I’m in the very same situation currently. A coworker vibe coded a PR for me to review. I asked: normally I would ask “why did you do xyz”, but what are you going to do now, proxy that question to your LLM? And is the LLM going to construct a “why” based on the nonsensical code it produced? Is this how we want to work?
The review is currently stalled in absence of answers.
You're going to ask your own LLM session why they did it that way. And if there's a problem, you raise it. And if this individual keeps opening buggy PRs, you have a talk with them about how to use the LLM to catch these issues before a PR is opened.
Seriously, it's much easier to review the AI generated plan, instead of reviewing their code. What I found is that, if the change surface is small enough, AI can get it right under the correct assumptions, but the noob mistake is to let AI loose and come up with its own, often not correct assumptions. So you got to step in before they generate the whole slop...
Incredible what mobile phones and browsers can do nowadays. I remember implementing this paper from 1993 (the absolutely OG for this topic and very readable): “Display of The Earth Taking into Account Atmospheric Scattering” by Nishita et al: https://www.researchgate.net/publication/2933032_Display_of_...
In another comment I recalled reading a paper while implementing Rayleigh and Mie scattering a while back...this was definitely it!
Getting it working was a "holy shit, we can actually model this complex real-world phenomenon pretty well with a few relatively simple calculations" moment. I went from a static blue skybox to a full day-night cycle just like that.
Tell me you've never worked with system languages without telling me you've never worked with system languages (telling claude to "write it in Rust" does not count).
The problem is where it's measuring joules of energy. To use cars as an example:
It measures joules of energy as in "how much heat the gasoline we burn produces", some of which we convert to mechanical energy to drive the car, but the majority is just waste heat going out the tailpipe.
By comparison an electric car powered by solar has no tailpipe. There's still a bit of waste heat from electrical resistance, but nowhere near as much.
If we measure like this, by converting a gasoline car to electric (powered by solar for the sake of ignoring some complexity), and driving the same distance, we somehow managed to cut our "energy demand" in half. Despite the fact that we're demanding the exact same thing from the system.
If we measured "joules delivered to the tires of the car" we wouldn't have the same issue. At least until someone starts arguing about how their car is more aerodynamic so joules delivered to the tires should count for more in it.
Edit: We could also go in the other direction. Instead of reporting it as 1kw of solar energy (electricity) it could be 4kw of solar energy (the amount of sunlight shining on the solar panels)... No one does this for obvious reasons, but it's more similar to that primary energy number for fuel in many ways.
The total energy supply figure is a primary energy mix - for the fossil fuels it represents the thermal energy of the fuel. You can look at the final energy consumption section a bit lower to get a different picture taking into account conversion losses.
That is still subject to the primary energy fallacy. Those reports are in terms of primary energy, i.e. how much heat is released by combustion of fossil gas. But in order to replace fossil gas in a chemical plant, you need much less electricity than the primary energy of the fossil gas suggests.
> For all energy sources, the IEA clearly defines energy production at the point where the energy source becomes a “marketable product” (and not before).
Doesn't that mean if you are burning coal to make electricity, you wouldn't count the heat output because the generated heat is not a marketable product.
Looking at the chart for TFC, the wind and solar case looks even worse. Wind and solar supplies 2 million TJ compared to 36 million for coal.
All I was really trying to say from the outset is that I'm surprised at how important coal still is and how little we use renewables. I see articles here all the time about the massive advancements in solar (and wind to a lesser degree) and I had it in my head that renewables were a much larger part of the energy mix than they are.
There is lag created by sunk capital costs. Coal is still producing considerable electric power in the US, but the last time a new coal-fired power plant came online was more than a decade ago, and there are none under construction (although Trump was trying to get one built, to considerable skepticism and inertia). The average age of a coal-fired power plant in the US is 40+ years.
I thought I had read that some cables have systems at each end to detect and correct when there's a splice with swapped cores, however, I can't find a reference; you could imagine such a system would mitigate partial breaks by assigning working paths to priority customers. However, I see a lot more about automatically aligning multicore fibers for splicing, I suspect proper alignment may be more practical than fixing it at the ends.
Even if it's not automated, the number of cores in multicore fibers tends to be pretty low, and there's a standard for marking fibers [1], that's similar-ish to the standard for marking copper pairs [2].
They do other unholy things. I don’t know what, but consistently while playing music on my HomePod opening that site makes it stutter within a few minutes, fully stop working shortly afterwards and it needs a reboot to work again.
The review is currently stalled in absence of answers.
reply