading to supply/demand forecasting difficulty and mismatches?
TX generates 138% more electricity in 2023 vs. 2009 [1]. By contrast CA generated 5% less electricity in 2023 vs. 2009 [2] (roughly same for any 2 years around those start and stop years).
I do wish this topic were treated less politically and more pragmatically. At this point everyone seems to agree we need more electricity in the future with less carbon intensity and so far, TX is one of the few states actually achieving it at scale.
The absolute amount of clean generation mostly just correlates with the amount of electricity demand.
Texas uses 475 TWh / year of electricity, California 251, Florida 248, Ohio 149, and down it goes.
Meanwhile Texas only has 33% carbon free electricity vs 49% in California which are still below outliers like Vermont 80%, South Dakota 81%, and Washington 83%.
Unfortunately, WV and Kentucky are coal country with under 7% carbon free.
Yea. It's trite, but a lot of things are literally bigger in Texas. It's easy to do really misleading stats around Texas because it just does more of everything than the rest of the US. It can both build more green energy and still pollute more than everyone else.
California is also significantly bigger by economic output while using half as much electricity. There's probably some hidden factors there, like Texas engaging in more electricity intensive industries than California does or California using more fossil fuels for things Texas uses electricity for (this one seems improbable), but it seems like a very strange discrepancy.
I have to assume most of the electricity use in TX is cause it's over 100 degrees in most of the state for most of 9 months every year, and a lot of California barely gets over 75 for more than a few weeks. Texas lives and dies on air conditioning, which is a nearly fully electrical use. Colder places use energy, but it can be a mix of electricity, gas, propane, and other options. AC is almost always electricity.
Maybe there's also an industrial angle, but you can live in CA without power almost all year. There's a lot of Texas where if the power goes out for a few days, all your food is spoiled and you're probably about to die from heat with almost no alternative.
I had my AC go out in October in Texas and it was 88 degrees in my house within 6 hours. A lot of houses and apartments really aren't designed for natural ventilation in the way that you'd want in order to be able to naturally cool a space.
It’s a bunch of things. Home solar and insulation requirements isn’t directly showing up on these comparisons but they matter. In 2023 California got 19TWh/y from home Solar vs 3.7 for Texas. https://www.statista.com/statistics/1419901/us-residential-g...
The California as a wide range of temperatures, directly next to the ocean is milder assuming the wind is from the west, but it gets much worse with El Centro seeing average July high’s of 107!
https://www.currentresults.com/Weather/California/city-tempe...
PS: A less commonly considered advantage for solar is rooftop panels provide shade and by converting sunlight into electricity it effectively lowers their albido reducing local heat gain.
I mean, that table says the average high for almost all of is in the mid to high 90s. That's not that far off from what I said.
Additionally, sure, California has plenty of desert that's as hot as Texas, but most of the population is in LA, San Diego, or San Francisco, which are all coastal and relatively temperate. Most of Texas's population is in Dallas(97), Houston (93), San Antonio (95), or Austin (97) which are all generally in reliably hot areas. Houston may have an average summer high of 93, but that's also coupled with a brutal humidity.
The equivalent California population centers are SF (66), San Diego (75), LA (82), and Anaheim (87). All generally livable temperatures, especially given their overnight lows are in the 60s. Most of Texas has summer overnight lows in the mid to high 70s.
Yes, Texas isn't that much hotter than the rest of the US south, but it's also insanely more populated. Houston with it outlying suburbs nearly has the population of all of Mississippi put together. For the most part, large quantities of people don't live in the other really hot parts of the south except for Florida.
If the argument is temperature and therefore AC alone explains the drastic difference in per capita electricity use the numbers just don’t add up. If half of Texas electricity use was AC and all of California stayed 75 the sure however…
I agree there’s a big costal population in CA, but those city temperatures are really deceptive as even just the suburbs of those cities get hot. San Bernardino which still is part of the LA metro area and just 50 miles from the ocean jumps to 95f in July. The closer to the ocean you get the cooler things tend to be but most of the greater metro area is in the 90’s not 75f.
Just looking at the Central Valley Fresno, California population 550,000, 97F in July. Sacramento 525k 95f Bakersfield 420k 98f. And that’s not counting a host of smaller cities and towns or the southeast.
Something like half of the population of California see’s 95+ temperatures in July where they live. So yes San Francisco being surrounded by the ocean on one side and the bay on the other stays surprisingly cool, but it’s just not that much land, 875k people live in SF proper and most of the state is very different.
> But if you're going to claim that most of Texas only drops below 100 for winter, you're going to need a citation.
I didn't intend to imply that. I'm not insane, I literally live in Texas. I know it gets under 100 during the summer. I'd be shocked if anywhere in the world spent a single 24 hour period over 100 most years.
I was being hyperbolic, but not by much. In 2023, Houston saw 97 days over 95 degrees. Dallas saw 84, San Antonio 110 days, Austin 109. For most of those days, the low is still over 70. It gets to around half the year in most of the major population centers if you lower it to 85 degrees. Los Angeles only rarely gets over 100 days over 85 and SF almost never sees more than 100 days.
The practical effect is the same - you generally run your AC nearly 24/7 to keep your living areas below 80 degrees from April through November. It was already in the high 80s last week in a lot of the state.
In each of your examples, you show 3-4 months of days over 95. But yet you said "over 100 in most of the state for most of 9 months of the year". Not by much would be a fraction, not a difference of 200-350% of your example.
Just looking at electricity is unfair though. How much of Vermont's, South Dakota's and Washington's heat come from clean electricity and how much is natural gas? If you want to compare apples-to-apples you have to look at things like residential carbon intensity separately from industry.
Careful once you start adding elements individually you’ll tend to focus on those supporting your viewpoint rather than the total. Electricity is after all used by more than just residential homes. If you want to look at “Total Energy” in a more brand context than electricity that’s not just heating and electricity it also includes gasoline and industry etc.
So on average people in Texas drive 16,171 miles per year vs 10,949 for Washington state etc. Similarly the oil industry both flairs and burns a great deal of fuel in order to crack longer hydrocarbons into shorter ones etc.
However, in the widest definition all of this is just a rounding error. It’s actually plants on farms which have the largest energy supply/demand by a huge margin, but being free we rarely consider it in such calculations.
Yes, but a state like Vermont is going to outsource 100% of the carbon intensity it takes to produce gasoline and natural gas to another state. Then use a very small amount of electric energy compared to fuel and gas. Comparing only their electricity usage is very disingenuous.
Vermont is exporting a great deal of the electricity it generates as well.
A nationwide map of carbon intensity by demand would have very different numbers in the electricity column. Such a viewpoint is reasonable, but you need to maintain consistency in definitions for such comparisons to be meaningful.
> TX generates 138% more electricity in 2023 vs. 2009 [1]. By contrast CA generated 5% less electricity in 2023 vs. 2009 [2] (roughly same for any 2 years around those start and stop years).
Isn't some of that because Texas has no major connections to other grids? As your quote from the article points out that means that to meet increasing demand Texas has to build more power generation in Texas.
If California, which is well-connected to other states' grids, needs more electricity they can buy it from other states.
> I do wish this topic were treated less politically and more pragmatically. At this point everyone seems to agree we need more electricity in the future with less carbon intensity and so far, TX is one of the few states actually achieving it at scale.
That's good but its not the whole story. You've got to also look at the mix of renewables and non-renewables. Texas is moving in the right direction, but other states are too and many are farther ahead.
In 2022 Texas produced 139 TWh of renewable electricity [1], which is more than any other state--about 50% more than second place (Washington at 88 TWh)
But Texas also was #1 in non-renewable electricity production at 386 TWh. (Second most is Florida at about 240 TWh).
Percent of electricity from renewals is probably a better way to compare states. Of the top 10 electricity producing states California has the highest renewable precentage at 43%. New York and Texas are at 28% and 27%. The rest of the top 10 is a few around 12-13% and the rest in the 3-10% range.
CO2 Mt/TWh is also useful as it can give some idea of how dirty a state's non-renewables are. For example Delaware and Montana emit 505 and 504 Mt/TWh CO2, but Delaware is only using 3% renewable electricity whereas Montana is using 52% renewable. That suggests that Montana's non-renewables are quite a bit dirtier than Delaware's non-renewables.
Texas is 406 Mt/TWh. Combined with their 26% renewables this suggests that their non-renewables produce around 556 Mt/TWh CO2. There is way dirtier out there (Wyoming at almost 1100 and West Virginia at around 960 for example), but of the top 10 electricity producers only Ohio matches it. California is 380 and Florida is 401.
> Isn't some of that because Texas has no major connections to other grids? As your quote from the article points out that means that to meet increasing demand Texas has to build more power generation in Texas.
> If California, which is well-connected to other states' grids, needs more electricity they can buy it from other states.
Well, that's not quite right, in order to meet increasing demand the increased capacity still has to be built somewhere, but it doesn't have to be built in California.
Either that or you could make the other states use less electricity.
Is that CA number because more properties are generating lots of Solar PV and so need less from the grid?
You still have the problem, especially with PV, of the duck curve that requires other producers to ramp down (and potentially be paid for ramping down).
Texas only adds more clean generation because it’s way less capital intensive than building a natural gas or coal plant. Those plants require $500M+ minimum and the returns just aren’t that great. My wife is an energy attorney in Texas and handles power purchase and interconnection deals like these all day long.
Solar and wind deals require far less capital, go up faster, and aren’t subject to the supply risk of natural gas or coal.
Texas also has a lot of clean energy thanks to sun and terrain. The Edwards plateau creates some of the best wind generation opportunities in the US.
Texas also attracts energy heavy industries because it has relatively cheap power. Which we’ve learned partly results from not paying anyone to have excess capacity… which is all fun and games until you have winter storm Yuri roll in and your only option is to “shed load” which btw kills some people.
Another aspect of Texas is that we have demand response contracts whereby certain users get paid simply for the ability to “take” power when required. This is very attractive to bitcoin miners. Prices here go negative from time to time which is pretty wild.
All of this attracts a lot of energy-intensive industries to Texas.
> Texas is the only state in the lower 48 that has no major connections to neighboring power grids.
This may be true but it makes it sound like there's no connections, which is false. There are interconnections between other power grids but depending on their definition of "major connections", it may be technically true.
>This may be true but it makes it sound like there's no connections, which is false
It sounds more like there are no major connections, which is what a plain reading of that sentence says. And as you said, that's true, there's no alternate interpretation where interties that can carry around 1.5% of the grid's load are "major connections".
And you're ignoring the fact that the lack of interconnections is a deliberate political decision, not a technical oddity. By keeping their grid independent, the Texas grid can escape regulation by FERC. The limited interties they do have are used for used for scheduled and emergency power trades and are not treated as interconnections supporting interstate or international trade.
Except:
They had to interfere with the free market to force electric suppliers to harden their infrastructure against cold weather -- something they hadn't done before because of the race to lowest operating costs.
There's also other horror stories from under regulation:
> they just trusted that supply and demand works…and were correct.
Except when they weren't, like the sibling comment about 2021 winter and all the hoopla over needing to force them to invest in winter proofing; after a winter failure which cause hundreds of deaths due to multiple days long power outages.
And the more current rolling blackouts during the summer that no one is happy about but have little to no recourse for. It's so much fun having to reset digital clocks on appliances every morning. Or, working late and the unexpected joy of having the power grid tell me it's time for bed because the power for the entire neighborhood, including street lights, will be off for the next 3-5 hours. Hope you have good battery powered fans on before you went to bed or enjoy waking up in a humid mess.
There should be a high sales tax on all advertising. It's the perfect thing to tax:
- Luxury product companies advertise the most (utilities rarely do). The tax will decrease consumption of the most economically "unnecessary"
- Advertisers charge more money to show ads to rich people than on poor people - even a flat % tax will naturally impact the rich more than the poort
- Prices are set by auction at the maximum of the advertisers willingness to pay - this means tax revenue will hit the supplier of ads not the buyer
- Tax on ads is fundamentally a tax on consumption not income
- It's a zero sum positional product - ads make use of finite physical space/attention. This is close to a negative externality that people cannot contract away (eg i cannot contract the billboard company to not show me ads for $5/mo). On the web, this could technically be done, but the marginal value is so low and the number of billboard providers so high that transaction costs prevent it from happening
I am actually willing to have space ads if there were a super high tax on it. Imagine to show a message to all of humanity for a week, you must first pay $100B of taxes to all of humanity. That might be worth it.
I've been surprised there's less discussion of advertising taxes than much more complicated things like land value taxes, even though advertising/marketing is 1/5th of GDP in developed markets.
The first few paragraphs of this article suggest the structure is not "here's a well-reasoned economic argument for why ad taxes are utile", but "I hate Google and similar companies, so we should screw them over with punitive targeted taxes".
Paul Romer has received the Nobel Memorial Prize in Economics. While that's just an appeal to authority, it still has more substance than dismissing a long post without even having read the majority of it. I encourage you to read the whole thing before concluding that there is no well-reasoned economic argument in this text.
This would explain o1 poor performance with problems with variations.
o3 seems to be expensive brute forcing in latent space followed by verification which should yield better results - but I don't think we can call it generalisation.
From firsthand experience, this simply cannot be true. I can give them totally novel and unique physics problems I just made up- that requires tracking the movement of objects through a series of events, and it answers most correctly. Moreover, they find analogies between disparate concepts and fields of study and make useful suggestions based on them- which is arguably the same process as human creativity.
I think ultimately the disconnect is people theorizing about what it can or cannot do with an incorrect mental model of what it is, and then assuming it cannot do things that it can in fact do. The irony of discussions on LLMs is they more showcase the limits of humans ability to reason about novel situations.
Yeah, because if the AI boom becomes the AI bust, we'll have another 2008-level economic crisis on our hands.
The investments into AI are in the hundreds of billions (maybe even more if you factor in the amount of people studying and researching AI), but the returns are in the tens of billions (if even that).
If you exclude the "growth" coming from the industry sniffing its own farts (e.g. Nvidia selling insane amounts of insanely overpriced GPUs to InsertYourFavAICorp), the actual amount of "useful goods and services" produced (api accesses, chat subscriptions, ai-enabled app growth etc.) are tiny compared to the investment levels.
The AI train appears to have no brakes. A massive crash or AGI are the only options now. Both are going to be bad for average humans.
Their methodology shows they can create an infinite variety of problems.
This is the same thing as synthetic training data.
It doesn't matter if models are trained on the output of the generated data or not. If the model ends up being able to solve newly generated variations, you'd have to admit that it understands the underlying problems.
I think what it shows that it has minimal "understanding" of the problem - otherwise such small variations wouldn't pose a challenge. Training it to handle these specific small variations doesn't change that.
If it were a complete failure on variations I would be inclined to agree. Instead it was a 30% drop in performance. I would characterise that as limited understanding.
