conjecTech

The answer is pretty obvious: those plants aren't new and solar wasn't an option when they were built. There haven't been new coal plants built in the US in some years, and the reasons are purely economic. Wind and solar made up 76% of the generation added last year for that reason[1]. The companies have sunk costs in their existing coal plants, but the economics of solar/wind have become so favorably they've been prematurely retiring them in droves.

[1] https://www.eia.gov/todayinenergy/detail.php?id=42495

Solar's return-on-energy is quite large - on the order of 10x, so you could bootstrap the process of creating all the solar panels you would ever need with a trivially small investment of non solar energy. In practice, it will be higher, but given those economics, every solar panel you manufacture deters 10x the same amount of non-solar electrical consumption.

You need to be precise with how you talk about the shift in energy demand due to EVs. It's not a net energy increase. EVs are far more efficient than gas vehicles. I assume what you mean is increased electricity demand. You can back this out based on typical driving metrics - 250wh/mile is a typical efficiency for electric vehicles. If each driver does 10k miles/year and there are 200M drivers in the US, then this equates to a 13% increase in electrical demand for full adoption[1]. You could probably add on 2% more for efficiency losses in transmission and charging, so call it an additional 15%. That's effectively worst case as it assumes no reduction from the removal of the current infrastructure(ie pumps for gas pipelines, heating for refining, trucking, etc). Those are huge users of energy (though much in forms other than electricity). Oil refinement alone accounts for ~4% of all energy use(not just electricity) in the US.

EVs are so efficient, transitioning to them will probably consume less energy than just the energy we currently spend producing and distributing gasoline. We'll probably need to add to electrical infrastructure, but compared to the savings that is easily justifiable.

[1] https://www.wolframalpha.com/input/?i=250+wh%2Fmile+*+10000+...

It's a good observation, but it only impacts throughput when ship capacity becomes the limiting factor. If you added more ships right now (or shortened wait times), it wouldn't change throughput.

It's probably paid for by the better fee structure on payment processing: credit card processing is generally a base fee + a percent of the purchase. By having people put $20 on a Starbucks card once rather than making 5 different $4 transactions, they reduce the fixed charge they incur by 80%.

I'm sure it varies somewhat from company to company, but the general ethos was to reduce warehousing to an absolute minimum. From a corporate finance perspective, this has a bunch of benefits. Excess inventory is effectively unproductive capital. Additionally, it's costly to store and is subject to theft and damage as well. A lot of the focus on JIT came out of post-war Japan where capital was scarce and this was done out of necessity.

In some cases, an incredibly aggressive supply chain and payment schedule can actually create financing for a company. Aswath Damodaran's corporate finance class mentions one of the major alkaline battery producers as an example. At any point, they only have 2 days worth of production in inventory, and they require payment within ~7 days from clients, while their own suppliers are paid after 30 days. That means at any point, they are holding about 20 days of revenue worth of cash without paying any interest.

This is an issue with latency, not throughput. Our supply chain is actually running at considerably higher throughput than we ever have before. The ports provide numbers for activity on their websites - Long Beach and Oakland are both operating at 30% higher capacity than they were pre-pandemic.

What has changed is the queues and the pricing. The shortages we are feeling are the gaps that have surfaced between when retailers trigger reorders and when they are replenished. They grew accustomed to being able to do that just-in-time as forecasted inventory went to 0, but the added delays have made this much harder.

Additionally, the shipping prices have made many things simply uneconomic to ship. I saw an analysis on the difference in cost recently for a grill: prior to the pandemic, shipping from China to LA contributed $5 to its price, with shipping costs up 8x, that's now $40. Rather than fulfill orders at negative margins, many people are simply waiting for lower prices after the holidays. It's not that they're being held up in transit, these things aren't being shipped to begin with.

I'd be curious to hear where you think all of that complexity has moved. While ICE are definitely more reliable these days than they were 50 years ago, I don't think the major issue is with them directly, but with the sheer amount of ancillary components they imply - ie belts, alternator, fuel injectors, spark plugs, transmission, radiator, starter motor, starter battery, etc. Those are the source of the vast majority of problems and maintenance for ICE cars. EVs need none of them. For electric vehicles, I'd view the cooling system as the largest source of complexity. And I think those will simplify as batteries become better designed (lower internal resistance from tabless for instance), more dense and less expensive.

I think it's important not to conflate Tesla and electric vehicles. They are building luxury vehicles that have the added complexity and added unreliability that implies.

There may be other factors besides what they nominally "want". Depending on their source of financing, landlords changing the rent on these spaces can be both logistically difficult and may cause a huge negative cashflow shock because of the way it effects capital requirements on commercial loans. Louis Rossmann covers it in a video (https://www.youtube.com/watch?v=NdfmMB1E_qk) based on this reddit comment (https://www.reddit.com/r/nyc/comments/innhah/nearly_twothird...).

I do agree that logic puzzles which require a single observation and/or have a binary outcome tend to be very bad interview questions. That being said, I think the original post demonstrates that the issue with the manhole cover problem isn't necessarily the question itself, but how it's evaluated. I'd be quite pleased if I were asking the question and the interviewee gave that answer (though I might raise the point that the configuration necessary to drop a rectangular cover is also one of the most ergonomic for holding it). The general notion that there is a single right and wrong answer is incongruous with a class of question used to evaluate the candidate's reasoning ability. Though I wouldn't be surprised if these were historically evaluated appropriately and candidates simply misinferred why they were being rejected.

I don't get the derision for these kinds of problems. And I think it's unlikely Feynman would have found this unhelpful. These are commonly referred to as Fermi problems. Fermi was a physicist of Feynman's generation - they were probably standing together at Trinity when Fermi did the famous yield approximation. The other physicists held this ability in such esteem they gave it a name and remembered him for it.

The question is meant to gauge whether you can create a simple model of a system that is completely foreign to you, but for which you are familiar with the fundamental rules governing it. The absurdity of it serves to remove bias. You don't have to worry about whether or not the candidate happens to have experience filling airplanes with ping pong balls. So much of effective programming is creating a model for the processes we are trying to automate. Coming up with effective abstractions uses many of the same skills.

A lot of the construction boom is probably attributable to the new found geographic freedom COVID gave, coupled with the low interest rates from the Fed's monetary policy. There was a sudden mismatch of locale supply and demand. Housing starts were also quite low over much of the last decade, definitely below the equilibration point for 2019-2013.

It'll be interesting to see how things play out. I think the shift of older millennials out of cities combined with the relatively smaller generation replacing them will likely cause a reversal of a lot of the urban price growth in the past decade in places like New York. Rents are already reflecting that, but sales prices have more hysteresis.

Housing starts are at ~1.7MM units/year right now, that's enough for about 4.5MM people at current household sizes. US population growth was 1.6M in 2019 and <1MM in 2020. Even with generous estimates on losses of existing housing, I don't think 3x is appreciably off.

I saw a lot of people in the trades barely scraping by 5-7 years ago when housing was bottoming out. Housing stock is growing at something like 3x population growth in the US right now. A few more years of that and they may be back in the same situation. Being employed in a heavily cyclical industry definitely has rich years, but it's also got lean ones.

It is, and they literally address this in the first paragraph.

This is horrific for employees. Ben Kuhn already nailed the math here[1]. The optionality embedded in long-dated grants is a huge fraction of total comp at high-growth companies and represent almost the entire right tail of outcomes. It also requires assumptions about the future, which is why companies generally abstain from quantifying what it's worth. It seems like the companies doing this are trying to arb that uncertainty by retaining all of the potential gain for themselves in exchange for a slightly larger nominal income.

Don't do it.

[1] https://www.benkuhn.net/optopt/

Woo boy, I see a lot of people throwing shade at Atlanta here. That's not exactly a new phenomenon. As a native ATLien, let me quote one of our prophets:

The south got something to say.

I think all of my units are correct. The primary point of the second paragraph was that the kind of analysis you are pointing to is inappropriate here. Thinking of this from a server heat-sinking perspective isn't helpful because the car side of the system doesn't have to operate indefinitely in steady state and is a large thermal mass.

I've worked on very high current automotive applications in the past. While you raise some valid points, there are generally solutions to them. I've found its best to think of these things on a unit level - which is why so many battery characteristics are quoted in that way. If you can think of solutions for how to get around these issues for a single battery on a bench, there tend to be tractable ways to scale them.

Heat dissipation is one of the only things that doesn't scale linearly. 6kw sounds like a lot of power, but over 10 minutes, you are only talking about 1kwh of energy - about enough to boil 10 liters of water. Given the mass of the 100kwh battery, a large fraction of that could just be absorbed as temperature gain. If you assume the specific heat of iron (which is an underestimate) and the weight of the current 100kwh batteries, its about a 13 degree C change.

Looks like the market share for electric hit 50% in December for Sweden if you count Plug-in Hybrids[1]. So a bit more optimistic! Clean Technica does great reporting for these numbers.

[1] https://cleantechnica.com/2021/01/05/sweden-hits-record-50-e...

Interesting...as someone else pointed out, this is just for feeds. If I recall correctly, Twitter uses Redis as the store for user feeds. I wonder if this is a result of the recent deployment of Graviton2 instances for ElastiCache[1]. I'd think Twitter would be at a scale where such a managed solution wouldn't be hugely valuable, but I thought this use of Graviton chips was really smart. I wouldn't be surprised if Amazon gave them a very good deal to lock in a big customer for that offering.

[1] https://aws.amazon.com/about-aws/whats-new/2020/10/amazon-el...

I think you just made the Justice Department's case for them. An H-1B is intended for individuals with specialized skillsets. The fact that Facebook engages in this form of mass, generic recruiting underscores the fact that they are not seeking out rare skillsets and don't really need to utilize this program to find people matching the required criteria.

Visa and Mastercard are both publicly traded. You can go look at their financials. I believe Visa still holds the crown for the highest gross margin and price-to-sales ratio of any stock in the S&P 500. Mastercard isn't far behind. These are not marginally profitable businesses. They are money printers that rely on tightfisted control over merchants and the payment clearing infrastructure.

It's nowhere near 3.5%. It's probably an order of magnitude less. My mother spent a decade as the controller of a luggage store chain. She told me the amount they spent on credit card fees was at times in excess of their stores' payroll. It is rent seeking at nearly unfathomable levels. Even the shady check cashing operations for the unbanked are able to operate profitably on a 1% cut.

It's not as simple as thinking in terms of those first order effects. Mailing subsidies reduce the cost of non-local alternatives for goods. It can often make it impossible for small businesses to exist in less urban environments and can lead to downward spirals where rural communities are self-exterminating because processes like this consistently siphon money out of the local economy.

Imagine if shipping of all consumer goods were made free. Amazon would be INCREDIBLY more dominant. And in fact, this is a large reason they started with books. The USPS has a special program built for a world 100 years ago called Media Mail which allows you to ship books for next to nothing. It made it impossible to run a bookstore profitably. Removing shipping subsidies to rural communities may very well create more of them and will definitely make them more resilient.

My apologies. I didn't realize there was an explicit policy.

Daniel, calling this a duplicate when the original post is 9 months old seems really questionable.

It's likely the Technical University of Munich based on the hyperloop piece. Though I had to do a double take because it sounded incredibly similar to some of my experiences at Georgia Tech. If you want to know the kinds of universities that do these kinds of things and do them well, its useful to look up some of the student design competitions and see who regularly attends. From recollection, MIT, CMU, Olin, Dolhousie University, Vanderbilt, IIT, RPI, Embry-Riddle, Harvey Mudd are a subset of the places where you can have this kind of experience.

I've had the good fortune of knowing the TripleByte team personally. I'm not at all surprised to see this being handled in such a sincere and agreeable way.

Ammon is a sincere and truth-seeking individual. He's willing to be convinced that his opinion is wrong, a character trait we don't do enough to praise and which I've found to be exceedingly rare these days. Situations like this highlight exactly why I've trusted them with my data in the past and will continue to recommend TripleByte to friends in the future.

Fellow techie. Did you forget about the mandatory athletic fee of about $250/year[1]? It covers about 7% of the athletic departments total expenses[2]. I believe there were other backdoor mechanisms that allowed for transfers of school funds to the athletic department, though I don't recall the details well enough to contest it.

[1]http://www.bursar.gatech.edu/student/tuition/Spring2020/sp20... [2]http://www.budgets.gatech.edu/Fees?id=6

Very little of that is true demand. That corresponds to 10x American GDP annualized.