Best example of this is Claude's own terminal program. Apparently renders react at 60fps and then translates it into ANSI chars that then diff the content of the terminal and do an overwrite...

All to basically mimic what curses can do very easily.

This is the fun part of whole AI-built things that a lot of people don’t want to accept - it really, really, really does not matter if the code “nice, maintainable and etc.”. Does it work? Is it somewhat extendable with AI? Are users generally happy and adoption rate is high? That’s it.

If you have high time preference yes. Over longer time horizons I think the issues with vibe coded software will reveal themselves in the same way badly written software does.

In this case, no, it does not work. CC is full of reported rendering issues, including some that affect me.

You're approach seems straight-forward in theory -- just check every possible move and make sure that none lead to a checkmate. The only issue is that "checking every possible move" is a huge state space (way above what is computable). Not only that, but there are cycles (so you need to deduplicate). And if the game is a draw, then that means the number of moves is technically unbounded (since there would always be a move that makes the search tree deeper), as by definition, there is no way to end the game. So the question is 'when do you stop searching?'. It could be that checkmate is possible, but you haven't searched the 1 in 1 billion part of the strategy tree. In practice, its probably down to some heuristics and a reasonable depth search, but its not formally verifiable. Its a variant of the halting program -- prove that there is a stopping point for this game.

The reason occam's razor works is useful is it draws the one line connecting two points, rather than any squiggle that passes through them.

One other point -- I think the left has effectively shifted the conversation on Israel very quickly. I think immediately following Oct 7 atrocities, public support was overwhelmingly with Israel. By raising awareness of the situation, it has now become more slanted towards "peace in Palestine." I see no reason a similar type of shift couldn't occur on any issue if a coordinated effort to discuss it and raise awareness existed.

And by doing so, it would likely cause change and or discussion by those in power.

> it would likely cause change and or discussion by those in power.

The reason this is an absurd comparison is because on the Palestine issues, it’s a desire to stop using / selling weapons into a conflict and on the Iran issue “causing change” would be starting another war in the Middle East.

> By raising awareness of the situation, it has now become more slanted towards "peace in Palestine."

"the situation" changed from "more than a thousand Israelis murdered by Hamas" to the total destruction of Gaza, the death of tens of thousands and worse.

It's not exactly surprising that there was a shift in where public support is directed.

Sorry I think the GP's point is correct. I feel the same about how we hear very little about modern-day slavery, but lots about much more minor workplace issues in the west. I'm not saying don't discuss modern workplace issues, and don't battle for even better working conditions -- but the silence is deafening. If American children were working 12-16 hour days in sweatshops, it would be nonstop in the news.

By not speaking out, it lessens the moral standing of those making a huge ruckus over certain issues, but remaining silent on arguably far more serious ones.

The power to cause change in democracy rests mostly in influence over decision makers who hold the power and money. The ability to get the news and media and celebrities talking about an issue is what gives protestors and those shouting on the left power to change things. Ultimately politicians and the elites want to be "in the right" to hold onto their power and money.

As an example, suppose 80% of the population was suddenly in an uprising about atrocities in Iran, and the next major election hinged on this subject. If some political party takes the right actions, they win the presidency house and senate. Do you think nothing would happen? Trump has literally said he wants to annex Greenland -- anything is possible if leaders feel they have political mandate.

Sitting in comfortable silence or talking about relatively easier issues just allows the more complex issues to go unsolved.

Again, nothing against pushing for peace for people in Palestine, but claiming that we should just ignore things in Iran reduces the legitimacy of the cause.

The pro-peace activist in WWII, who knew of concentration camps, but never mentioned it, and even told others not to discuss it. They claimed there was no point, nothing could be done. But the legacy wasn't the pro-peace activism, it was denial of the glaring situation they ignored.

while I think this is a fun idea -- we are in such a dystopian timeline that I fear you will end up being prosecuted under a digital equivalent of various laws like "why did you attack the intruder instead of fleeing" or "you can't simply remove a squatter because its your house, therefore you get an assault charge."

A kind of "they found this code, therefore you have a duty not to poison their model as they take it." Meanwhile if I scrape a website and discover data I'm not supposed to see (e.g. bank details being publicly visible) then I will go to jail for pointing it out. :(

I think if we're at the point where posting deliberate mistakes to poison training data is considered a crime, we would be far far far down the path of authoritarian corporate regulatory capture, much farther than we are now (fortunately).

Thanks that was the term I was looking for "attractive nuisance". I wouldn't be surprised if a tech company could make that case -- this user caused us tangible harm and cost (training, poisoned models) and left their data out for us to consume. Its the equivalent of putting poison candy on a park table your honor!

That reminds me of the protagonist of Charles Stross's novel "Accelerando", a prolific inventor who is accused by the IRS to have caused millions of losses because he releases all his ideas in the public domain instead of profiting from them and paying taxes on such profits.

Easy to cherry pick examples and counter-examples. See the luddites for counter-example. Artisans making high-quality textiles are no longer broadly in demand. Lots of pro examples too, I just don't find analogies helpful. It may be that like clothes, there's only so much need for software. We don't really need 1000 browsers or operating systems after all, 3 or 4 good ones is enough (and 90% of people use 1 or 2), despite there being free very good alternatives (unit costs 0, demand still low).

> It may be that like clothes, there's only so much need for software.

Clothing demand has increased greatly in the past decade due to fast fashion. Much of this clothing is designed to cost a few bucks, last a few wears, then get thrown out. It's an ecological disaster.

Maybe we'll see something similar happen with software — as production costs fall, trends will shift toward few-use throwaway software. I highly suspect this is already happening.

Sorry I meant -- only so much demand for people to produce clothing. the thesis "AI will not replace software engineers" could go the same way as textiles (or not). Massive demand for software, no need for developers. Same as I'm sure having AI art will probably increase the total consumption of art/music, but probably reduce the jobs for artists.

> trends will shift toward few-use throwaway software

software has worked this way since the rise of the internet and SaaS. consumers rarely need to install anything locally other than a browser.

All this talk about how you can vibecode all your apps now, "why use OSS?" is making me laugh. Sure for a little website or a small tool, maybe even a change to your somewhat complex codebase that you thoroughly check and test.

Is anyone replacing firefox, chromium, postgres, nginx, git, linux, etc? It would be idiotic to trade git for a vibe coded source control. I can't even imagine the motivations, maybe "merges the way I like it"?

Not sure, but anyone who's saying this stuff hasn't even taken the basic first level glance at what it would entail. By all means, stop paying $10 a month to "JSON validator SaSS", but also don't complain with the little niggling bugs, maintenance and organization that comes with it. But please stop pretending you can just vibe code your own Kafka, Apache, Vulkan, or PostGRES.

Yes, you can probably go faster (possibly not in the right direction if inexperienced), but ultimately, something like that would still require very senior, experienced person, using the tool in a very guided way with heavy review. By why take on the maintenance, the bug hunting, and everything else, unless that is your main business objective?

Even if you can 10x, if you use that to just take on 10x more maintenance, you haven't increased velocity. To really go faster, that 10x must be focused on the right objective -- distinctive business value. If you use that 10x to generate hundreds of small tools you now have to juggle and maintain, that have no docs or support, no searchable history of problems solved, you may have returned yourself to 1x (or worse).

This is the old "we'll write our own inhouse programming language" but leaking out to apps. Sure, java doesn't work _exactly_ the way you want it to, you probably have complaints. But writing your own lang will be a huge hit to whatever it was you actually wanted to use the language for, and you lose all the docs, forums, LSP / debugging tools, ecosystem, etc.

Perhaps this is a helpful model, rather than worrying about the "billions spent" and whether its inference vs training.

How much would it cost you to deploy a model that you and maybe a few coworkers could effectively use? $400k probably to buy all the hardware required to host a top-tier model that could do a few hundred tokens per second for 10 concurrent users? That's $40k per person. Ammortize the hardware over 5 years, thats $8k per person per year (roughly), with no training costs (that's just you buying hardware and running it yourself). So that means, you need ~$800 per user monthly just to cover hardware to run the model (this is with no staffing costs, internet, taxes, electricity, hosting, housing, etc).

So just food for thought, but $200 claude code is probably still losing money even just on inference.

Since they are in the software realm, they are probably shooting for a 90% profit margin. Using the above example, it would be ($800 + R&D + opex) x 10. My guess is assuming no more training (which probably can never be profitable at current rates), they need $20k per month per user, which is why that number was floated by OpenAI previously.

After a hypothetical AI crash, the cost of hardware will plummet. It will suddenly become quite affordable to spin up a GPU or five on-prem to host a couple of models for internal use.

The only reason hardware is so expensive now is to scalp the hyperscalers. Once that demand crashes, the supply will skyrocket and prices will crash.

Fair point -- but my overall point about how much users may have to pay to make these companies profitable stands. Maybe if prices stay depressed for years, but these companies are doing buildout at current prices, and they need to make returns on hardware they are buying now. I suppose they could bank on prices coming down in 2 to 3 years by a factor of 10, then current price ($200 per month) might be profitable (disregarding training, employees, power, etc).

A RTX 6000 pro costs 9k. You can run good models with 96gb of memory.

It can easily serve 10 people or more depending on the overall usage pattern (coding vs everything else).

So now imagine Hardware getting better every year, models getting better too and everything overall gets more efficient.

M1 vs M4 apple increased performance by 100% in 4 years

And there are inferencing optimized chips like groq.

Don't forget kv cache and overall optimizations.

I think your math is off.

And ai is already better than interns. An intern costs you at least 1k per month probably 2k.

For me the math works just fine.

That's a bit different than art. I put it closer to "why do you care if your girlfriend is AI or real? Isn't it just the end emotions you care about?". There is a deep human connection to art, creativity, expression of human emotions and feelings. Reading a poem about losing a loved one and connecting with it, only to find out it was written by a machine is a deep betrayal of that. Its like finding out the love letter you got in school was actually a mockery by the person you had a crush on -- what does it matter? the letter made you feel good right, and that's all you were after. It matters because intention and emotion of other humans matters to most people.

Not everything is purely about being able to output a product and/or produce a tangible good or service. Some things are about people and how people feel.

Another example. I run a charity that takes money, but just generates AI videos simulating helping children. What does it matter? Ultimately the person donating just wants to feel like they made a difference, and they get the same feeling either way, believing the money is well spent. It matters because no one is really being helped, no virtue is actually being enacted in the world.

In the same way, generating all our art and music from AI would represent a massive harm in the world -- effectively extinguishing massive portions of human creativity, and all the people who get to feel useful in creating, editing, and distributing it. In a cold capitalist view, what does it matter, I just want to see a pretty picture for a moment. In terms of actual real value in the world, it is negative and selfish, assuming the only value is my temporary enjoyment of product.

Firstly, thank you for posting this! I'm one of the people who primarily values the art on its own merits, and not on whether it was made by a human chiseling with rocks and ground up flower petals for ink, or an AI generating something. The primary part of that value assessment is definitely how it makes me feel. Your post is the first time I felt I may actually understand the other side.

Speaking only for myself, I can absolutely understand where you are coming from. It makes a lot of sense when put this way. But, I think the difference here is that what you are describing is deceit, and it's the deceit rather than the output, that would bother me in all of your scenarios.

For example, your strongest point in my opinion, is the AI girlfriend versus the real girlfriend. That's a phenomenal argument because it is in my opinion an accurate analogy so how's the logic side strong, and it's also a horrifying one, so it hits hard on the emotions as well as the logic side. The beauty of this is not lost on me, you have created amazing art with that argument! That's the kind of art that really resonates with me.

But zooming in on that scenario, I think the key is disclosure. If the person dating the AI girlfriend knows that it's an AI girlfriend, that doesn't float my boat but I know people who would actually prefer an AI girlfriend to a real one. Again, not for me, but I recognize that it is for some people.

Same with seeing a pretty picture on the screen. If it's being presented to me with deceit behind it, either a person claiming they snapped the photo or made the art digitally when it is actually just AI, then it does ruin the art for me. If it's disclosed though that it is made by AI, I can evaluate it on its merits. Just like in your table example above, I may appreciate the effort and personality behind a more flawed piece that was made by hand, but I also appreciate the precise lines and geometry of a machined output. The key is the honesty and disclosure behind who created it. I get a different value out of the handcrafted piece than I do the AI generated piece. One isn't necessarily better than the other, just different.

Where I do feel a little hesitant on the AI side, though is as you get at the capitalist destruction of art. Without a doubt, the middle level of artists will be hollowed out. I suspect there will always be a place for the traditional artist, but I do worry it will be diminished. On the flip side, I've been able to use AI to take photos of my pets or family, and reimagine them in interesting ways. I know it's not real, I know it's computer generated, and I'm not hanging those pictures on my wall. I simply do not get the same joy from seeing those pictures as I do the originals. I could be wrong here, but I feel like that is the heart of your point, and I think it's a good one.