I gave Opus 5.5 one prompt and six hours to visualize Invisible Cities
quesma.com323 points by stared 10 hours ago
323 points by stared 10 hours ago
A couple of years ago, after reading the book, I sketched some of them on Procreate with an isometric grid, but not all 55. I only arrived at number 4, since each took multiple hour...
https://camillovisini.com/drawing/fc4jc6-le-citta-invisibili...
https://camillovisini.com/drawing/p2yrdj-le-citta-invisibili...
It’s interesting, with a human made piece of art, I’m always drawn to « zoom in », to see how you did little details, if you have hidden something here and there, how you connected different parts.
With AI generated stuff, sure it looks amazing at first glance but I definitely don’t want to zoom in, since I know it’s all a cardboard facade with no passion. Look a bit too closely and you’ll see the bridges that make no sense, the bunch of nonsensical threejs cylinders etc.
I always consider AI-generated content to be a draft — or, in its parlance: something that takes the shape of the final product.
I don't dismiss the ability of models to create. I dismiss their ability to create finished goods. Humans should always be involved, especially with creative works. Models are great at modeling existing creations and solutions — head starts on long roads — but it's the person that needs to walk it and take the journey.
This pairs well with the saying "art is never finished, it is abandoned."
Abandoned one-shot drafts are akin to speaking words but never finishing a sentence before moving onto the next thought. There's no there there.
Are they really? Nothing of value is produced unless the exact artifact is transmuted into a professional good? There’s nothing to be gained from sketching preliminary drafts or notes because they’re never polished for publication?
Not what I meant. The point was work under a certain grain size of effort has very little (or negative) value to others.
We're in an environment awash in 'content' where every person has an infinite slop producing machine, and a lot of people are publishing nearly zero effort drafts, awed by the appearance of meaning.
What happens when you zoom in to a human-made painting and notice an error?
You draw what you saw in your minds' eye. I can guarantee the author of this didn't visualize the cities like they came out.
Since this project (both yours and OP's) seem more for personal fulfillment and enjoyment, I wonder who got more enjoyment out of the process, and also who will remember the results better in ten years.
One is a toy. Created on a whim, it will end up in the bin with other toys. They might pick it up and paw at it later, stir up some memories, and then put it back.
One is art. When completed, they can look at it and be reminded of the journey of creation, a labor of love. It could be put away, or displayed for all to enjoy.
> One is art. When completed, they can look at it and be reminded of the journey of creation, a labor of love. It could be put away, or displayed for all to enjoy.
How is this not applicable to both things? This just reads as gatekeeping to me.
The amount of time or effort put into something does not determine its value.
I'd much rather see yours than the slop version. One of the nice things about the book is that everybody sees the cities different in their minds eye, this link above came from no mind.
That looks pretty nice, I didn't know Procreate could be so good at precise line art!
Love it. I see isometric angles - I updoot.
> Leaving there and proceeding for three days toward the east, you reach Diomira, a city with sixty silver domes, bronze statues of all the gods, streets paved with lead, a crystal theatre, a golden cock that crowns each morning on a tower...
Text in the second person is causing my my D&D instincts to kick in.
If you haven’t read the book yet, I’m tempted to warn you against opening the visualization. Part of the joy of reading this book is letting your mind’s eye run wild, and I fear if I read it for the first time after toying with this site, I’d be remembering the cities rather than conjuring them.
That is the exact problem with the city of Dinipro.
Citizens of this city high in the sky imagine it so fully in their minds that no one there needs to open their eyes to walk the streets and live their normal lives. But if one makes the mistake of opening their eyes and tries to see the city directly, they would immediately fall through the clouds to the ground deep below.
On that note, I'd quite recommend the Coyote vs. Acme movie, some great world building there.
I heartily recommend the audiobook, narrated by John Lee.
The book is a trip and the dreamy sort of narration matched it perfectly. My wife and I would spend evenings listening to one city at a time, on a walk around our little park, one headphone in each ear. I don't know if it was just the environment or the time in our lives, but we both really look back fondly on this book.
I read the book decades ago, then I lost it in one of the many relocations. I always thought I should buy it again but never acted on the idea... A few years ago, I was walking by an antiquarian, and noticed a thin unmarked gray spine. I entered, picked it up, paid, and walked away.
PS I didn't open the gallery. Nice idea but I don't care about visualizing something I should feel.
In fact, having a physicalized 3D representation of the cities is IMHO antithetical to the text. I don't think OP ever stopped to ask if there was more to Calvino's work than 'cool city descriptions'.
I fully agree, thank you for this warning.
How Opus 5.5 imagines these cities is a question for Opus. The real experience is how we imagine - and I believe there are no two people who have exactly the same mental image of an invisible city.
Checking Amazon, several of the available editions are illustrated, which I suppose would have the same problem.
Don't worry about opening the GPT-6 Astra one though, those ones are so bad that your brain will immediately expunge them.
> I’d be remembering the cities rather than conjuring them
Having looked at some of them (and it being one of my favorite books) ... it won't do that. These visualizations are as terrible as they are unnecessary.
Maybe the interactive link in the article for the optic focus can lead you on a fun journey instead. And the bottom-right of every visualization is a link to yet another fun diorama to explore.
It's a techie creating the torment nexus but the imagination version. No need to picture in your mind the invisible cities! Time to make them visible!!!
Talk about missing the entire point of the piece of literature.
I opened it because I had no idea what Invisible Cities meant here. I thought they might be some real cities that don't come up on maps.
The website gives you a good hint before opening up the visuals. So, I got an idea.
I don't really know about Marco Polo. My only knowledge comes from the Netflix series, which doesn't talk about these invisible cities.
I understand why that’s impressive. But I don’t feel much. I guess, I was just used to envy how much people can be absorbed by some idea to spend so much time implementing it. Now it feels like a single shotted nice-typography-and-all presentation at work that someone did 10 min before the meeting.
There's also an eerie feeling of emptiness to the outputs. They lack real intent. Glossy and technically impressive, but they don't tickle your brain. I think it's because they don't really communicate or give you that buzz of new insights. They are pretty but not _helpful_.
For example, I didn't learn about lenses or earthquakes or fusion messing with the controls on these.[1] I didn't know what to look at.
However, there were interactive exhibits at the Children's Museum I spent a long time exploring and loved. That's because they were carefully designed to really teach the concept?
"What we obtain too cheap, we esteem too lightly: it is dearness only that gives every thing its value. Heaven knows how to put a proper price upon its goods; and it would be strange indeed if so celestial an article as freedom should not be highly rated." – Thomas Paine
I've been thinking of this line recently — really gets at the heart of the matter for me. There is no substitute, no shortcut, no alternative for effort.
That might be true but I don't think that's the problem here. I have seen plenty of AI generated images that evoke awe, nostalgia, other emotions - engaging me for a short time, and a few carefully prompted ones that work for a little longer. This fails to relative even these.
Yeah, the demos look really cool but you're thrown into them and there's nothing explained as to what you're looking at or why you should care. Funny, an article explaining something like how a camera focus works with a few illustrations would be far more useful than a fancy 3D interactive.
It turns out the more technology you throw at something does not equate to ease of learning.
Isn’t that at least in part because the intention here was to demo the functionality?
Put in the hands of the educator, they will be able to add those illustrations, design for the light bulb moment, etc, and the tools at their disposal just got much better.
If it makes you feel better.. I find this impressive but it doesn't pique my curiosity. I may get biased, finding a reason not to care.. but honestly I think if this were presented differently I would. And it's not so much about using AI either. I think you could use Ai to make something compelling but so far the one shot prompt outputs to me lack focus
It's low entropy. The amount of entropy is prompt + model but now that you have seen the model various times, the entropy comes down back to the prompt only.
You'd not be amazed by seeing a 2 sentences prompt and thus you are not amazed by this work either.
Another thing to keep in mind is that if you asked twelve different artists to render these cities from their descriptions, you would likely end up with twelve very different renditions.
Now imagine the same thing but each person is using Opus 5.5 - I’d wager there would be a lot of commonality among each of the outputs.
Depends, really. Ten different artists using Opus 5.5 and with some experience with LLMs, outputs may differ quite widely due to:
- Each phrasing the prompt(s) differently
- Each having different history accumulated in context
If you asked me to issue the same prompt as you to write some web artifact, chances are mine would be substantially stylistically different, simply because at some point in the past half-year, Clause started styling everything in LCARS form for some reason.
Whether the differences would be substantial or surface-only? Probably depends, but chances for substance increase with iteration. Real artists don't zero-shot their masterpieces either.
Sure, but I'm referring more to this specific post e.g. "They gave Opus 5.5 one prompt, so it was more like a one-shot".
I’m saying if you provide the same prompt to Opus 5.5 across 12 different people, you’re going to get some similarities that immediately arise.
I’ve mentioned this across many HN threads - I call it the LLM corollary to “garbage in, garbage out”: "generic in, generic out". If you prompt with more specificity, you’re obviously going to get different and potentially better results. But at that point, the artist is doing more of the work.
A while back, I tried a similar test asking for LLMs to come up with engaging D&D puzzles that would fit organically into an underground labyrinth environment. This is a pretty general prompt. Let's just say that the results were about as inspired as a tepid bowl of tapioca.
Looking at the posts on HN AI topics lately, this seem to be the depressing sentiment.
If something that previously took weeks for a human to design, now takes 10 min, that means a whole week of salary went up in smoke.
When have we automated enough?
> If something that previously took weeks for a human to design, now takes 10 min, that means a whole week of salary went up in smoke.
To me, this is the depressing sentiment. The idea that, somehow, the goal isn't to produce useful things but to spread work out over a period of time in order to maximize payment.
At a high level,I conceptualize my job as "doing useful things." Whether that's design, writing code, debugging, sysadmin, CI/CD, writing doc, whatever. I just want to be useful. If there's some tool that makes that easier, I am happy about it. I don't think that tool will result in my getting paid less. But if it does, oh well.
>But if it does, oh well. I'm glad our ancestors didn't take that line of reasoning we'd still be feudal serfs. "The lord has come and taken all our grain again? Oh well, at least I've been useful to him."
If our ancestors worried more about technology obsoleting human labor we'd still be subsistence farmers. Priorizing usefulness is why we're rich today.
Weird way to look at it, from another view we can now print a week salary in 10 minutes. Might devalue the output slightly but we still have the product.
Who is "we"?
This is so often the essential question, and equally often unanswered
Undoubtedly a fundamental reconsideration of political economy is necessary if we are going to go all in with AI. As of now, this technology will exacerbate the existing 'singularities' and wealth disparity distortions in the socio-economic body. And now, basic fucking income is not the answer. Maybe pitchforks.
Derrek Fox, Rupert Hunt & myself of course, who else could I possibly have meant?
The same value was created, and a whole week of free time was made available?
> If something that previously took weeks for a human to design, now takes 10 min, that means a whole week of salary went up in smoke.
The problem also is, if a human took weeks to design this, it would be much better. This is quick but also bad.
This really sounds like a bias selection filter. Humans spend a lot of time on stuff that's shit too.
Every decision a human makes has some level of intent. AI cannot intend, it can only suppose.
The achievement and commitment of others inspires us to reach further and further. Now that's being eroded I wonder what the future will hold.
Yeah, I haven't read the book but these models don't evoke any particular feeling in my glance at them. And I have seen AI-generated stuff that's decent at creating a mood. I suspect that what AI can't do currently is produce a finely tuned mood, it just has the equivalent of "presets". I mean, Hollywood is hardly better than that today as well.
Invisible Cities is my favorite piece of literature hands down. The audiobook narrated by Richard Higgins is also done so beautifully and can't recommend it more.
As neat as this project is, this does no justice to whatsoever to the book. On the surface Invisible Cities seems to be a book about random cities' descriptions. In reality though, this book is about semiotics, meaning, language, and it's limits.
I recently saw a project that intended to turn House of Leaves into a game. The graphics were shoddy and the mechanics were dull, but by far the worst aspect was how it "paraphrased" passages from the book into Claudespeak (allegedly for copyright reasons).
Seeing the same thing done to Calvino (and Weaver's) sublime prose feels downright criminal.
We read this in my undergrad phenomenology class and it blew my mind. I didn't know art could be this rich or deep
Calvino's other creations are good too and recommended. I really enjoyed Cosmicomics!
The first city I clicked on at random had the opening phrase:
> Arriving, you rejoice at its bridges, each different
The model city had like five bridges total, three of which were didn't bridge anything at all but instead sat in the middle of the river and went along its flow, the "bridges" for some reason pretending to be small islands.
Astra Penthesilea has towers that sometimes clip into the green... hedges? Pipes?
Opus Eutropia has a few isekai circle cities with one of them... ON the river, giving no space for ship traffic.
Slop has never been this beautiful before!
Not to mention that the slop site claims that the cities are Invisible, when I can clearly see them.
Like many other people in this thread, I feel like passing what would essentially be a hobby project off to Claude robs the creator of most the fulfillment and gives the viewer less reason to appreciate it. But with the models being as good as they are today, I genuinely feel conflicted. If your end goal is to share something with others, why spend a weekend on a project like this when you could just hand it off to Claude?
Your appreciation should come from the one shot prompt. It's a direct comparison to Astra's one shot. I feel like you and many others are trying to "appreciate" the wrong thing.
If the creator truly wanted viewers appreciate the end product they'd start with the one shot output and spend their weekend iterating on each city either by hand or with Claude. But they didn't want to do that, they wanted to show off what Claude/GPT could do in 6hrs. That's the whole blog post.
This isn't a ceiling, though. This is the new floor: What you can get with more or less zero human effort and ~$100.
I expect we're going to see more and more amazing "hobby projects" that did involve both a few hundred dollars in tokens and hours of human creativity and effort in combination.
But more human effort and say a $.50 pencil would produce something more impressive for sure if the human was reasonably skilled.
Why would that someone want to listen something you haven't created? What have you "shared"? Claude could have made the tools that let the creator build their cities from their own imagination, but that's not what this is.
Kublai Khan received an ornate letter signed by Marco Polo: "In Madrid, City of Lost
Things, no item remains where it was set. If one drops his key in the dirt, he may
never re-enter his home, and, even if he manages to stoop and recover the key, he may
rise to find a tulip garden where his house once stood. In complementary fashion,
things lost by others are forever turning up: A pocket watch on a coffee table. A fond
memory in your recollection. I even know of a prince who turned up in a prison cell.
When he appealed to the guards for his release, he failed to find the crown on his
head, and when he was asked his name, he searched his thoughts, but could not find it.
Indeed, the only hope now for the release of this prince of Spain is for you to send
back 300 ducats for his release. Of course, he will reward you handsomely once he is
out. Yours truly, Marco." Kublai Khan cocked an eyebrow and declared before his court,
"Hey, everyone: Looks like we're about to get ripped off by the guy who traded gold
for paper!" The court erupted in booming laughter.
— Italo Calvino, Invisible Cities 2: This Time It's Visible.Interactive demo: https://p.migdal.pl/invisible-cities-opus-5.5/
Source code: https://github.com/stared/invisible-cities-opus-5.5
Understanding how the models work, is it reasonable to say the generated visualisations are copy or, at least, derived from someone else's work?
Yeah, I was wondering why the two results were so similar, despite lack of detail in the prompt. Both have an array of "miniature" cities on a game-like map, all of them drawn surrounded by a circular border. Something sketchy is happening here. Either more prompt or intervention is going into it than the author claims, or it's copying some preexisting portrayal.
I can't imagine a worse thing happening to a better book.
To Someone who is passionate enough to write about it, do something to build about it, share about it?
unlike someone who is just whining about it?
Wtf are you to someone who has put effort to spread about the book? I never seen/heard of the book, and now I am iintersted in it thanks tot he author.
It's kind of amazing to me, I imagine many here are pro-creativity but all of these responses are basically saying don't create and being negative. I can't get over how rude these responses are.
Creativity apparently is restrained to specific things.
One of my favorite things about the National Treasure Cinematic Universe (two movies and a TV show) is the existence of characters that I like to call "treasure grumps". Their role in the story is to warn the main characters that searching for treasure is a terrible way to spend your time, and can only lead to personal and familial ruin.
Meanwhile, here in the non-fictional world, an estimated 20-30 million dollars have been spent excavating Oak Island with absolutely zero treasure of any sort found.
I was finishing Invisible Cities yesterday, because the book was due at my local library today. I was wondering whether the cities could be adapted into some visual art form. And obviously I thought about feeding it to Opus. 24 hours later, I see this post…
That is what happens in the city of Ternopoli.
Citizens there gather their tools at the bottom of a giant staircase and climb up countless stairs, only to find out that what they were planning to do has already been done by the time they reach the top.
Dissapointed, they descend to grab a new set of tools and make new plans for tomorrow, only for the same thing to happen again.
Very cool, thanks for posting!
And shame on all the people shitting on it because AI made it. If you valued art for art's sake, you wouldn't be so bothered when somebody makes something art-adjacent using some technique you don't approve of. It in no way detracts from the kind of art you admire. Don't be type who just has to let others know that you find something they like to be beneath you.
It's amazing to me how miserable some people can be. Incredibly rude and dismissive to come in a comment section and shit on something creative.
TBH only scanned through ~20, impressive technically but disservice to Invisible Cities to render it as... uninspired, bland, miniatures. All the wrong of details to focus on to capture even notional urban verisimilitude. Looks like a 8 year old drew scribbles and their dad finished the project.
I copied the prompt from the article and inserted it into Claude, here is what it created:
https://claude.ai/share/6bbd421a-147d-435d-bc00-7e93b393b943
(the standalone version at the very end of the chat is the one that runs in the browser)
Well, overall, by faaaaar not that beautiful as what the author from the article received? :-D
(though, it was done with Opus 4.8)
The whole point of the OP's article is: Opus 5.5 is so much better at one-shot design! Also, your link is not public - I'm curious to see your Opus 4.8 comparison
Access granted for all people who requested.
Wondering: I clicked share & copy link - this should be correct?
Does this bring any happiness? for me, No. effort need a purpose.
The uprose could be as little as trying something new. Some of you have seem to lack curiosity and you are jerks, too, shitting on somebody for trying something new.
The demo does not work on Firefox or Safari on the iPhone SE, for what it's worth. I can't scroll down to close the splash screen.
This is probably one amazing positive these models have brought forward - the ability to break out of a single modality, and utilise others to help us learn and visualise. While the visuals here are amazing, my son for example prefers to learn by listening and talking, so we convert lot of his study materials into audio and real-time voice roleplay.
$74 vs $ 10 vs $25 for the same prompt, The interesting number is not actually quality, its what we can get per dollar like 6 subagents runs in parallel.
Yeah for now, this week, on that model, all of which are constantly changing. I would love to know actual cost of running those inferences on such hardware, including electricity, training, etc. i.e. what we should be paying. I can imagine that job would cost multiple hundreds if they were charging enough to actually be profitable companies.
Such a drag- a few carelessly lumped together Ai demos. I mean the headline says it all, there is not much more to it. ok, no thanks!
both the linked camera lens example as well as the last one the author linked runs at ~5fps on vivaldi and pegs CPU 100%. i have never seen any 3d visualization website have performance this bad, i assume they're not using webgl given my GPU is 0% utilization.
First off - I'm a senior dev with ~30 years of experience in graphics engines, game dev, web dev, WebGL stuff, blah blah. I am generally skeptical of AI, but... after reading the top of every thread in this comment section, I feel like people forgot that they could zoom in on the cities to see all the details.
The amount of _stuff_ crammed into each little city is almost unbelievable. Drawbridges open and close, glowing lines sketch patterns, camels walk across a tiny desert. Buildings fade into existence and then burst in a shower of sparks. Flags fly, smoke drifts, buckets cause ripples as they dip into underground pools. The city made of plumbing has tiny people in the bathtubs! There's a freakin' roller coaster with moving cars and a ferris wheel!
I absolutely understand the AI hate that's being commented here, but... in this instance, I just can't feel it. I keep finding more and more little details every time I click on a city, and I am astounded. I just now noticed there's a little guy in the city of strings hanging up new strings at the top of the hill. Damn. My favorite is Marozia, where dark buildings open like treasure chests and release shining towers and flocks of glowing birds.
Yes, there are also a lot of stupid AI errors - every spoked wheel has the spokes off-center for example. But holy crap, I just can't hate this. I don't care about who built it or the random bits of jank - this thing is _beautiful_ in a way that I haven't seen from AI before, and it warms my old GPU-powered heart.
Even as I was writing this I kept finding more things I didn't notice on my first few passes through the cities:
Phyllis fades from color to black and white except for two houses. The fairground has a big top with tiny bleachers and a guy inside. The happy city has people with their arms raised in a V. People in the marketplace city wear hats and some are carrying bundles on their heads. There are weird cactus plants in the hanging city. In the half-fairgroud city, there's a statue on one of the trucks. The city that copies the dead has a set of pallbearers carrying a wrapped corpse down the stairs. The people on top of Argia are laying with their ears to the ground, listening for the animated sound waves coming from below.
And on and on and on and on. I just can't even.
There's a tiny cat on one of the roofs in Raissa! The rats in Marozia hop along their paths! The center of Theodora has bookshelves with individual books!
But who cares about those details if they're just random? Why are they there? What do they mean? What was the intention? Sure it's jaded, but just cramming in elements doesn't necessarily impress, other than technically it's better at not overlapping random stuff.
On the down side, $84 spent on a slot machine might have actually yielded some benefits.
While reading the book years ago, I did not visualize Isidora as a city with 9 houses, like a children’s book, or a little video game level intro, but so unlike Kublai Khan or Italo Calvino.
Yeah. This fails to load both in Opera and in Chrome on Android. I'd tone the hyperbole down a bit.
wow that's awesome! super cool :), fyi so inspiring that I did something similar for d&d forgotten realms, including a timeline: https://narfman0.github.io/realms-atlas/ fable drove it, opus agents did the work, took 30-45 minutes. (intend on extending to planescape, elaborating on significant historical events, and some text to speech narrating cool events)
rock on
It reminds me of Total War. The soundtrack and way that the map is shown. Is Total War an inspiration? Anyways, I liked the visualization. But I never read the book, so I followed (partly) droidjj recommendation
It's one of my favorite books - and I think it deserves better; it's a conceptually very dense text and just slapping together a few 3D models misses the point
it's also not very astonishing that the current models are able to create this - it looks exactly as expected
and I actually think getting an interesting AI generated rendition of the text from a single prompt is a worthwhile endeavor
maybe even a good replacement for the pelicans: it would test if a model can engage with a text beyond the surface level
but this ain't it
Funny note about Opus 5.5.
I once instructed it to spawn at most 20 subagents and it spawned 40, with an adversarial reviewer for each of the 20.
It doesn't follow these instructions very well.
Cecilia looks like an average residential area in Russia.
I just have zero interest in looking at any of these "AI did a thing for me." posts.
The creation really is amazing. I harnet heard of the book but got a long way through it!
Latent space is amazing isn’t it
Loads nothing on iphone lockdown mode fwiw.
I'm a little torn by this sort of demonstration, because while many of the scenes have obvious markers of slop (impossible intersecting geometry, bridges to nowhere, and so on), the scenes largely do work to convey the intended concept/emotion, and the low-poly aesthetic is executed decently well.
I guess I shouldn't be surprised that another visual medium is starting to fall to LLMs after the success of diffusion models for image generation. Nevertheless it's wild that a mostly text-focussed model is building little worlds like these.
What is this for? It doesn't deepen the understanding of the book. The illustrations aren't attractive on their own. They're not a very good representation of the descriptions in the book.
Is creating not enough? Somebody has done something artistic and all you hear is negativity.
this is amazing, wonder how much better it's going to get in a year from now
I ran cat /dev/urandom for six hours but I didn't need to write a blog post about it
Just curious - was the result a 3D animation? If so, I'd love to read your blog post!
If you're going to spend tokens on 3D stuff, then please build me a better FreeCAD first.
"Attention whore" used to mean someone who makes garbage to get attention. Now it's a machine that uses attention to make garbage.
Thanks, I hate it. I imagine Calvino turning in his grave at such an abdication of human imagination.
Well it still has the trademark "colored left-side of the boxes" ai slop, but otherwise looks (and sounds!) very impressive.
Well it still has the trademark "colored left-side of the boxes" ai slop, but otherwise looks very impressive
What a beautiful website, wow. I've been online 10,000s of hours, have seen pretty much everything and this one goes straight in my Top 10.
I'm so glad I'm able to witness this Cambrian explosion of software :).
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]
[dead]
[flagged]
Couldn't your bot even spare the tokens to read the article? The one-shot broke apart after 90 minutes, not 6 hours, even with explicit instructions.
Why, I once encountered a running session longer than a day!
The agent had spun up a backward shell script to watch for the shutdown of another process, but wrote a bug in the script that would have left it running indefinitely until I got home and noticed it.
This was with Fable, no less! And it happened a few more times, though I caught them sooner.
I’m not sure runtime is an important metric at all. Shouldn’t we aim for 0 runtime with maximal results?
How do you get it to spend six hours? I’ve done projects where it would have benefited if it put in extra work.
/goal spend at least 6 hours doing … works
The supervisor agent will keep the session in a loop until 6 hours have passed and eventually the agent will decide to use up the remaining time rather than fighting with it
I wonder how much co2 is being thrown into the atmosphere everyday through the steady stream of “look what I made this LLM do” and endless “benchmarking”?
Sigh
One useful hint here is the API token cost - in this case "about $74 in API tokens" for Opus 5.5.
We know Anthropic run inference at a margin, so that $74 means that the cost of the electricity involved is substantially less than $74. I'd love to know the actual cost there.
I wonder how you only reach $74 in 6 hour runtime. I would have guessed this runtime goes into the hundreds of dollars.
Or I'm realizing that I'm overpaying my tokens...
A lot of that 6 hours isn't spent waiting for the models to return tokens, it's spent executing the code that they've written.
Approximately nothing - the carbon cost of electricity is already very low, and most of the inference costs are GPU and datacenter amortization. You could require all AI inference to be carbon-neutral and that'd barely raise the API prices.
I wonder how much co2 you use to get through the day, and how much you use to remark on other people's creations in a way that suggests you disapprove of their utilization of their available resources. How much co2 do your projects emit, since you seem to be interested in those metrics?
I strongly suspect one HN post (or two now) is not even in the same ballpark. I could ask Claude to figure it out but that would be a complete waste of time and energy. ;)
This has given me a great idea: I have a modest proposal on how to reduce net CO2 emissions while not reducing token consumption at all
USE carbon dioxide to get through the day? What?! We're not plants!
I assure you that if your body did not emit co2 throughout the day you would not continue metabolic processes for an especially long period.
> I wonder how much co2 you use to get through the day
What a terrible thing to say to human! Being (presumably) human, it is their inalienable, natural-born right to do so.
Shame on you.
Intuitively I'd say that just-for-fun AI projects emit orders of magnitude less CO2 than what just-for-fun road trips or air travel emit.
So unless you can show that it approaches all the other frivolous consumption that humans love to waste resources on, maybe we can get back to experimenting with cool technology and talking about it, on this website called "Hacker News"?
[flagged]
> The days of techies gatekeeping these experiences are over.
Why do people keep using the word "gatekeeping"? "Techies" made open source. Anyone could have taken the time to learn what they wanted, the tools and the resources were available.
The disdain for tech will always baffle me. Yes, this stuff is hard. A lot of it requires a certain level of skill. Even now, you still require some level of skill to get good results.
No non-corporate tech person has ever wanted to prevent another soul from using their work or learning their craft.
The top-level comment feels a lot like the perspective of somebody who never cared to try.
With robots for agriculture, delivery, and food processing, the days of farmers gatekeeping food will be over
Also, the "techies" are naturally going to be a lot better at pushing AI to its limits. This doesn't change the playing field - it means more people can do more things, but the upper end of the skill range is going to be just as far ahead, if not further ahead, of everyone else as it always was.
> Anyone could have taken the time to learn what they wanted, the tools and the resources were available.
False in reality. Just ask any dev who has had to listen to family and friends' app ideas over the last decade. The good news for them is soon they will never be asked again.
Having a concept is completely different, their unwillingness to dive into implementation cannot be blamed on the "techies". I would wager that the same people who were pitching unrealistic app ideas then are still not trying to make apps using AI.
What AI has largely changed is how the communication barrier (or perception thereof) is dealt with between two sides: one which consists of those who perceive gatekeeping or attribute their lack of substantive progress to it, and the other side which consists of the time, resources, and skills the former feels entitled to or indebted towards.
I’ve had a running theory for a few years now that some tech folks are more open to receiving an error message from a compiler or other tool, than they are to critical feedback from a peer. If they’re blocked and finally have exhausted their other avenues (pre-LLM: web searches, forums, etc), they’ll reluctantly engage with a peer, and do so with extremely limited patience.
The first paragraph of this reply is extending that theory to non-tech —-> tech, but replace compilers with LLMs that _appear_ to be more useful (or at least busy) working with what limited context you provide them, especially when it’s insufficient.
It won’t quite tell you your app idea is unrealistic or that the implementation is nearly impossible, but after burning a bunch of time and tokens, those people will eventually get the point.
Or reluctantly reach out to a tech person anyway.
This is more comparable to a mass produced plastic replica than the handcrafted work of an artist. Experience and skill are not orthogonal to creativity or taste.
The examples were a one shot. The creativity hasn't even been applied yet.
Experience and skill is what gives you taste. Creativity is nothing without it.
And scarcity gives things value. No one bats an eye at a mass produced cotton t-shirt with a detailed custom print on it. There's services online where you can order a thousand with whatever print on it for quite a cheap price.
But take one of those shirts 80 years back in time and it'll likely impress everyone.
Perhaps that's what's going to happen to software too. If you're an idea guy you might get a monkey paw situation, at least if you planned on making some sort of serious income from your software idea.
Now, you just need to use someone else's words and someone else's AI built on everyone else's stolen art! But sure, yes, thats you being creative... Don't use your imagination at all.
There are three main categories of AI ethical dilemmas for me:
1. IP infringement
2. Climate impact
3. Cultural impact
I think that 2 & 3 apply to all AI right now - artistic, coding, writing, whatever. However, I don't think that #1 applies to everything. Code has been openly shared for as long as it's been possible to do so. There have probably been cases where AI trained on closed-source code, but I hardly believe that was ever necessary at any point to get where we are.
My point in saying this is that your critique of AI is valid but I think a little less so here.
#1 is way more concerning to me
That's fine, I'm only making the point that it's not really relevant to coding. Programmers have been outright copying each other's code for decades to the chagrin of nobody.
I totally understand. I'm a developer with almost 20 years experience myself and it has taken me a while to come to terms with this too. Fortunately, I made enough during the good times to see me out once my job ceases to exist. I'm sorry for everyone who didn't.
Or, you can use your imagination building on top of the sum of all human knowledge distilled into a language model. Your choice.