It's an accelerator. A great tool if used well. But just like all the innovations before it that were going to replace programmers it simply won't.
I used Claude just the other day to write unit test coverage for a tricky system that handles resolving updates into a consistent view of the world and handles record resurrection/deletion. It wrote great test coverage because it parsed my headerdoc and code comments that went into great detail about the expected behavior. The hard part of that implementation was the prose I wrote and the thinking required to come up with it. The actual lines of code were already a small part of the problem space. So yeah Claude saved me a day or two of monotonously writing up test cases. That's great.
Of course Claude also spat out some absolute garbage code using reflection to poke at internal properties because the access level didn't allow the test to poke at the things it wanted to poke at, along with some methods that were calling themselves in infinite recursion. Oh and a bunch of lines that didn't even compile.
The thing is about those errors: most of them were a fundamental inability to reason. They were technically correct in a sense. I can see how a model that learned from other code written by humans would learn those patterns and apply them. In some contexts they would be best-practice or even required. But the model can't reason. It has no executive function.
I think that is part of what makes these models both amazingly capable and incredibly stupid at the same time.
Funny enough Chinese State owned banks have been doing much the same for quite some time. No one ever defaults, loans are extended as long as it takes. Presumably the threat of being called into the next party meeting to explain yourself is sufficient motivation for the people running the business to pivot as many times as it takes until they find a way to make money. Worst case the state swaps someone else into leadership.
I say this to say... who knows? I guess if you shuffle deck chairs fast enough everything works out fine (?)
The larger you are, the larger the rounding errors are, the more money that can disappear due to a failure and explained away or extended or written off or whatever euphemism you want to pick. But the sum of rounding errors is less likely to itself be a rounding error. It works until it doesn't, and Evergrande collapsing with $300 billion in Chinese real estate debt will be a case study for years to come.
Isn't the real underlying risk here concentration, as opposed to diversification?
If you have unlimited capital and time horizon, because you're a nation with the power to tax and print money, then you can keep this game going for a long time.
The only thing that mandates it stops are if (a) too many of your loans are correlated with the same thing that crashes (e.g. energy, tulips, AI, etc) or (b) too many of your loans are tied together in a single entity (either because it combined multiple smaller entities or because it tied itself into all their financial arrangements).
Ohhhh a live one! Sir do I have a wonderful bridge in Brooklyn to sell you! :)
Fun fact: banks fund this sort of nonsense constantly. I've asked about this before: why they do it. They must be making money I just don't know how. The LBO guys pay themselves massive management fees and dump the debt on the company so they walk away scott free.
My wild guess was the banks offload the eventual IPO onto investors and so make their money on the IPO fees and funneling their own clients the dead-man-walking shares. But I honestly don't know.
> wild guess was the banks offload the eventual IPO onto investors and so make their money on the IPO fees and funneling their own clients the dead-man-walking shares
The banks get paid back their debt when the next PE fund buys the company or the company pays it off. Unless an IPO is being done to pay off debt, which it never is, the mechanism you describe doesn’t occur.
The list of companies imploded by LBO/PE is quite high though. Why do banks keep lining up to fund such deals? They must be making money somehow. These companies aren't worth much in liquidation. Are they able to extract enough value during the dead-man-walking period to make it worthwhile? Especially for retail or similar deals where the bank isn't going to foreclose on a bunch of real estate or assets worth selling.
I was not saying this is how they make money - I was saying I honestly don't know. If you do know please share. I would love to understand why the banks are so keen to fund what looks to my eyes like super shady vulture capitalism. We start with a profitable company and end with a smoking husk. The Wall Street guys are doing it to steal as much value as they can before it all blows up. Someone is eating the eventual loss. Who? Or are you saying the majority of these deals don't end up with the company being eviscerated?
"usually" and "typically" are doing a lot of heavy lifting here :)
Access to knowledge, equipment, and budget varied dramatically prior to widespread internet access. Someone setting up a BBS might not even know about multi-line modem cards or serial port expansions. Even if they knew about them they may not have been able to reasonably obtain them. Or they may have been operating on donations, surplus, or discount equipment. Or they simply may not have had the luxury of time to research all of that as user demand meant they were too busy laying tracks in front of the train.
Many BBSes ran on 1-2 lines per PC because that's what they understood how to build or the hardware they had access to. You might be surprised at just how many lines some BBSes setup this way had!
People forget there was a time that anything outside the standard PC was extremely expensive, often had flaky or nonexistent software support, locked you into a fly-by-night vendor that might go out of business tomorrow, was only available via a distributor who wanted to have you talk to a "sales consultant" before they'd sell you something, etc. Many many people chose sub-optimal implementations because it was an off-the-shelf PC they could replace at any time with trivially simple software requiring no special CONFIG.SYS drivers or TSRs to fiddle with. Especially if you'd ever been burned previously.
> It's a wild misunderstanding of how BBSes worked.
That's quite the assumption.
There were a lot of different BBS hosting programs. They wildly varied in what they supported and how they were implemented. Further even within a given piece of software the ways you could configure them and the consequences also varied. Even if a given software supported concurrent users on a single PC for various reasons a BBS might choose not to host that way.
Making the default MAC address and machine serial number depend entirely on the NVRAM/RTC battery is... a choice. You'd think they would have used some fusible links to burn the original values into the hardware to fall back on if the battery died.
Nobody burns in MAC addresses. MAC is usually stored on eeprom. Either in BIOS or small i2c one on the NIC. Storing it in the nvram means you save $0.1 BOM, dont have to partition your bios chip and serialization becomes cheaper and faster. Its not like losing one is a problem, just bang a random number with Sun prefix and you will be statistically fine.
> The IDC article says that DRAM prices are not expected to come down again
Sure thing. I'd take a look at IDC & similar firms' forecasting history before worrying too much about what they say.
There is an AI boom right now. There will be a consolidation cycle at some point. When that happens half the players, if not more, will disappear. The huge hardware budgets will go with them.
We also can't be certain that the DRAM makers aren't capitalizing on this opportunity because they can. Remember: all of them are convicted monopolists. As in actual prison time convicted. And fined. And lost civil lawsuits. Multiple times.
I just can't see AI paying enough of a premium on HBM to justify the DRAM spikes. Frankly I can't see the volume either. Wafer starts on DRAM are dramatically bigger than you are probably imagining. DRAM is in practically everything these days. AI servers is but a drop in the bucket. 10% of the market? Yeah right, if its 4% I'd be shocked. And you are telling me a shift of 4% of wafers to HBM is driving these prices and shortages?
I humbly suggest if you look at the numbers something smells funny.
Disclaimer: none of us has access to the actual data, a lot of it is inferred by industry players. Some are well connected and usually accurate but that is not evidence. Therefore it is possible this is a genuine market action and nothing nefarious is going on.
HBM is not normal memory. It uses a lot more area per bit and has lower yield too. So a Gb of baseline DRAM and a Gb of HBM are very different measurements, the latter equates to so much more in terms of volume.
True which is why I posted the caveat. It is possible that the prices we are seeing are legitimate market dynamics. I just think given past history we should be skeptical of that. It smells like market manipulation to me.
Of course given the dynamics of business these days and how financialized these companies are it wouldn't surprise me if this is also just a result of refusal to invest capex in expansion. So long as the other 2-5 players in the market do the same there's no benefit to expanding or lowering prices. We all just sit on our hands and let market forces raise prices (and profit) for us.
Whatever the case it seems capitalism has gone off the rails in multiple ways.
> At least allow us to use names instead of numbers.
You can for the destination. That's the whole reason you need the "&": to tell the shell the destination is not a named file (which itself could be a pipe or socket). And by default you don't need to specify the source fd at all. The intent is that stdout is piped along but stderr goes directly to your tty. That's one reason they are separate.
And for those saying "<" would have been better: that is used to read from the RHS and feed it as input to the LHS so it was taken.
My question is: who is lending the money for these leveraged buyout deals? They seem to leave the lenders holding the bag at some point when it all implodes. Do these deals really pay off often enough to be worth financing them?
> who is lending the money for these leveraged buyout deals? They seem to leave the lenders holding the bag at some point when it all implodes. Do these deals really pay off often enough to be worth financing them?
For this deal, it is:
- ~$57b Debt financing: Bank of America, Citigroup, Apollo
- ~$46b Equity backing: Ellison's dad, RedBird Capital Partners
- Sovereign wealth funds on the equity side: Saudi Arabia, Qatar, and Abu Dhabi
They keep financing these deals so they must be making money on them but there is so much value destruction in so many of these cases. Someone has to be left holding the bag of hot garbage in the end so who is taking the loss?
We know private equity offloads the debt onto the walking corpse and usually pays themselves huge fees and bonuses for their "assistance". So it is clear how they make money. I just don't get how the banks make out.
US style plugs and derivatives (and Australian, Japanese, Brazilian, etc) - all invented by Hubbell - are "good enough".
Are they objectively good? No. Do they regularly fail, cause fires, or shock people? No.
Even my kids when young understood how to grip the plug without touching the metal contacts and to this day still have not been shocked. In theory can something fall and hit the pins just right to cause a short? Sure. You could also get struck by lightning. In practice it just doesn't happen very often.
For the US/North American NEMA style there are some improvements and some clever things about them. Modern receptacles have shutter doors that stop you from putting anything into the holes unless the ground pin or neutral pin unlocks it first. Many plugs also cover the rear part of the hot/neutral with plastic so if the plug is not fully inserted there is no exposed metal.
The plugs also prevent mixing voltage and amperage. The typical two vertical blades (5-15) are for 15 amp circuits. 20 amp circuits (5-20) have one horizontal + one vertical blade. The receptacle has a T shaped slot to match - that way you can plug a low-amp device into a high-amp circuit but not the reverse.
Similarly the 240v version of this plug (6-15/6-20) has the same property: 15amp and 20amp versions. The 15 amp is two horizontal blades. The 20 amp is one horizontal + 1 vertical but swapped places compared to the 120v version. I do wish more builders installed the 240v receptacles in kitchens in the US. There is no technical reason we can't have higher power kettles and whatnot. If code required these in garages and kitchens more appliances would be available for them.
(I find it insane that Brazil continues to be dual exclusive voltage; all of North America is dual concurrent voltage. Every home/office has 120v and 240v available. In Brazil it depends on what state/city you live in - some get 120v, some get 240v. Even worse they use the same standard plug design for both so you'd better hope the plug is the right color or has the right sticker. And you can't be sure you can take electrical appliances from one city to the next! At least they should have adopted different plugs for different voltages.)
The huge advantage of these plugs is compatibility. We already have them. The cost to change designs is massive. The benefit extremely small. It just isn't worth doing.
Note: The 240v NEMA plugs I am referencing are not "dryer plugs" which are physically much* larger and designed for much higher amp loads in the 30-60 range. The 6-15/6-20 are literally identical to the standard 120v plugs but with different blade orientations. They were designed to support 240v appliances in everyday use since all of North America is dual voltage. In practice 240v is only ever used for large appliances like ovens so the 6 series doesn't get much use which is a bit of a shame.
> The plugs also prevent mixing voltage and amperage. The typical two vertical blades (5-15) are for 15 amp circuits. 20 amp circuits (5-20) have one horizontal + one vertical blade. The receptacle has a T shaped slot to match - that way you can plug a low-amp device into a high-amp circuit but not the reverse.
Alternatively you can just run everything at 230V and then you don't need a million different plugs as any wall socket can provide up to 3.5 kW, enough for any home appliance except for the most power hungry ovens and IH stoves.
Rewiring an entire country is a wee bit expensive. Even if the wires are rated to 300V (they usually are), transformers would have to be replaced, and they cost quite a bit. Also different sockets and circuit breakers, and a whole lot of billable hours by licensed electricians.
(But changing the voltage is easy compared to changing the mains frequency. Japan still has to live with 50Hz in one half of the country, and 60Hz in the other.)
The transformers are fine as they are. We already have 240VAC in the typical US home: Two legs of 120v, with one being 180° out of phase. That part exists and it works fine -- the big industrious parts of the infrastructure are already supplying 240v.
US 240v is a bit different than the way the rest of the world usually does it, where they have 1 leg of 240v and 1 neutral, but AFAICT that detail is not a big deal for the stuff that actually uses electricity.
The wire itself, broadly-speaking, is fine.
Suppose we decided that tomorrow at noon to begin the move to 240v.
We just refactor our breaker panels and update to some new objectively-good whiz-bang outlet format (because we would certainly never borrow existing designs from anywhere else on the planet; we in the States have a big problem with Not Invented Here when it comes to policy), and finally get rid of twist-in Edison light bulb sockets, and that part is done.
But then all kinds of stuff doesn't work anymore.
Fridges, garage door openers, microwaves, light bulbs, clock radios, natural gas furnaces, and anything else that doesn't work with 240v: That stuff is dead in the water without converting back down to 120v using an autoformer or something.
Sure, we'll eventually get things updated; when we don't count survivorship bias examples, it's plain to see that stuff just doesn't last all that long anyway (and never actually did).
But for a time: There literal mountains of stuff that just won't work without help. And that's a tough pill to swallow.
---
What we could start doing is embrace our existing dual-voltage home wiring methods, and putting 240v sockets in some places where it's actually beneficial. Places like kitchens (for heating water and food), say. But broadly speaking: Nobody does this because nobody sells safety-approved residential appliances for the US domestic market, so it's a lot of money to spend to get it done for no benefit. It's a catch-22.
> What we could start doing is embrace our existing dual-voltage home wiring methods, and putting 240v sockets in some places where it's actually beneficial. Places like kitchens (for heating water and food), say. But broadly speaking: Nobody does this because nobody sells safety-approved residential appliances for the US domestic market, so it's a lot of money to spend to get it done for no benefit. It's a catch-22.
Yes, this is the problem I've been complaining about.
If code required at least two NEMA 6-15 outlets in kitchens and garages builders and remodelers would install them.
Once they started rolling out there are a number of existing appliance makers who sell 240v appliances in the rest of the world who could begin offering that variant for sale.
It would be a good way to kick-start the market without any major changes! It is literally just running one extra romex and installing one double-pole breaker yet it makes double the power available. The builder that remodeled my house ran multi-branch style (3-conductor + ground) to save on copper so if I can find a junction box where that wire splits off I can install one right there without even running a new wire as the breaker is already a double-pole breaker.
> Alternatively you can just run everything at 230V
Sure and everyone can just stop using Python 2.x tomorrow right?
Backwards compatibility is a big deal. Even moreso when it involves physical infrastructure. who wants to pay billions upon billions of dollars to make the change? How long will it take to roll trucks on all those linemen and electricians to convert/retrofit everything? Does the customer pay? The government?
And at the end of it everything is just the same as it was before. There's no huge benefit to be had for doing it.
That's why I said I wished we had more use of the 240v NEMA plugs. So we could begin supporting higher power appliances over time without some huge switchover expense.
For that matter I wish 3-phase was more available. I have a small machine shop in my garage that would greatly benefit from it.
> US style plugs and derivatives (and Australian, Japanese, Brazilian, etc)
Brazil no longer uses US style plugs (though you'll still find them in older installations), it nowadays uses a much safer EU-derived style.
> I find it insane that Brazil continues to be dual exclusive voltage; all of North America is dual concurrent voltage. Every home/office has 120v and 240v available. In Brazil it depends on what state/city you live in - some get 120v, some get 240v.
This is wrong; it's very common to have for instance both 127V and 220V in the same building, sometimes even side by side in the same wall faceplate; 127V is phase to neutral, 220V is phase to phase (on the common 3-phase system). Yes, it does depend on the city, some cities use 220V exclusively, and there are a few other variations, but AFAIK the 127V/220V 3-phase combo is the most common.
> Even worse they use the same standard plug design for both so you'd better hope the plug is the right color or has the right sticker. And you can't be sure you can take electrical appliances from one city to the next! At least they should have adopted different plugs for different voltages.
Yeah, at least it's better than the confusing mix of legacy sockets we had before (which already were mixed voltage - and yeah, we already used the "120V 5-15 NEMA plug" aka "computer plug" even for 220V).
> This is wrong; it's very common to have for instance both 127V and 220V in the same building, sometimes even side by side in the same wall faceplate; 127V is phase to neutral, 220V is phase to phase (on the common 3-phase system). Yes, it does depend on the city, some cities use 220V exclusively, and there are a few other variations, but AFAIK the 127V/220V 3-phase combo is the most common.
That's good news. I'm glad my info is out of date!
> Similarly the 240v version of this plug (6-15/6-20) has the same property: 15amp and 20amp versions. The 15 amp is two horizontal blades. The 20 amp is one horizontal + 1 vertical but swapped places compared to the 120v version. I do wish more builders installed the 240v receptacles in kitchens in the US. There is no technical reason we can't have higher power kettles and whatnot. If code required these in garages and kitchens more appliances would be available for them.
If we had more 240v circuits in garages and on the outside of the house you could use electric motors for more yard tasks. Batteries and gasoline is used often in the US because our branch circuits provide about half the power of a 240v branch circuit. You can buy electric mulchers that are powerful enough to grind tree limbs but they can’t run on a 120v circuits.
We get 3 phases to each home, phase to neutral is 127v, and that's the standard voltage, so loads are divided over the 3 phases.
230v we get through phase to phase connections. We also balance those for the 220v loads, but it's kinda risky due to the nature of our grid, being an island.
Whenever there's a fault they disconnect the zone affected but sometimes in the process we get VERY short but massive overvoltage events.
Since everybody generally uses 127v, as the system trips the 127v line voltage increases for a bit, often within spec but because we take 230v from between the phases it spikes to heights beyond spec and burns the devices.
We use a center-tap neutral except for commercial/industrial that receives three phase.
Most small-to-medium homes/businesses have two hot legs coming off each side of the transformer coil. The neutral is connected to the center of the coil and bonded to earth/ground so it becomes a 0v reference. Each hot leg to neutral is 120v. Between hot legs gives 240v. That neatly supports both voltages in a backwards-compatible way. Typically clothes dryers, hot water heaters, ovens/stoves, etc are 240v appliances. Lamps, USB chargers, and other small day-to-day stuff is 120v.
There are two failure modes that can happen but they are rare and usually only affect the customers attached to the affected transformer or a single customer.
1. Floating neutral. If the neutral becomes disconnected that causes floating voltages as the electricity backs up across the neutral and returns via the opposing hot leg. This presents as randomly fluctuating high/low voltages to 120v appliances but most 240v appliances don't use then neutral and don't care.
2. Damaged hot leg. One hot leg partially arcs to ground or is otherwise damaged. This causes half the 120v appliances to flicker/brown out. 240v appliances will see random low voltages.
Three phase is often delivered as wild leg/high leg delta so a neutral can be derived. It is usually setup so one phase (eg A/C) is center-tapped to make the neutral and two hot legs. This gives three phase power per normal and the same setup as a normal home would have: A/C forms two 120v legs wrt the center tap neutral. However you get 208v between the other phases and neutral so for high density housing you also need to balance the phases resulting in some apartments having 208v power rather than 240v. Thus most 240v appliances also support 208v here but unless you've lived in an apartment or worked on commercial/restaurant systems you'd never see that voltage.
Our breaker panels have 3-phase variants. You'd usually install both: a 240/120 panel for "normal" loads and a 3-phase panel for 3-phase and 240v split phase loads. Breaker design is the same: 3-pole takes up three slots and the bus bars alternate by 3 so every third point is on a different phase.
I used Claude just the other day to write unit test coverage for a tricky system that handles resolving updates into a consistent view of the world and handles record resurrection/deletion. It wrote great test coverage because it parsed my headerdoc and code comments that went into great detail about the expected behavior. The hard part of that implementation was the prose I wrote and the thinking required to come up with it. The actual lines of code were already a small part of the problem space. So yeah Claude saved me a day or two of monotonously writing up test cases. That's great.
Of course Claude also spat out some absolute garbage code using reflection to poke at internal properties because the access level didn't allow the test to poke at the things it wanted to poke at, along with some methods that were calling themselves in infinite recursion. Oh and a bunch of lines that didn't even compile.
The thing is about those errors: most of them were a fundamental inability to reason. They were technically correct in a sense. I can see how a model that learned from other code written by humans would learn those patterns and apply them. In some contexts they would be best-practice or even required. But the model can't reason. It has no executive function.
I think that is part of what makes these models both amazingly capable and incredibly stupid at the same time.
reply