You cannot walk 100 miles in a day, but anyone with a bit of training can cycle that in an afternoon. It's not a matter of effort or laziness, it's simply not possible.
The stitching comment is just pedantic. People don't hand stitch like a machine not because they are lazy, it's because it's ineffective and impractical. It's not impossible to drive screws directly into softwood by hand, but you'd never get anything done that way. It's not lazy to choose another method, it's just being an effective worker.
As for coding agents, it's clearly effective to write software without them because we have been doing it for decades so far. It still remains to be seen whether they are laziness (ie. it's just a motor) or if they are actually enabling new possibilities (ie. we get software that really wouldn't have been possible before).
Minor nitpick with your first sentence: I do not believe "anyone with a bit of training" is an accurate way to describe what is required for pedaling a bicycle at 20 mph for 5 solid hours.
At a time when I was in better shape than the average person, the most I ever worked up to was 40 miles, and it was brutal. In my entire middle-aged life, I have met exactly one person who believably claimed to have done 100 miles in a day. I'm sure it's easier than many/most people think, but not 'anyone' easy, with just 'a bit' of training.
Well, I should say I live somewhere very flat. I'm certainly not talking about Tour de France style mountain stages. I'm not used to that at all. Also, for 5 hours most people will need to be in a good group/peloton. Doing it solo, like a time trial, is difficult. But I do think most people could get around 6 hours (16.7mph average) within a year of training in sportive conditions (ie. with some drafting, not strict time trial conditions). When I trained my solo average over 50 miles was 18.6mph and I'm not particularly special.
> Can you write 10k LOC a day without a coding agent?
Easily, by writing a code generator. Or disassemble a 1k C program.
> They just allow things that wouldn't be economically possible without them.
It’s not economical to have more code. It’s economical to have correct code and the way to do that is with the power of abstraction. Instead of dealings with opcodes, you have assemblers and macros. Then it’s C, next it’s python, and what would have been a 10k program is now just a few lines.
You can't generate 10k LOC in a day with a coding agent either, not if you want to actually understand it and review it (which you must do if you don't want the quality of your work to go into the toilet).
In my life, the bigger issue are people who are not asking Claude, but asking me.
Because it would take 15 minutes to find out the answer myself, and 3 minutes to let Claude find it, I ask Claude. So how should I answer? Should I disclose that I used Claude to find the answer? Or should I act like I did it myself, appearing to have super-powers that I don't possess?
> Or should I act like I did it myself, appearing to have super-powers that I don't possess?
Do you tell people you used a hammer to hit the nails? Do you go out of your way to explain to people that you don’t use your bare, bleeding knuckles to punch in the nails?
Give people some credit and don’t treat them like idiots.
Also don’t treat your LLM as some “super-power”, it’s just a fancy hammer and still hits your fingers if you don’t know how to use it.
Is it not more akin to asking a handyman to go and do the work for you? They are the ones wielding the hammer, youve just sent them off to go and do it on your behalf without getting your hands dirty. If you then claim to take credit for it, I can see how that might be misinterpreted.
As I see said quite often, an LLM is deemed of a similar ability to a junior software engineer. If I were that junior and my work was repeatedly ripped off by my senior colleague, I would be pissed. Not that an LLM cares in this instance, but the original requester might want to know
That seems like an extreme interpretation if I'm honest. As ever, its more likely somewhere in between your hammer analogy and the handyman. In any case, there is almost zero downside to listing your sources when answering a question as it provides the person asking the question with more context as to the origin of the answer
If you provide sources, provide the sources the LLM found for you after checking them yourself. Otherwise the recipient will have to vet them.
Slight rant: I’ve started to get a bit fed up with people antrhopomorphizing these things, it’s unhealthy. It’s definitely not a handyman and it really is just a tool. I know they’re designed to seem conscious and that just makes them dangerous to use. If my colleague starts treating his hammer as a person, it’s definitely not normal.
I generally provide the source for my information. Even if I already know the answer, I will generally provide a reference on the web or in a document. If I asked an LLM, I will provide that as part of my answer, and I'll generally chip in whether the answer makes sense to me or not; in some cases I'll say that I have no clue if the LLM response makes sense.
Needless to say, I have probably not advanced as far as someone who hoards information, because companies tend to value the information hoarders, but this is just the way I'm designed to operate. I despise information hoarding.
I usually do so I understand what I am replying, but seriously, the latest Opus and Fable versions get it right more often than I would. The time of AI hallucinations is over, as long as they have the context.
Even Mythos can hallucinate[1] with a codebase to analyze. The argument I often see is that humans are fallible, too, but the issues we're seeing in this topic are about those who should know better—including in senior positions—heeding LLM responses/advice over human and concerningly not actually thinking about things at all. Ie: they're not being treated by many as just useful tools/tentative feedback but as authoritative answers/solutions.
There still needs to be critical thinking involved on the human side, even if there's a high rate of accuracy in certain dimensions.
That's a bad example: it did not hallucinate about the codebase (which it had access to), but about a rule outside the codebase (which is not in the context, I assume). Yeah, missing requirements can cause it to rely on things it thinks are true. Same as humans do all the time when they lack context.
That's a very good question that I also don't have an answer to (yet). But I believe part of the reason for that is that were currently still in the middle of this shift, so there's issues of both knowledge about AI and how people relate to it. And of course there's so many different situations here.
Your non-technical parents ask you a question? Use the best tool for the job (might be Claude), then just help them with the task.
Talking to a friend looking for genuine advice? Maybe show them how you use Claude while also answering the actual question.
A lazy person trying to make you do their work? Don't become their Claude interface.
Someone asking a question but actually just trying to open up a conversation? Maybe check-in what their intention behind the question is, and if it's indeed conversation / exploration, leave the phone in your pocket and do that.
I am a software engineer. Most of the times, the people asking are from product management or devops. Devops contractors are the most difficult ones. They have access to the code, they have access to Claude. But they don't use it.
When Google started taking the world by storm, and for a few years after that, I had the same problem.
The way I solved the problem was by making a big show about the fact that I used Google to type in literally what they asked me word for word, and that I looked at the first couple of hits for a few seconds.
Most of them got the message pretty quick. There were a few holdouts though who took longer.
In the end it's really just the same thing as with stackoverflow answers. If you knew how to combine certain keywords for an answer you were searching for, then it took you like 5 seconds to find the relevant SO answer for that. Even then people got mad about it 'why can't they just search themselves if it's so easy to find?' - well, maybe it's not that easy and obvious to a lot of people and that will get even worse now
This was common before LLMs as well. Instead of googling, they ask you, then you google in front of them and show them how easy it was to figure out. At least in my experience, they learn to google by themselves eventually, but there is some type of disconnect in being proactive.
There are typically middle paths, like pointing where they should look approximately (if you are an expert on a project, you typically have a better idea where to look). And then they can look or point their agents at it, or whatever they prefer.
In the past "RTFM" or "LGMTFY" have been responses, so why not something like, well, "I'm not aware of the top of my head, you afaik <brief explanation>. Also feel free to ask Claude on the technical details :)"
"Hey, here's what I found. By the way I used Claude to get this info. It might be worth trying that yourself if you have similar questions, it's quite good at this in my experience. Give it a go"
Give a man a fish, and you feed him for a day. Teach a man to fish, and you feed him for a lifetime.
Literally just ignore these people if they don't get the message. Let them complain to their supervisor and hopefully their supervisor will bring up the fact that you asked them to use AI instead. It's an easily defensible position. Some people need to learn the hard way.
Public embarrassment is also good. I've done it and had apologies off people. They tend to get the message pretty quickly if they're made to look dumb as fuck.
> Let them complain to their supervisor and hopefully their supervisor will bring up the fact that you asked them to use AI instead. It's an easily defensible position.
That is not an easily defensible position at all. LLMs cannot give reliable answers, so it makes perfect sense to ask a human instead of the LLM. Any manager who would take the side of someone saying "just ask Claude" is a very foolish manager indeed.
"Do your own research" is definitely a defensible position. No one is saying take Claude's output as fact.
These people would still ask and waste your time without AI. The fact is, they don't even bother trying to find an answer themselves. And it's often something Claude is extremely good at finding out. Which is why my first suggestion was if you used Claude, then let the person asking know that it worked well for you and encourage them to use it. It's their own competency at their job to judge the output, they shouldn't be pushing the hard bit onto someone else.
If someone uses AI or some other research and then comes to you with their findings and other questions then that is infinitely better to me. I absolutely despise people asking me basic questions that involve me literally doing the research job for them.
If they come to me like "Hey, I have X problem, I used Claude like you suggested and it came out with this response. I'm not 100% sure it's correct because of Y. Would you be able to take a quick glance to see if it's on the right track?". I am much happier with this question because it's clear they have actually tried to research themselves.
Ignoring someone who doesn't even do the most basic research is 100% a defensible position. People should be doing their job.
The problem with asking Claude (or any LLM) is that they will get things wrong much more often than a human, while sounding perfectly confident about the information. It's a recipe for disaster unless you're asking it about something where you have enough expertise to be able to sniff out the BS. So if people are asking you a question, that means they figure you're able to answer the question accurately. If you can do that with asking Claude then whatever, but the person asking you probably couldn't do the same. They would certainly get an answer, but they wouldn't be able to have confidence in its accuracy. You're providing value, even if it's not necessarily immediately obvious to you.
Have you really used the latest models (say Opus 4.8+ or Fable)? When was the last time they gave a wrong answer to a question about the code base? Half a year ago that could have happened. But not today.
Storage of energy in the dimensions required for weeks with almost no wind is an unsolved problem. Right now, we can't store enough to make it through a single night without using fossil fuels. Most countries don't even have the capacity to store the energy for a single hour.
Geothermal and storage is worth investing in, investing in nuclear is not. NPPs are simply too expensive and you won't find investors unless you heavily subsidize it with tax money. Only compnies willing to invest in NPPs are those with virtually free money: the handful of AI big tech giants.
We're building the factories to make them as fast as we can find investors with money to spend on them, because they're already cheaper than fossil fuels which are themselves cheaper than nuclear.
As we can store fossil fuels for weeks of demand, we can store the required energy in chemical form, no? Germany has huge natural gas storage, for ezample.
Iceland is a small area, small population outlier literally sitting on two plates rubbing up against each other and a relative excess of active volcanism and large lava fields.
About 85 percent of the total primary energy supply in Iceland is derived from domestically produced renewable energy sources. Use of abundant hydroelectric and geothermal power has made Iceland the world's largest electricity producer per capita.
Countries like UK, Ireland, Germany, Danmark, Netherland, Poland also have dark winters. Solar in winter is only about a fifth of production in summer months.
Iceland has geothermal, which is amazing, but won't help central Europe.
Those places also have a huge wind resource in the form of the North Sea, for Ireland whatever the other bits of that submerged continental shelf are called, and for Poland the Baltic: https://globalwindatlas.info/en/
This is important because wind season anti-correlates with PV season.
Yet there are whole weeks practically without wind in winter. And most climate models estimate that global warming reduces wind further. Storage in the required dimensions is an unsolved problem.
> Yet there are whole weeks practically without wind in winter. And most climate models estimate that global warming reduces wind further.
Not for the combination of [continent-wide, offshore, the Atlantic coast].
The Atlantic continental shelf isn't correlated with the Baltic.
Offshore Dunkelflaute is much shorter than on-shore.
> Storage in the required dimensions is an unsolved problem.
Also false. The required quantities even of batteries is comparable to electrification of road transport.
Also, independently of batteries, hydro plants come in two flavours that people often mix up: generation, and storage. Of these, only generation requires an associated upstream supply and watershed; storage can be done with a hole in the ground, of which there are plenty, and just the currently and soon-to-be closed coal mines in the EU can be converted to around 13 TWh of pumped storage: https://gfzpublic.gfz.de/rest/items/item_5038282_1/component...
And it's not like we can't dig more holes if we wanted to, we dig holes to get out rocks even though rocks are one of the cheapest things you can buy.
> The Atlantic continental shelf isn't correlated with the Baltic.
How do you transmit the power? Do the countries on the Atlantic even want wind power, set up the transmission lines, and pay for it?
> Offshore Dunkelflaute is much shorter than on-shore.
That reduces the storage requirements a bit, but there is not nearly enough offshore. And offshore is expensive. That's why it's getting harder to find investors.
> The required quantities even of batteries is comparable to electrification of road transport.
That doesn't prove it's possible, but makes it even harder to get the required quantities.
> hydro plants come in two flavours that people often mix up: generation, and storage
Hydro requires the right geography. Is there any potential for hydro left in central Europe?
> EU can be converted to around 13 TWh of pumped storage
Max comment nesting reached. Instead of going through your response one by one, let me just say one thing: the premise and promise of the whole energy transition are based on studies that completely underestimated the physical and engineering challenges in practice. That's why we are where we are. Energy costs have gone up; CO2 emissions have not. I wouldn't believe any study before it has been implemented at scale, at least in a region or small country.
> Do the countries on the Atlantic even want wind power, set up the transmission lines, and pay for it?
They can sell the power, you know. This already happens.
> That reduces the storage requirements a bit, but there is not nearly enough offshore.
It more than covers it.
Even a Dunkelflaute period is reduced, not zero, output. Those seas are huge, and the normal wind there is fast; even on a continental-shelf-spanning Dunkelflaute (much rarer than "just Germany and Poland"), the capacity in those waters *even with that reduction* is more than EU demand just by itself, which in practice means "not building as much as possible".
> And offshore is expensive. That's why it's getting harder to find investors.
Compared to onshore wind. But even offshore wind is cheaper than new-build nuclear, coal, and has strong overlap with new CCGT.
Even then, it's not hard to find investors: the US went as far as banning it because the investors were actually happy to invest.
The limitation on investors right now is that we're transitioning faster than any previous energy transition before, while also having other things to also invest in. It's like how big tech in the US is running out of people to get money from to spend on data centres, the limit isn't actually building them, it's how fast they're getting built when your answer to "how many would you like to order?" is "yes".
> That doesn't prove it's possible, but makes it even harder to get the required quantities.
The fact that EVs are for sale, however, does.
> Hydro requires the right geography. Is there any potential for hydro left in central Europe?
I thought you might say something like that, *which is why I literally wrote what I wrote*:
people often mix up: generation, and storage. Of these, only generation requires an associated upstream supply and watershed; storage can be done with a hole in the ground, of which there are plenty
And then linked you to a document, with a quote from that document, about using coal mines. The document you replied to with doubts, supports the conclusion you've just denied.
Also, I already saw all of Tom Scott's videos when they were new. As per the document:
Finally, the realisable potential includes projects that are not only technically and economically feasible but also aligned with regulatory, environmental, social and implementation timelines, facilitating their implementation within short to medium time frame.
The following table shows that in total 2264 installations have been identified in the EU ETS to be power plants, which have reported emissions in at least one year since 2005. The fuel of 1440 power plants have been identified. Of this there are 124 lignite power plants, 284 hard coal power plants and 2 power plants using blast furnace gas. There are 57 power plants using oil products. The majority of the plants uses natural gas (858 plants). About 824 plants have not been matched to a fuel yet.
Not this time. Go to crookedtimber dot org and you can still see it. Just don't follow the instructions... but clean your browser afterwards!
It even supports Macs. But the Mac clipboard content is just "Oops...".
P.S.: even worse: the crookedtimber site itself is infected. No third-party attack. It registers a service worker on the user's browser that stays even when you leave the site. Be sure to clean up the storage data after visiting that site.
So done a much deeper analysis - there is an loader injected at the Wordpress side which triggers a read of a payload from a smart contract on the Ethereum chain. It stores this in localStorage, registers a ServiceWorker etc so it is persistent.
It then spins up the ClickFix attack - limited to one time per day. I haven't dug into the payload given by the ClickFix attack yet.
For people that have visited while it exists:
1. On Chrome go to chrome://serviceworker-internals search for crookedtimber and unregister the ServiceWorker
2. On Firefox go to about:debugging#/runtime/this-firefox, search for crookedtimber and unregister the ServiceWorker.
If you want to be even safer - just clear all local data for CrookedTimber.
Claude's analysis: "it skips logged-in WordPress users (wordpress_logged_in_* cookie), skips /wp-admin and /wp-login.php, and rate-limits to once per day per visitor via localStorage. That's why site admins don't see it — and why you only saw it on first load."
Doing a bit of digging too - it also filters only to users running MacOS or Windows and skips everything else which is why it will be invisible to the devices I tried it on.
The Amiga was a great computer when it came out, but it was not 10 years ahead of anything. Things progressed incredibly fast back then.
The graphics were not 10 years ahead, not even 3 years. First SVGA cards came out in 1987, with a 320x200 in 256 color mode that was far more practical than Amiga's 4096 color mode, and with resolutions of up to 800x600. By 1995, resolutions of 1280x1024 with 256 colors were widely available, and Nvidia released their first graphics accelerator.
The first SoundBlaster cards came out in the late 80s. The SoundBlaster 32 released in 1995 was far superior to the Amiga.
And the state-of-the-art CPU in 1995 was the Pentium 133, faster than a 7Mhz 68k by orders of magnitude.
The Amiga 1000 was an awesome computer. But technically obsolete by 1990.
The early Soundblaster cards sucked. Even the SB16. The first real comparable product to the Amiga was the Gravis Ultrasound in 1992.
Don't even mention SVGA, with it's pre-VESA non-standardised chaos, and abysmal high res performance. OK, the 320x200(240)x256 was actually quite usable for games, quick enough to run Doom in 93. But for video editing, there wasn't really anything on the PC that even compared to the Genlock, at least not for that price.
And to top it off, AmigaOS had preemptive multitasking (in 1985!), Windows 95 was the first real OS to do that (not counting OS/2 and NT) with graphics (not counting Linux or any unixes), yay, that's 10 years for you.
In 91 I sold my Commodore 16, and got an XT with Hercules graphics and the built-in PC Speaker. Not even on the same planet as the Amiga.
In 92 or 93 I got my first 386, with VGA but no sound card for a while, though I secretly wanted an Amiga 500 instead, but it wasn't my call. However around 92/93 PC gaming and the demoscene (!) really took off, so I got a soundcard as soon as I could, and life was very swell after that.
I think the Gravis Ultrasound is in the same category as the Apple IIGS sound card, which came out in 1986. And I put the Apple IIGS sound card much further ahead than the Amiga Paula.
The IIgs had the Ensoniq DOC (ES5503): 32 oscillators allowing for ~16 stereo voices at the same time. But it only had 64K of sound RAM.
The GUS (that kickstarted alsa on linux) is based on the DOC II (ES5504), which is more or less a 16-bit version of the DOC that could address 1MB of sound sample RAM.
VCF East: The Story of Ensoniq – Al Charpentier, Bill Mauchly, Joe Friel: When Bruce Crockett, Bob Yannes, Charles Winterble and Albert Charpentier left Commodore in late 1982, they named the company Peripheral Visions.
Listen as Albert & Bill will tell their tangled story of how they become Ensoniq.
> By 1995, resolutions of 1280x1024 with 256 colors were widely available, and Nvidia released their first graphics accelerator.
You're mentioning a lot of technologies from 1995 which were goundbreaking no doubt, but lets not forget that that is indeed 10 years later than the A1000, and that Commodore went out of business entirely a year before that.
A VGA card was not in every aspect better than the Amiga custom chipset, otherwise Doom would have happened much earlier. The 256 color mode only really made sense when coupled with a powerful enough CPU which had enough oompfh to update each pixel each frame (for instance Doom's recommended spec was a 486). For some 2D graphical effects, the Amiga could do much more with much less hardware power and for a much lower price.
But where the Amiga really was ahead was in software, it took the PC until Win95 (arguably, until NT4) to implement 'proper' multithreading and to finally get a desktop UI that wasn't a complete joke, and the Mac only got preemptive multitasking with OSX in 2001. Amiga was basically the poor man's UNIX-like workstation when you couldn't afford to shell out somewhere between 10k and 100k for a NeXT, Sun or SGI workstation.
> Amiga was basically the poor man's UNIX-like workstation when you couldn't afford to shell out somewhere between 10k and 100k for a NeXT, Sun or SGI workstation.
Or 6K (1.5K used in 92/93) for a well-specced 386 that can actually run UNIX, multiple DOS versions (including preemptive multitaskers), OS/2, Windows (3.x/NT). And that's just the surface of the mainstream.
> But where the Amiga really was ahead was in software, [...].
You confuse "software" for your specific preferences in GUIs. And software? The PC software catalog in DOS alone towers over Amigas, with a couple of notable outliers (the odd exclusive here and there, especially in cost-effective video production).
> The 256 color mode only really made sense when coupled with a powerful enough CPU which had enough oompfh to update each pixel each frame (for instance Doom's recommended spec was a 486).
Yeah, like 2D platformer Stormlord (VGA version released in 1989) or Red Baron (1990). Played the latter first in VGA on a 286. Ran smooth. Amiga 500? Stutterfest. The game that allegedly sold more 386 machines than any other? Wing Commander. 1990. The late OCS port was technically inferior. Yeah, yeah, I know. 2.5D only counts when one invokes Doom.
The Mac II (1987) had 640x480 256 color non-interlaced graphics. Multiple display support (which I don't think the Amiga ever had). And square pixels!
Much more expensive than the Amiga 1000 of course, but those capabilities had moved way down the Mac line by 1990-1991. And by that time the high-end had 24-bit color, much higher resolutions, accelerated graphics, and laptops (something else the Amiga never had).
The software had moved on too (eg TrueType, QuickTime, AppleTalk).
It was obsolete as soon as the 486 was released, pushing 386 (especially 386SX) systems with SVGA into a lower price tier. Commodore's answer, the Amiga 1200, was released too late with too little (late 1992 with AGA graphics and no changes to the sound, plus no fast RAM?)
I stuck with the Amiga platform until 1994, then I got a cheap 486 after the Pentium was released.
I never had any issue like that, but I don’t use it in German. Not surprised it’s more stubborn on German. After all it’s only imitating people speaking that language.
I wonder what the topic was though. Was it a legitimate request, or just some discussion?
I get your point, but the RSS format (and Atom, and JSON feed..) and the feeds are also partially to blame: there is just no common HTML subset specified for RSS. Some feeds are formatting their posts specifically for RSS. But they are in the minority. Most are just dumping whatever HTML they put on the website. With CMS-specific CSS classes and Tailwind CSS, without declaring them anywhere. Or even requiring JavaScript.
RSS Readers have no chance but remove most styles from the HTML and apply their own, hoping it renders ok.
I think it's an interesting idea, and I would actually buy it, but there's one thing missing for me: a simple API. I use neither Stripe nor GA. All I would need is a simple REST URL that I can ping with the event type, data (amount or number) and my key. Create a small markdown description that my coding agent can read, and I could get it running far faster than by using any integration with third-party services.
Three more comments:
- My main concern with applications like that is memory and CPU consumption. I haven't registered it, but after starting, it consumes 85MB. That's tolerable these days, but more than other menu bar apps I am using, and they have far more functionality. I appreciate small applications being a small footprint.
- Maybe I am not the target group, but I really prefer the more 'family-friendly' sounds. I don't want to explain to my kids why my computer is moaning.
- There may be a market for notification tools that offer a bit more flexibility (like letting users configure the text that's displayed). Or even offering a widget. I am aware of free tools like Übersicht for that, but I don't want to waste 1 GB of RAM for a small 64x64 widget, just because it uses a full HTML+JS rendering engine.
Hello,
Thank you for your valuable feedback!
Those are some great insights, really appreciate it, so yes, i will go ahead and implement your idea for a simple API, probably will ship it in the next 2 days, will msg back in here.
Gonna also look at the memory and CPU consumptions bits more closely and try to optimize to the max.
Regarding the sounds, so there is a range that you can use really, you are in control, you can even override the sounds files and add your own! also i will be adding even more sounds (in total gonna target around 50-100 sounds).
You can also configure the text that's displayed this is already impelmented.
Same with coding agents. I could do it by hand. It just would take much longer. Often so long that it would be impractical.