In haven’t seen that as a limitation because the agents also are able to grep for keywords to find files to explain things. So they don’t necessarily have to ingest the whole codebase into context.
Oh man, can anyone imagine a non-Terminator scenario for this?
Update: I'm not saying people shouldn't develop this, we're never going to squash human curiosity. But when I see this kind of stuff, I'm deeply troubled by how bad actors (state and non-state) will use this.
I hope our security services are working hard on countering these potential threats.
Im more worried about these type of things causing us to blast each other and ourselves back to the 1920s or so during conflicts when small explosive EMPs start being viewed as less damaging than drones and robots. A fast explosive on the back of a neodynium magnet and a few coils of copper can make a hell of an EMP blast. The only reason we don't use them now is due to all the collateral damage, but if drone bombs represent even more damage they become viable. Yeah it will destroy all the radios around and fuck up a bunch of expensive equipment, but you would still have soldiers with guns rather than just smoking craters.
Nobody really uses much undirected EM warfare in my opinion because it represents a huge escalation in a war, similar to the use of indiscriminate chemical weapons, or even nuclear weapons.
It would be devastating in the local battlefield, potentially frying radio or other equipment depending on the size of the device or how close you could lob it towards the enemy before going off; but with the low wattages many non-military communication devices use today you would also be blasting horrible noise to all of them beyond the local area and disrupting communications across potentially multiple neutral countries.
It would be a large act of aggression against any countries around them and NATO, and at scale possibly even piss off far away countries like the US and China. Especially large EMP devices could even be temporarily misidentified as a nuclear explosion and gain the immediate full attention by any nuclear powers watching out for it.
An EMP large enough to be an international incident is a nuclear explosion. The effective range of one pumped by a conventional explosive is very small.
For undirected EMPs, consider that a lightning strike is in the region of about 5 gigajoules, and the difficulty of pumping that much energy through a coil.
My expectation is that a CHAMP warhead (not missile, warhead) would have a EMP range of something under 300m (probably less), even then vs. unshielded electronics (compare the effect on your body of being in a car that's struck by lightning vs. being on a 100m away on an open field that's struck by lightning), and only even that much by having a high gain (highly directional) antenna: https://en.wikipedia.org/wiki/Counter-electronics_High_Power...
That said:
> but with the low wattages many non-military communication devices use today you would also be blasting horrible noise to all of them beyond the local area and disrupting communications across potentially multiple neutral countries.
You could do EMP, but you could also do some sort of point-defense turret. Drones are lightweight and fragile, so it doesn't need to be big - just fast and auto-targeting.
Didn't they try this in Ukraine and it doesn't work? Any point installation is quickly overwhelmed. The only answer to FPV drones so far seems to be more FPV drones. Though they're not using fully-autonomous drones in Ukraine yet, so that might still play out.
“[As of May 2025] Ukraine has developed and successfully tested the Sky Sentinel – an AI-powered, fully automated turret designed to shoot down Russian drones and missiles… the M2 is known to have an effective range of 1.5 kilometers against airborne threats. Each unit costs approximately $150,000. Developers estimate that protecting a city would require 10 to 30 turrets… Given that each Iranian-made Shahed-136 drone used by Russia costs around $100,000, Sky Sentinel offers a scalable and cost-effective solution to a persistent and deadly threat.”
Yes, a wall of $150,000 bushmasters with some servos can take hurl enough lead in the air to protect a city from a single gulf-war era $100,000 lawnmower engine on two meter wings bumbling in a straight line at jogging pace.
We’re currently in the “a shipping container on a semi can could launch dozens of $2000 racing quads with molotov cocktails zip-tied to the bottom with enough agility to thread a needle faster than a turret can swing its own mass”
And the writing is on the wall for some near-future “any nation state could drop sci-fi cluster bombs that shed ten-thousand 250gram racing quads that can overwhelm even the most advanced point defence just by numbers and it’ll be cheaper than a conventional 2000lb bomb”
>with enough agility to thread a needle faster than a turret can swing its own mass
If it has a range of 1.5km, then the drone needs to be moving really fast in order to move a couple of degrees per second.
If it's going like 80mph, then that's 36m/s, which comes out to <1.4 degrees/s at 1500m away. For 1000m, you get ~2 degrees/s. Not to mention that this velocity is at a right angle to the turret, and you have to close the distance.
The point is that you can send lots of tiny drones to converge on the turret. They have to cover 1.5km (though that's only the distance if there's no cover, i.e. the gun is on a perfectly flat plain with no trees, buildings, etc). If the drones are travelling at 36m/s then that's about 50 seconds to cover 1.5km. If you drop 2000 drones, you'd need to kill 4 each second. That seems like a lot.
I feel like search and rescue after an earthquake where a drone swarm can canvas and categorize if it saw movement or not is one possible "non-bad" use.
Fire departments and police in Germany are deploying more and more drone units, too.
Firefighters use them to search for missing persons but also to get aerial images and a better overview of larger scenes as "running around" is often not possible or doesn't help that much with the overview.
Police is using them to take pictures of accidents. It's easier to see tire marks and the whole "history" of an accident from above. Really reduces their time on a scenery to take pictures of everything.
If by burrito you mean shaped charge high explosives with lethal shrapnel, triggered by facial recognition, delivered by drones the size of house sparrows at the speed of sound, then yes, burrito delivery.
Same as others have said, with consumers wanting to try 20 different models all the time, you have to reduce the friction of trying out the model.
I hope this is just a huge mistake that they aren't allowing anyone to actually try it.
Word of mouth from HN and others (not just advertising a link or press release) is how I've started to use about pretty much every single AI feature I use.
Assuming this isn't a mistake, it says that this company has the wrong management/leadership structure if they think they can sell a brand new developer focused coding tool without letting actual developers try it. Maybe we don't know something?
Apart from that, just fix broken stuff. Practice like any other skill, like others have said.
Like decide what skills to learn based on what’s broken.
Need to solder something together? Buy the soldering iron. Need to figure out which capacitor you need to replace with? Learn how to identify capacitors and navigate digikey.com
Risks to consider in general are (also of the top of my head don't take as exhaustive):
* Electrocution
* Burning yourself
* Setting stuff on fire
* Fumes, both from soldering and overheating/burning plastic
There's a kind of balance between habits and awareness. Rely too much on awareness and you will miss some safety issue during a complicated repair. You need good habits, but rely too much on them and you won't notice when you finally make a mistake.
Those apply during repair processes. It's also possible to render a device unsafe to use, for example by damaging a li-ion battery or by a 'bodge' repair that circumvents a safety feature.
A tip on how to reinforce good safety habits taught to me by my shop class teacher is to refer to any dangerous tool you are using by the kind of injury it is likely to inflict. The lathe is the deglover, the angle grinder is the eye splinter injector, the welder is the retina tanning machine, the soldering station is the IQ diminisher, etc. This helps to put you in the mind of "I need the eye splinter injector for this task, how shall I go about avoiding getting splinters injected into my eyeballs?" instead of getting complacent.
Are you trying to mix two payment models together though?
To provide a counterpoint:
With almost all traditional employment, people are paid for their time. Their output is more an expectation based on the time put in.
So if you want to bill by output, that's always an option as a consultant (fixed price).
But as a business owner, I would absolutely expect my employees (and vendors) to use these tools to go faster, just like laptops make them go faster than typewriters. It's not like it's causing the employee 4x more effort for that 4x output, is it?
If you were digging ditches by hand, but now you have an excavator with you, are you expecting 10x payment just because you can now do it 10x faster? Nobody will pay for that. Instead, you would now charge by a different metric (say, length of the ditch, rather than hours). And you would now be able to handle more customers since you can provide the same outcome to more people in less time.
I have no problem if someone wanted to charge by outcome. I think the concept of paying for "time" is going to go extinct in this new world. Think of lawyers who bill $500/hour. Now they can't say "I need 4 hours to have my team review this" and send you a $4k bill for their whole team to bill, because expectations have changed.
But framing this as an "us vs. them" thing by comparing apples and oranges isn't going to get us anywhere.
> With almost all traditional employment, people are paid for their time. Their output is more an expectation based on the time put in.
Yes and no. It's true that employees are rented for a limited amount of time rather than owned like tools. However, employees can earn vastly different sums of money for the same amount of time. Some make $7 per hour, while others (a privileged few) make $7000 per hour or more.
> I would absolutely expect my employees (and vendors) to use these tools to go faster, just like laptops make them go faster than typewriters. It's not like it's causing the employee 4x more effort for that 4x output, is it?
Ok, but are you going to let employees who use these "labor-saving" tools work 1/4 of the hours per week as before for the same compensation as before? I suspect not.
> But as a business owner
> But framing this as an "us vs. them" thing by comparing apples and oranges isn't going to get us anywhere.
Of course that's what a business owner would tend to say.
My point is that there's no inherent reason why the benefits of "increased productivity" would trickle down to employees. Individually, employees may look at A.I. and think it's great, but collectively, if all employees have access to the same tools, then employees don't benefit from A.I., because it doesn't fundamentally change their relationship with employers.
Think about how students started using A.I. to cheat and get ahead. This turned out to be short-sighted, because everyone has the same access to ChatGPT, and thus A.I.-fueled cheating became ubiquitous in school, and the cheaters are back where they started, no longer ahead of their peers and no better off than before. Indeed, they may be worse off now, because potential employers have to expect that all college graduates are incompetent cheaters who didn't learn anything.
This is not true at least with the current crop of AI.
I have seen 4-5 people use the tools (won’t say where and in what capacity, but only one of them was an employee of mine) and the results were wildly different.
One person just copy pasted AI responses and added no value. But that meant they also copy pasted substandard responses and it actually reduced the quality of their work (in fact, their greatest asset was client management)
Another one was a coder, and even though he was using windsurf every day, he was getting stuck on the most basic items. And having access to AI made no difference in his productivity because he doesn’t have the base knowledge and experience to be able to discern what is a quality suggestion and what isn’t. And he doesn’t know what questions to even ask the AI to help unblock him.
I actually think these tools will WIDEN the gap between what a good engineer can do, and what a mediocre one can. And it will widen income gaps. I’m not going to say this is a good thing. I’m still a human being.
For example, with regard to students, it was the mediocre, lazy students who started using ChatGPT to cheat. The good students didn't need it. And as I said, the cheaters are no better off than before. They mistakenly thought they had some completitive advantage, but the same tools were available to everyone.
With regard to engineers, the issue isn't the difference between good engineeres and mediocre engineers. The issue is that all good engineers have access to the same AI. Thus, engineers who were equal to each other before AI are still equal to each other. It's not a competitive advantage over peers, and thus there's no incentive to pay them more. AI doesn't change the supply of good engineers, nor does it give individual good engineers a competitive advantage over other good engineers. Any additional productivity comes more or less for free, because it's due to factors external to the engineer rather than internal to the engineer.
In the end, it's all supply and demand. That's not to say it has no effect either. Increased productivity from AI can increase wages by raising the upper bound for salaries (companies won't pay an employee more than the value he can create), and creating competition between companies (if a single programmer can generate $500k worth of value, and their salaries are only $100k, then we'd expect other companies to spring up to arbitrage this difference).
> companies won't pay an employee more than the value he can create
True, but the wealthy corporations aren't simply paying for "programmers". Cheap programming labor is available, especially in less wealthy countries, and corporations could avail themselves of that if they wanted (and they sometimes do). When some programmers make more than their peer average, it's because the employers consider those programmers to be "the best", the smartest, most skilled, hardest working, or whatever. In that regard, A.I. is largely irrelevant. The best programmers and the worst programmers can both use A.I., so it's not a distinguishing factor in the labor market. What matters the most is the supply of the type of programmers that companies are seeking.
A.I. might (or might not) increase a company's revenue, but there's no reason the company can't just book that as profit. A.I. doesn't increase the demand for programmers; if anything, it might reduce the demand for programmers, if the programmers they've already hired become more productive. Adding more employees has diminishing returns, as we've seen with mass tech layoffs in recent years. Hiring is not an automatic productivity boost, if you're measuring productivity as an addition to revenue. At the end of the day, producing more stuff doesn't matter unless the company can sell that stuff.
> if a single programmer can generate $500k worth of value
How do you even determine the monetary value produced by a single programmer in a large team effort?
(Ask yourself this: how much "value" is generated by human resources or administrative assistants or janitors? You might say $0, but nonetheless they get paid more than $0, because the work needs to be done. You can't a functional corporation without them. And their compensation is based on supply and demand, not on "value generated".)
Let me put it this way: I used to work for a small company, and some of my coworkers also worked at a BigCo at some point, either before or after SmallCo. Their personal productivity didn't change much over time, from one employer to another, but their employer's revenue per employee changed dramatically from one employer to another. A tech company's revenue depends on many different factors, many of which have nothing to do with the employed programmers.
Let me put it another way: the more that employers rely on publicly available automatated tools—for example, LLMs—rather than on the individual skills of employees to increase productivity, the more the employees themselves become interchangeable cogs in the machine, and thus by supply and demand, their compensation will not increase.
Of course, if using LLMs were an elite skill that required inborn talent and/or years of specialized training, then you would expect LLMs to increase the compensation of those with that unique skill, but I've seen no evidence that this is the case.
> But as a business owner, I would absolutely expect my employees (and vendors) to use these tools to go faster, just like laptops make them go faster than typewriters. It's not like it's causing the employee 4x more effort for that 4x output, is it?
s/these tools/drugs
And before you claim that it's an "apples to oranges" comparison, let it be known that we have no data on the long-term cognitive effects of offloading thought to machines.
> But as a business owner, I would absolutely expect my employees (and vendors) to use these tools to go faster, just like laptops make them go faster than typewriters. It's not like it's causing the employee 4x more effort for that 4x output, is it?
what this means is that the business owners capture all of the benefits of the increased productivity (which is what we've been seeing for decades)
>what this means is that the business owners capture all of the benefits of the increased productivity (which is what we've been seeing for decades)
Not really. Even though labor's share of GDP has steadily dropped for the past half century, it's on the order of 4% for the past 70 years. Real GDP per capita on the other hand has more than tripled in the same time period.
Because there's an enormous and growing gap between top compensation and lowest quantity pay, a better comparison would be to look at _median pay_ as a share of _GDP per capita_.
"business owners" by definition, means capital. AI engineers, quant traders, and bankers making 7 figures TC doesn't magically mean they're "business owners" in "business owners capture all of the benefits of the increased productivity".
You can keep seeing this through an adversarial lens. And I could just respond with some flippant remark like “well, they’re the ones that pay the bills” - but all that does is create a false battle between parties that have a shared interest (that of making a living and providing value)
Either way, I do care tremendously about the power of efficiency. That can sometimes lead to better profits, but more likely leads to happier customers and more business, which helps the employees.
Maybe because of supply and demand, but not because they bill hourly, which is exactly my point. It won't be simply because the excavator can get it done 4x faster.
I mean that the price multiplier isn't based on the productivity multiplier, as the parent was implying. Of course, the cost of the excavator is built into the price too.
Also for smaller jobs (like with coding), I might now be able to hire a mini excavator (using AI as an analogy for the excavator) myself for a weekend and just do the job myself, and leave the bigger jobs to the pros. That's already happening (look at RR Buildings on YouTube, two guys build massive structures with the help of tools).
My guess is that adding "fuck" would've changed the show's potential returns and content ratings, which is a pretty big change when projecting revenue, ad sales, etc.
Rather than "hey I just wanted to add one word and they pushed back"
Or it could just be a certain way of framing things?
But for the benefit of this debate I'll state my bias. Almost all my travel is as a group.
But in most other facets of life, we save by buying more, right? (buying wholesale, buying bulk, etc).
So I've actually had the other feeling... if I'm buying 4 tickets at once, can I get a bit of a volume discount? And I'm not sure that's what the airlines are doing here (I don't ascribe any altruism, it's probably more that families were getting cold feet with increasing prices), but I like that I get some kind of cheaper rate when I'm buying 4x of something, just like in just about every other purchase in life.
But it’s a “solution” to a nonexistent problem. The vast majority of stuff people own sits unused for decades but gets moved around to whatever home you’re living in out of irrational loss aversion.
This is human, and fine. But I’m 100% certain this author is grossly overestimating the value of the junk they believe they are “saving” and how much of it they will actually use. This is the rationalizing of a budding hoarder.
The time pressure does these kids a huge service by forcing them to clear out stuff that doesn’t actually matter so they don’t feel the need to buy a 4,000 sq ft McMansion to store their college bean bag chair and every other piece of junk they’ve ever owned.
Ultimately, the authors children will run across these “salvaged goods” in a decades-untouched pile in her basement upon her death in about 66 years.
What's more wasteful, throwing away mass produced items that still have 50% of their life left that you won't actually use (but are rationalizing that you will because 'free')?
Or heating a 4,000 sq foot home for 60 years to store said items in unused piles and then having your children eventually throw those items away anyways when you die?
Realistically it's going to be 1 of those 2 scenarios.
This is the rationalization that I've used successfully before.
Every item takes space, and space costs money. Space often costs quite a bit of money. If I'm paying $2000/m for a 750sqft apartment (not unreasonable for many metros), then every square foot costs me ~$3/month minimum (often quite a bit more because I also have to heat, clean, and organize it as well).
How many months of storage does it take to mean you're literally losing money by keeping an item?
For old shoes? Not that many... if you haven't worn a pair of shoes in a few years, you've likely spent more than $100 storing it.
For bad furniture? Not that many... it takes a lot of space. That crappy couch no one wants to sit on anymore can be costing you almost $100/month.
For that old bag you overpaid on and don't even like that much? Now you're just lighting more money on fire for every month you hang onto it.
---
Mentally shifting the equation to a more realistic "opportunity cost" model for items is healthy. Posts like the top level article are, respectfully, a chock of bullshit. Those kids might have messed up when they originally overpaid for a luxury item, but I don't believe for a heartbeat they're messing up when they choose to let it go.
Go work in a place that takes donations, and you'll quickly see they do exactly the same calculus. Goodwill might take your whole bag of donations, but they will absolutely throw 70%+ of it away right off the bat.
They can't afford to try to reuse your crappy stuff for months, quietly losing money storing it.
The nicer consignment shops won't even pretend - they'll just flat out refuse to accept most of your junk. It's not even worth sorting through.
There is absolutely an opportunity cost for all of the stuff you own. I won't publish my entire thinking on this, but after seeing my parents collect, hoard and store things for years and years, I place a high value on not having something (I tell myself that I am letting the store hold it for me.)
I still have too much stuff and its a fraction of what my parents had.
My experiences volunteering at goodwill and habit for humanity ReStore.
A bag will come in full of items. A small subset of that bag will be pulled out to go on shelves. The majority of the bag will be sent to "salvage" (trash/recycling).
Some folks have a curated selection of items they bring in that mostly get accepted, but usually the folks doing that just take it to a real consignment shop, or sell it on something like facebook marketplace.
Generally - a large portion of the donations are exactly this situation: used items in a big trashbag that come in after a move (or eviction [or death]).
They go in the trash. Where the owner probably should have put them a while back. I also don't like how much churn exists in our modern consumption economy - but it's not doing favors to pretend it doesn't exist. Much of our "stuff" is low cost, semi-consumable items that will end up trashed - by design. I don't like it either, but saying "Donate it" is like pretending recycling is going to solve plastic pollution.
Gotcha. I think I had a different situation in mind where the things are already pre-selected. If there's bags of garbage or stuff nobody needs, sure those would have to be disposed of. Totally depends on what the input is.
Sure, and that's fair and I'm not saying you shouldn't donate like that (frankly - that about the only way you should donate if you want to actually be helpful).
I think the reality is that online marketplaces basically fill that gap, though.
Most folks who have items that they know aren't trash try to sell them themselves online these days (or best case - use things like "BuyNothing" groups).
But if you can't move it on facebook marketplace, the sad reality is that goodwill and the like are also probably going to put it in the trash if you donate it to them.
Even places like homeless shelters are very picky (exclusively "new" or "gently used" donations of specific items).
There's just a lot of work in "matching" used items to people who want/need those items, and most times the value of the item is lower than the value of that work (even when that work is heavily discounted by volunteers).
I have a bunch of stuff that someone might want (or probably more realistically think they want) but there is a very real limit to how much effort and time I'm going to put into connecting to that someone.
What really bothers me with the article---something I think I share with top-level commenter dpkga---is throwing away perfectly fine articles of clothing (i.e., "tennis shoes"). Not to mention, expensive. These, presumably wouldn't cost a fortune to transport even overseas. I get it that maybe the student had to make a choice between, say, bulky books/notebooks/school work which may come in handy for the future but that's where the "expensive" part comes in; why even buy something expensive if there was any chance you'd discard it before its service life is up? (Other commenters have provided answers elsewhere that I consider plausible.)
The furniture (i.e., "bean bag chair") I can totally understand why they'd discard it. The only thing that bothers me then is, will they buy a similar item for next year? Because if they will, then this stuff doesn't "doesn't matter" and therefore the problem actually exists, if only because it feeds into the mindless consumer attitude which leads to over production of goods that end up in landfills if not in someone else's hoard pile.
But then, having a dorm "thrift store" is not a bad solution. Let some curated amount of stuff stay behind and be available to other students. Some stuff needs to be just trashed by responsible adults but otherwise, a university or dorm can find some space.
It just seems that some such organization never get to this.
Better yet, offer to store a reasonable amount of students' belongings over the summer. My college offered this my first couple years but stopped by the time I graduated in 2010.
Yes, we had summer storage undergrad for dorms. Still put too much energy and effort into moving stuff around. In a more digitized world, I like to think there would be less paper etc. and I maybe wouldn't even have a big stereo system.
They dont want a bunch of students hanging around with no school and extra time on their hands. Agree its way to short, as someone that had to deal with it, but also understand the trade off they were making at the time.
I've always wondered... Making agents edits (like vibe coding), all the tools I've tried (Cursor, Zed, VSCode) are pretty equal since most of the brains are in the underlying models themselves.
But the killer app that keeps me using Cursor is Cursor Tab, which helps you WHILE you code.
Whatever model they have for that works beautifully for me, whereas Zed's autocomplete model is the last thing that keeps me away from it.
What do Cline users use for the inline autocomplete model?
Cursor changed their pricing recently and now charge a 20% markup on LLM API calls to use their "Max" models (which from what I gather are the full extent of the LLM context windows you get in the API anyway).
I also don't love that Cursor generally plays "context compression" games since they have an incentive to keep their costs minimal. I just don't love any of these tools that try to be a bit too smart as a middleman between you and the LLM (where smart is often defined as 'try to save us the most money or be maximally efficient without the user noticing').
Cline also tries to be smart, but it's all for the benefit of the user. I like the transparent pricing -- you bring your own API key so you're paying the underlying API costs without a middle man markup.
Am I being pennywise and should I just use Cursor directly? Maybe...but I've found my Cline results to be generally better for the more complex queries.
I actually haven't explored the inline autocomplete space that much yet. I really like Cursor Tab (or Supermaven or even plain 'ol Copilot) -- they all seem really good. I'm sure there's a Cline equivalent out there, but I don't know what it is (yet).
Has it improved lately? I used it for a year and until I switched to cursor, I felt the AI coding stuff was a complete sham, because copilot was essentially useless.
reply