Hacker Newsnew | past | comments | ask | show | jobs | submit | tomasphan's commentslogin

In my beautiful hometown of Philadelphia they have a novel way of repairing potholes that I've yet to observe in other cities:

1. Do nothing for 9 months. This allows the pothole to mature until ready for step 2.

2. Put a traffic cone in the pothole.

3. After a couple weeks of public notice (traffic cone) dump hot asphalt into the hole, making sure to top off several inches above street level.

4. DO NOT WAIT for asphalt to cool down before opening the street. This allows for asphalt to stick to tires, shoes etc.

5. Make sure to leave a significant bump and don't compact the asphalt so next winter it will open up again.

6. Make sure to put any utility covers (manholes, drains etc) directly in the wheel path for maximum damage.

7. Profit!


Philadelphia is blessed by several feet of sub-street layers (stone fill, belgian block, concrete backfill, and terrible asphalt), embedded rail, pipes, and utilities that are all owned and managed by different local, state, and private entities. Oh and fairly wide temperature swings throughout the year, generous precipitation, salt, and let's not forget the drivers themselves. It's a miracle the roads are in as good a shape as they are - but it does have a traffic calming effect :)

It would take roughly 5000 square meter area to cool a typical small data center heat output (1 MW). Not great, not terrible.

That is a very tiny amount of compute though.

Apparently, OpenAI plan to build 250 GW of computing capacity by 2033.

To put that in space, based on your numbers, that's 1,250 square kilometers of cooling - an area roughly equivalent in size to Los Angeles


That's a lot of weight to launch into orbit

Yeah but these hyperscalers are building data centers that are 100 or even 1000 mW

Right, filtering is the reduction of information while diffusion/generation is creation.

It doesn't have to be a reduction. Swapping the colour channels would be a filter, but it's perfectly reversible.

Luckily we live in a society where its ok to use power for personal pleasure, such as running an A/C in the summer which accounts for much more electricity use than LLM inference.

https://www.eia.gov/tools/faqs/faq.php?id=1174&t=1


[flagged]


> U.S. data centers consumed 183 terawatt-hours (TWh) of electricity in 2024, according to IEA estimates. That works out to more than 4% of the country’s total electricity consumption last year – and is roughly equivalent to the annual electricity demand of the entire nation of Pakistan. By 2030, this figure is projected to grow by 133% to 426 TWh.

https://www.pewresearch.org/short-reads/2025/10/24/what-we-k...

There are ~10M cows nationally. The average energy consumption is ~1000 kWh/cow annually. Summing up, the entire dairy industry consumes ~10TWh. That is less than 10% of the national data center energy burn. [edit: was off by a factor of 10]


Not to mention dairy cows store chemical energy for human consumption, so we got some of the energy invested back.

> One dairy operation uses more resources than all the datacenters in the united states

citation for this claim?

https://www.pewresearch.org/short-reads/2025/10/24/what-we-k...

> U.S. data centers consumed 183 terawatt-hours (TWh) of electricity in 2024, according to IEA estimates. That works out to more than 4% of the country’s total electricity consumption last year – and is roughly equivalent to the annual electricity demand of the entire nation of Pakistan. By 2030, this figure is projected to grow by 133% to 426 TWh.


lol what? Can you please cite some sources for this claim?

Not at all. Agents communicating with each other is the future and the beginning of the singularity (far away).

Seems these thing pop up here ever so often. Either using firecracker or docker/containers. How is this different from the other sandboxes? BTW I love that you got LLM testimonials lol

I'm building an alternative to firecracker here if you're looking for something wayy different: https://github.com/smol-machines/smolvm

we've considered docker, firecracker, will add smol to working roster

context <> building something with QEMU

* required has to support LMW+AI (linux/mac/windows + android/ios)

there are scenarios in which we might spin micro vms inside that main vm, which by default is almost always Debian Linux distro with high probability.

one scenario is say ETL vm and AI vm isolated for various things

curious why building another microVM other than sheer joy of building, what smol does better or different, why use smol, etc. (microVMs to avoid etc also fair game :)


I needed Mac / win/ Linux / iOS / android for dioxus dev, so I built my own in rust.

https://skyvm.dev/


I focus on different design decisions.

Smolvm is designed to run locally, persistent (stateful), long running (efficiency), and interactive.

Worked with firecracker and other options a lot btw, most of everything is designed for ephemeral serverless workloads.


oh interesting our qemu use case is local!

Oh neat!

Feel feel to chat if you need anything, more user friendly docs are at smolmachines.com.


How would you communicate using an offline app?

By ‘offline’ they mean not connected to the internet. So peer to peer communication via wifi or bluetooth or USB or whatever else.

BitChat comes to mind.

Copy files onto Micro SD cards. Smuggle them out of the country or to a contact who has Internet access.

peer to peer RF like bluetooth or IEEE 802.15.4

With your mouth?

You are legally responsible for the actions of your agents. It’s in the name agent = acting on someone’s behalf.

Your English is very interesting by the way. You have some obvious grammatical errors in your text yet beautiful use of formal register.


in terms of "AI": agent is a marketing term, it has no legal meaning

it's a piece of non-deterministic software running on someone's computer

who is responsible for its actions? hardly clear cut


The person who chose to run it (and tell it what to do) is responsible for its actions. If you don't want to be responsible for something nondeterministic software does, then don't let nondeterministic software do that thing.

Hypothetical scenario:

You buy a piece of software, marketed to photographers for photo editing. Nowhere in the manuals or license agreements does it specify anything else. Yet the software also secretly joins a botnet and participates in coordinated attacks.

Question: are you on the hook for cyber-crimes?


Would a general person in your situation know that it's doing criminal things? If not, then you're not on the hook - the person who wrote the secret code is.

You can't sit back and go "lalalala" to some tool (AI, photo software, whatever) doing illegal things when you know about it. But you also aren't on the hook for someone else's secret actions that are unreasonable for you to know about.

IANAL.


IAAL (not legal advice) and your conclusion is generally correct. "Willful disregard" frequently nullifies potential defenses to liability.

You didn't have a reasonable expectation that it would, or even might, do that.

I guess you could say that you didn't have a reasonable expectation that a bot could accept a license, but you're on a lot shakier ground there...


Usually (but not always) there is a knowing element to criminal offenses.

The same as any other computer program: the operator of the program.

Not for long. Tools to create tools will get more popular. My 70 year old mother can login to Replit and vibe code anything (and has thankfully stopped asking me to help her). We’re close to something like that for phone apps and then it’s game over for us devs.


Well Replit supports building mobile apps also: https://replit.com/mobile-apps

Game over, devs


If you want to be that pedantic it does state "This site is satire. Not affiliated with Apple Inc. The bugs are real. The math is not. All estimates are made up. Your frustration, however, is valid." at the bottom.


The bugs are not real (inter alia, Mail app search does, in fact, work). Your being pedantic is my being reality-based.


All of the things on this website have affected my experience of macOS and iOS for a long, long time.

I definitely go straight to the Gmail website when I need to search for anything on my work account. Yes, I've got it set up to cache all the emails locally indefinitely. Have done for years. Even did so when my workplace used Office 365 instead of Google.

On Hacker News, of all websites, giving yourself the "Works on My Machine" badge is not a worthy contribution. It's dismissive of any experience other than your own.


He had it down as Mail search never works at the time of my comment. Never.

Don’t appreciate the drive by “that’s okay because it doesn’t work for me, btw, works for me is bad”. It’s fallacious. I’m not ignoring your lived experience. I’m calling out slop.


And I'm calling out the unearned arrogance of somebody who doesn't understand what hyperbole is, a fairly basic rhetorical device.


What arrogance? Is hyperbole a literary device, or genre? What’s the difference between hyperbole and making stuff up? When is it okay to say it’s making stuff up, when it could just be hyperbole? How is “you said it works for you but I say doesn’t work for me, trumped!” calling out arrogance?

Genuinely curious how anyone can use iOS autocorrect and conclude that it's not totally shit.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: