Korkstand

quote:

---

So, what happens when people average 45 in a 55? Does that perhaps cut the capacity of the road? I'd like to shut up the "Speed limit means I'm allowed to go a lot slower, you can never be too safe," crowd.

---

It does cut capacity but only a little bit. Safe headway is about 2 seconds in either case.

The bigger problem with going 10 under is the speed variance it creates which leads to more accidents.

quote:

---

The edge computer only sends lightweight METADATA back to the central server. Instead of sending a gigabyte of video,

---

I'm not even considering video at all in my AI ideas, but yes vehicle detection and tracking from video should absolutely be done at the edge. Then discard the video.

quote:

---

it sends a tiny text packet that essentially says: "Northbound lane has 5 cars, average speed is 35 mph, pedestrian waiting to cross."

---

Yeah so what I'm getting at is who/where is that data sent to? Certainly the next intersections in each direction, but that information is also useful to other intersections city-wide.

quote:

---

Because of EDGE AI, the central city server doesn't need to be a supercomputer. It just acts as a dashboard collecting tiny status updates and coordinating the broader grid.

---

It's the coordinating of the broader grid part that is a lot harder than it might seem. Keep in mind, if your intersection is emitting tiny packets of data, it will also be consuming tiny packets of data from multiple other sources. And all those tiny packets will have to be consumed by some machine and considered as a whole before emitting new packets to the edge.

Consider an accident 10 lights away... how should your intersection handle it? We want to avoid creating a mile of congestion that cascades down every cross street, right? Maybe the optimal strategy is to start delaying traffic within a certain radius from approaching the accident and accelerate traffic moving away from the site of the accident. Prevent congestion in the area by slowing vehicle accumulation and speeding dispersion. Responders can get to the accident and do their jobs faster, making it better for everyone.

As for the traffic we intentionally delay, do we let their navigation re-route independently, or do we send signals to coordinate their new routes to optimize? It's a lot to think about.

quote:

---

I want AI to take over all traffic light programming and timing

---

I wouldn't say I want it to "take over", but it should absolutely be used as a non-authoritative layer in the system.

quote:

---

What makes proper timing difficult?

---

It's difficult because of the sheer quantity and diversity of variables. If you only consider one intersection, you can tune and tweak it, take every variable into account, and get it "perfect" so that average wait times are minimized in all directions at all times.

But then add a second intersection where some of the same traffic interacts with both. Already it is likely impossible to optimize wait times at both intersections simultaneously. The best you can do is balance them to optimize the two-intersection system. And the problem gets harder exponentially as you scale up to hundreds of intersections. Each individual intersection will be pretty far from optimal in order to optimize the system as a whole.

And to truly optimize a whole city, the AI would need navigation data from every vehicle to avoid potential congestion. This might require an entire data center's capacity to keep up with it all if using AI.

quote:

---

Probably not, especially over the long term. EV's weigh and average of 20%-50% more than a similar model/size ICE vehicle. More weight = more wear and tear on roads.

---

If only ICE passenger vehicles were on the roads, the roads would last almost forever. If only EV passenger vehicles were on the roads, the roads would last almost forever. The weather and vegetation would dominate the wear and tear. A single semi causes more wear and tear than thousands of passenger vehicles. I don't think the proportion of wear and tear should be a major factor in how much should be paid in taxes. If it were, the tax burden on trucking would be enormous.

I don't know why anyone would have a problem with road use taxes. Either that or tolls.

quote:

---

I use my Windows devices more because so many legacy desktop apps at my place of work only run on Windows.

---

Funny enough, there are some legacy Windows programs that run better on Linux/WINE than on modern Windows.

I haven't tried to escape proxmox, but it looks to be as easy as exporting the machine then importing to something else. Containers might take a little more but it looks do-able.

If you only have 2 nodes and a workload that fits it well, I guess it would be simpler to eliminate proxmox. Personally though I still like the added layer of separation and flexibility that virtualization gives you. Like in the case of an OS upgrade, even with just 1 machine, it's pretty simple to take a snapshot that I can restore if something breaks. Or spin up another VM on that host with an already-upgraded OS and move the containers over.

quote:

---

Apparently one popular move is to mix the Mac and cheese and the bbq beef brisket.

---

This is my go-to at every potluck. Mac and cheese and meat on a dinner roll. DGAF if I look like an 8 year old.

quote:

---

Technology to send men to the moon on a rocket....but dont have one good camera.

---

Kinda like the first go round

quote:

---

quote:

---

home assistant dashboards and automations

---

Will you elaborate on this?

---

You're gonna have to tell us your starting point because HA is a whole world of stuff and elaboration could fill books. :lol:

Assuming you are starting at ground zero, here is what I will do with AI today:

quote:

---

Home Assistant is an open-source platform that lets you monitor, control, and automate devices in your home from a single system. It runs locally on hardware like a mini PC, Raspberry Pi, or server, and connects to thousands of devices—lights, thermostats, cameras, sensors, smart plugs, and more—across many brands. Because it operates primarily on your local network, it emphasizes privacy, reliability, and independence from cloud services.

What makes Home Assistant especially powerful is its automation engine and extensibility. You can create rules like “turn on lights when motion is detected,” “notify me if a freezer gets too warm,” or “shut down equipment if power quality drops.” Home Assistant often serves as the central integration hub—collecting data from devices, visualizing it on dashboards, and triggering actions based on conditions.

---

I've told my wife if I'm ever on life support, unplug me. Then plug me back in, see if that helps.

I've used codex off and on for a while, but as it's tuned for writing code but not necessarily conversing and "talking out" the direction of development, I've been struggling to produce nice things. Maybe I'm slow/stupid and everyone already figured this out, or maybe there's a better way/tool that I'm not aware of, but I've started doing the following:

I create a project in ChatGPT to collect chats related to a particular app. It guides me on what to build and how to structure it. When we decide we're ready, I have it produce a prompt for codex which is typically a 1k+ word outline of the specific changes to make, changes not to make, constraints, etc. This is *much* better input to codex than the few sentences I was writing before. Included in the prompt is instructions for codex to update a couple of design/state documents which describe the codebase. I feed those docs into ChatGPT so that it stays aware of the state of things to generate better prompts to feed back into codex.