HNHacker News
TopNewBestAskShowJobs

thedufer

3,203 karma · joined April 4, 2011

I'm a software developer in NYC. I previously worked at Trello, Inc (formerly part of Fog Creek Software). I have interned (Software Engineering) for a total of about 13 months spread among Gridpoint, Skyline Innovations, and Fog Creek Software. I graduated from the College of William and Mary in 2012 with a double major in Computer Science and Physics and a minor in Math.
submissionscomments
thedufer··on The worst possible way to be rejected for a job
I didn't realize this was becoming more common. It seems far more uncomfortable for everyone involved, although based on the comments it does seem like some people appreciate a phone/Zoom call. I don't really understand why, though (please try to explain!). An email gives me more time to process difficult news, and prevents me from needing to act positively in an unambiguously negative situation.

I've been called for a rejection exactly once, less than an hour after I'd left the office where I was being interviewed. I was kind of shocked (the interviews went really well) and I didn't realize until later that the recruiter never said my name, and had previously claimed I would absolutely not get the result for at least 3 days. A week later, I got an offer, and I always felt a little bad for the candidate I unintentionally caused to get ghosted because the recruiter dialed the wrong number and I didn't figure it out.

thedufer··on Car insurance in America is too cheap
Gap insurance covers a tightly-bounded amount of loss (it will never pay out more than the remainder of your loan), which is why it's so cheap. That's also why it's a great target for self-insurance - you know exactly how much you need to have to cover the worst-case loss. If you have the cash, you are almost certainly better off skipping this one.
thedufer··on New York will plant trees using new tech to maximize foliage impact
NYC uses ginkgos for this purpose. They still make up a fairly small portion of the trees in the city, though. They're also particularly hardy when it comes to handling pollution, which is obviously valuable in cities.

I believe this style of root system is called a taproot.

thedufer··on New York may ban noncompete employment agreements and Wall Street is not happy
The argument against this is that a company spends millions of dollars in research to learn something valuable, and anyone who didn't spend that money can trivially outbid for the employee that knows the results, since they can pay the employee some significant portion of the cost of the research that they didn't have to do and still come out ahead. I'm not sure I entirely buy this, but it's a lot more nuanced than "just pay your employees well".
thedufer··on New York may ban noncompete employment agreements and Wall Street is not happy
They're not applied as widely as it may seem. The terms are typically "up to" the length of time, and in practice firms waive 50-100% of the non-compete length pretty frequently, which is a decent sign that the cost is non-negligible. It is a bit tricky not knowing until you quit how long you'll be held to it, though.
thedufer··on FTC judge rules Intuit broke law, must stop advertising TurboTax as “free”
The original comment was confusing - they're getting $60B, and also have a mandate to modernize. Only a minority of that money is actually being spent on the modernization. Most of it is being spent on additional auditing of tax returns - it's well known that every $1 spent on audits brings in more than $1 of additional revenue, and we're nowhere near the breakeven point. It's expected that the additional auditing will pay for itself several times over - even if you count the modernization spend against it as well, the government will see a large ROI on this $60B (they're expecting additional revenue of over $200B from the extra auditing).
thedufer··on What is gained and lost with 63-bit integers? (2014)
It's a bit more complex than this, but roughly each block must contain either all pointers/ints or no pointers. There's a tag at the beginning of each block indicating what type of block it is (and some other stuff - the GC stores some state in the tag as well). If you have, for example, a record consisting of a pointer, an int, and a float, it'll be tagged as having all pointers/ints, and the float will actually be a pointer to a block tagged as data, that contains the float.

This is the "float boxing" problem - floats cost an extra pointer chase in most cases. The 63-bit int is an optimization that avoids needing the extra indirection for ints. You can still have 64-bit ints, but they aren't the default, and they cost this extra pointer chase.

thedufer··on Driving is more expensive than you think (2020)
In the context of NYC, it's the Metropolitan Transportation Authority, the entity in charge of most of NYC's public transit (subway, buses, etc).
thedufer··on Reduce vs. Fold in Common Lisp
The initial value is optional for `reduce`, but required for `fold`. If you don't pass the initial value to `reduce`, and the sequence argument is empty, then `reduce` calls the function with no arguments, which is unique - in all other cases, it is called with two arguments. `fold` always calls its function argument with two arguments.

This shows up more clearly in statically-typed functional languages, where variadic functions like this are far less common. In that case, you typically see that `reduce` returns an option type, whereas `fold` does not. The types would look something like `fold :: (a -> b -> a) -> a -> List b -> a` vs `reduce :: (a -> a -> a) -> List a -> Option a`.

thedufer··on Bank Failures Visualized
I strongly suspect that any sources that claimed that simply made the same error that you did.

Looking at the source you provided from the Chairman of the FDIC, it actually says that the top ten accounts held $13.3B - not _more than_ $13.3B, as you claimed. The next ten accounts necessarily held less than that, so say a generous upper bound at $26.6B for the top twenty accounts. In order for them to get at least half of the bailout, they would have needed to hold north of $55B in aggregate. The FDIC itself would have to have gotten the size of SVB's largest depositors wrong by over double in order for your original claim to be true, in the absolute most generous case.

thedufer··on Bank Failures Visualized
I think that's a misinterpretation of the data. Based on reports, SVB had around 130B of deposits when it was taken into receivership. Something like 90% of that was uninsured, so around 115B of uninsured deposits. The 20B hole means that without FDIC backing the uninsured deposits, they still could have paid out over 80 cents on the dollar. So if the top ten accounts had about 13.3B of deposits, they only received less than 2.7B of the bailout - less than 15%, not "well over half".
thedufer··on JPMorgan Chase Bank Assumes All the Deposits of First Republic Bank
Yeah, the amounts they're worried about are a few orders of magnitude larger than that. I don't think that's a compelling counter-example.

> If the Fed keeps trying to fix failed banks through indirect money printing (covering uninsured deposits by overpaying for treasuries/MBS)

Is this happening? The uninsured deposits are being covered by the FDIC insurance fund, which is funded by the banks. None of it is coming from the fed.

thedufer··on JPMorgan Chase Bank Assumes All the Deposits of First Republic Bank
I think you're describing "narrow banking". It doesn't exist in the US because the government doesn't allow it. As I understand it, they don't allow it because it would be obviously better for deposits individually, but would slow down the economy as a whole by driving up the price of loans, since all of the deposit money would suddenly be unavailable for lending.
thedufer··on A 1.5GB string
This service keeps around a lot of long-lived per-session state, which leads to a very different memory profile from a typical web service, like what you're describing. Your point is valid in general, but doesn't seem to be relevant here.
thedufer··on First-Citizens Bank to assume deposits and loans of Silicon Valley Bridge Bank
It sounds like you might be inferring that $100 in deposits is both a) $100 owed to their banking customers and b) $100 in cash. It is only the former, though - the cash was lent out to make the loans or used to purchase income-producing assets. A bank with $100 in deposits and $50 in loans is insolvent - it owes $100 to its customers whenever they decide to withdraw it, but only has $50 of assets (loans) to back that. In the reverse, the bank only owes $50 to its customers, and meanwhile has $100 of assets - if the customers all wanted their money back, it could sell the loans, and as long as it got more than half of its nominal value they'd be able to repay everyone.
thedufer··on IRS builds task force to explore running its own free e-file system
It does not, unless you define "equally" in a very particular way. Under consumption-based tax schemes, poor people spend a larger portion of their income on taxes than rich people (because they spend more of their income, and consumption taxes are a tax on spending).

I'm not sure what you mean when you say that progressive taxes are a sham. The goal of taxing rich people at a higher rate than poor people seems pretty obviously good. Maybe the implementation issues wherein unrealized capital gains aren't captured? But that's particular to our implementation of taxes, not to progressive taxes generally.

thedufer··on IRS builds task force to explore running its own free e-file system
Consumption taxes are regressive, so this would shift a lot of tax burden from the rich to the poor.
thedufer··on “Expect tests” make test-writing feel like a REPL session
The open source use pattern for expect tests in OCaml (via dune) is exactly as you describe (see https://dune.readthedocs.io/en/stable/tests.html) - you run the tests with `--auto-promote` to tell it to update. The editor integration is a very simple keybinding on top of more generic tooling.
thedufer··on 10 years of FTL: The making of an enduring spaceship simulator
FTL is in the app store on iOS. I remember playing it on an iPad years ago - looks like it's been available there since 2014.
thedufer··on A layoff guide for tweeps
My reading was that they're suggesting that in so-called two-party consent states, getting a third person in the room (whether or not it's done surreptitiously) and then having consent from 2/3 parties makes the recording legal. That's not true, of course - "two-party" is a bit of a misnomer. This feels like dangerous advice.
thedufer··on Subprime loans for college hiding in plain sight
Not explicitly, but can you come up with any other plausible explanation for you could "STILL end up with a diploma from a Major university"? That seems quite clear.
thedufer··on Subprime loans for college hiding in plain sight
I think you've misunderstood the suggestion. "excellent instructors at the most critical time of higher education" is a reference to the two years you would spend at a full university after knocking out a bunch of credits at a community college. This would allow you to save around half the cost, while still having a name-brand degree and top professors for your higher-level classes.
thedufer··on The Merge
That doesn't materially change the situation, though. It can't be cheaper to a) do the computation, b) generate a proof, and c) validate the proof many, many times than to just do the computation. You can cut down the number of orders of magnitude worse it is, but it'll still never be as inexpensive (let alone cheaper, as the original comment suggested).
thedufer··on The Merge
> Eventually the computational capacity will reach the levels that virtually any app can be replicated inside of it

This only makes sense if blockchains can catch up to traditional methods, because applications in the real world will just continue to get more complex as compute power allows. A defining feature of blockchain computation models is that the same calculation must be run many, many times to verify it. Given that each individual one must run on the hardware that blockchain is competing with direct usage of, there's no plausible story for blockchain to close the gap.

> with a much lower fee for the developers/owners/users and often having full open source code for the components.

See above - how could running the same calculation many, many times, plus the networking costs to communicate, ever compete on price with just running it once? At least one, and more likely all, of the parties must lose in the blockchain model.

> The blockchain development environment is fundamentally less terrible than the centralised database world

Terrible in what way? It can't compute on price or compute power/latency/etc.

> eventually the computational differences between the two will be so low

Again - how? See above.

> that centralised databases will be less desirable just because do the warp in the social contract.

I can't come up with a way to parse this that has any meaning, so I'm not sure what to say about it.

> Databases encourage centralisation of power to developer/owners, with aggressive bent towards monopoly. Blockchain encouraged opennesss and portability. It’s simply a better social contract.

This seems to be your only real argument - some vague social benefit. But how do we get there? See above - blockchains are far, far more expensive to run, so why would anyone migrate real applications to them? Even if this benefit were real, which I doubt, you don't seem to have any mechanism for making it work in the face of the very real costs of doing so.

Regulation seems like the only plausible way, but anti-regulation and anti-government is a pretty explicit goal of most blockchain folks.

thedufer··on V 0.3
Your website says on the homepage "No undefined behavior". Integer overflow in C is explicitly undefined behavior. What does the phrase "undefined behavior" mean to you? Because it's clearly something very different from what it means to the rest of the industry.
thedufer··on Three Arrows Capital reportedly facing insolvency
But the 1T number is mid-crash - I think the pre-crash figure I've seen was around 3T. Adjusted for inflation, crypto is still a bit short of the dot-com crash, but not by a lot, and it doesn't seem like it's over.
thedufer··on Hacker News “Who is Hiring?” top-level comments over time
Companies posting every month could just as easily be continuously growing (or, I suppose, have consistent attrition problems). Most of the places I've worked at have had a perpetually-open listings for SWEs, despite regularly hiring them.
thedufer··on I'm retiring from the engineering profession
You're so hung up on Austin, which isn't at all relevant to this conversation. Austin has nearly an order of magnitude more people than even the largest towns under discussion here. Austin has a median income 40% higher and median home price 160% (!) higher. If anyone is cherry-picking it's the person who, when talking about typical cities in the US, picks an internationally-known large city instead of something actually representative.
thedufer··on Twitter Deal Temporarily on Hold
> Matt Levine had a pretty different take, pointing out that if Elon Musk suddenly appears to be unwilling to buy Twitter, his financing might see that and - suddenly, but perfectly legitimately - evaporate, which would be a valid reason for Musk to walk away.

That was awhile ago. Since then, the funding agreements have been made public, and Matt's most recent take is that the funding basically can't disappear unless Musk can kill the deal via another route.

thedufer··on Twitter Deal Temporarily on Hold
> Musk should just pay the break fee of $1bn, and renegotiate for $42.69 or whatever meme number he fancies. Why pay $42bn total for something when you can get it for $30bn.

That's not really how the break fee works. There's a specific performance clause that allows Twitter to force Musk to go through with the deal as long as he has the money (which, shockingly, seems to have come through). They'd only give up on that and take the break fee if the deal was truly over and they were prepared to say no to a lower offer.

← PreviousPage 2 of 34Next →