TTPrograms

I don't think the posted algorithm is particularly novel, but the algorithm you cite is deeply different.

Also I note the only thing you have posted before is a link to this paper in particular.

I think there's an issue with the histogram rendering in this post. The rapid descent from the spike on the left is not consistent with high ECDF impact and the apparent binning resolution visible in the piecewise line-segments. In general histograms should not be visualized with connected line-graphs in this way - the standard bar graph depiction makes the bin-width apparent and resolves some of the issues the article needs the ECDF for (e.g. relative impact can be assessed visually by comparing the relative areas of the associated bars). The bar visualization also makes it possible to use varying bin sizes, which is extremely useful with any distribution that has tails.

The new TaskFlow API has been part of AirFlow 2.0 since its release in 2020: https://airflow.apache.org/docs/apache-airflow/stable/tutori...

AirFlow 2 was released in 2020. You're saying you knew that these issues were fixed, and then an article is published on your webpage in 2022 knowingly comparing against the technical properties of a major version release 2 years behind? That is not a good look.

FYI some of the Airflow issues are out of date / can be resolved with config changes.

AirFlow 2 is designed to support larger XCOM messages, so the guidance to only use it for small data no longer applies.

Your DAG construction overhead issue is likely due to dagbag refreshing. Airflow checks for DAG changes on a fixed interval, causing a reimport. The default period for that is fairly small, so for large deployments you will want to use a larger period (e.g. at least 5 minutes). I do not know why the default is so short (or was last I checked, anyway). Python files shouldn't do much of note on import regardless IMO.

I am not otherwise familiar with the improvements in Airflow 2, so I cannot say for sure if your other complaints still remain.

I don't think it's correct to think of a laser as a source in some thermal equilibrium. "Concentrating temperature" passively from sources in thermal equilibrium is forbidden, but there's nothing preventing "concentrating power".

Pulsed lasers bring material interactions into a highly non-linear regime - photon intensity is so high that multiple photon absorption is common. In the typical nuclear decay regime you are concerned with single photon absorption, and the gamma ray intuition is correct. There are also a number of approaches where various targets hit with ultrafast lasers produce controllable flux of gamma rays which are used in downstream experimentation.

Have we found diseases that could propagate across the world either (1) with virality greatly in excess of modern diseases, i.e. R0 >> omicron or (2) untreatable/curable and potentially latent for weeks to months?

A disease capable of coming close to ending civilization would need to have properties far beyond any disease observed so far. Either it needs to infect massive populations before we detect it, or it has to transmit over long distances (miles) despite e.g. moderate precautions like masking, air filtering. I think there's good reason to doubt such a pathogen could exist. The closest I could imagine would be an HIV-like immunodeficiency virus that can be transmitted via aerosol - but even that would have to cause disease much more severe than HIV without resistance among even .01% of the population.

Obvious no-true-scotsman. Believing that the goal of crypto is to circumvent laws regarding possession and theft is at most a fringe belief. The fact that this is at the top of HN demonstrates how devoid of merit crypto discussion here is.

I think in mechanical metamaterials the characteristic length defining the "metamaterial region" is rather the wavelength of pressure waves in the material you're considering - much like in electromagnetics you want the patterning (cell) length to be much less than the wavelength of radiation. In work like this they are effectively looking at 0.1 Hz or lower - near static loading - so I think pattern size can be quite large (around 600 m wavelength in bulk rubber for 0.1 Hz). This interpretation also replicates the localized behavior in the shock experiment videos. When the platform is dropped an impulse is applied with frequencies above the metamaterial regime for the material, so you see highly asymmetric response through the material - implying that the macroscopic "metamaterial" property characterization is insufficient to predict response, and so analysis must be done at feature scales rather than wavelength scales. The idea being that a "metamaterial" is a structure that can be treated as a bulk continuous material with a particular defined response as long as the interacting frequencies are all sufficiently low (far below the characteristic wavelength of the material).

I think the bending analysis you cite can determine the relative feature sizes desirable for certain "micro-scale" mechanical behavior, but it's possible to build a mechanical "metamaterial" much larger than that as well.

The sample distribution of viruses is incredibly important for this sort of analysis, and much of the argument here only makes sense through the lens of uniform virus sequencing. If you have imbalanced sequencing and imbalanced transmission you can also explain these differences.

The important thing is that mutations occur at a certain rate per virus per unit time. If you have an isolated population that's sequenced infrequently then (1) that strain will appear to evolve more slowly as there's a smaller population capable of mutating, and (2) once that strain is sequenced it's going to look far from what you've seen already since you haven't been tracking the intermediate mutations in this population.

The S/N ratio can be analyzed in terms of a random walk in high dimension. Variance in these walks grows over time (in terms of distance from origin, i.e. number of mutations), so the discrepancy doesn't seem super far from what's plausible under the null hypothesis. Perhaps someone can do the math on that.

The hypothesis merits further investigation, but the strength of the evidence presented here really requires some complex statistical analysis to determine if innocuous explanations fit. The analysis is far more complex than I would expect an epidemiologist or virologist to apply in the course of their work.

Thanks for the reply!

I've been investigating Julia AD recently - really hoping for arbitrary code diff, but nothing really works right now.

How broad will Diffractor's support of the language be? E.g. will it support mutation / exceptions / recursion?

Renting housing is generally not rent-seeking - that's exactly my point. If a property goes up for sale one can (1) choose to invest capital to purchase it, and (2) offer a contract for housing at a managed property to people who can't necessarily afford the capital investment of property purchase. Providing a service (housing access and property upkeep) in exchange for money is literally not economic rent-seeking (gaining wealth without any reciprocal contribution of productivity).

It's only for the particular school of Georgism (which holds that all undeveloped land value should be rightfully owned by citizens equally) from which the phrase originates that the transaction of "housing rent" is viewed as unjust enrichment (in that the majority of the cost is assumed to merely cover rights to the undeveloped land). Even through this economic lens it's difficult to make the "rent" definitions align, as the notion that a property might "go up for sale" is already a violation of the principle - the landlord is still providing a service with investment capital, they're just forced to work within the confines of the system for private land ownership.

The economic concept of "rent-seeking" as it's used now has relatively little to do with rental contracts as they pertain to housing.

https://en.wikipedia.org/wiki/Rent-seeking

E.g.: "The classic example of rent-seeking, according to Robert Shiller, is that of a property owner who installs a chain across a river that flows through his land and then hires a collector to charge passing boats a fee to lower the chain. There is nothing productive about the chain or the collector. The owner has made no improvements to the river and is not adding value in any way, directly or indirectly, except for himself. All he is doing is finding a way to make money from something that used to be free."

Staking is not any more rent seeking than mining is. Stakers calculate and propagate valid new states for the chain. I also don't see anything entrenching or perpetual about it.

It's not as though increased mining efficiency provides some net good for the Bitcoin blockchain, e.g. increased transaction throughput. It just provides a profitable edge for the miner. You'd just expect difficulty to increase, not energy usage to decrease. I don't think there's any particular reason to desire this sort of cross-miner competition that produces negligible societal gain.

I keep seeing the argument that Proof of Stake is more centralizing than Proof of Work, but it doesn't really add up with what I'm seeing in practice - Bitcoin mining centralized heavily around cheap electricity sources and companies that can afford SOTA dedicated mining ASICs. In contrast, anybody can become an ETH staker by purchasing the requisite amount of ETH (currently worth around $128,000, previously closer to $32,000) and make profitable return on it. You also retain a liquid asset that can be sold back at a later date. With the staking pools even this lockup amount can be reduced arbitrarily.

For a consumer to participate in PoW based on current miners it seems like they need to buy perpetually out-of-stock specialized ASICs for a sunk cost of $10-15k with the expectation that they'll break even in about a year and probably need to refresh in 3. The hardware management aspect as well as the variable cost of electricity seems to incentivize centralization of mining significantly.

Being able to standup a staking node with commodity hardware seems far more accessible.

I think the issue most have with the "no external stake" is that there was a common misunderstanding regarding Bitcoin value propagated for a while - that is, the cost of the consensus mechanism (compute + electricity) defines the price of Bitcoin. In reality it just sets a floor on the price of Bitcoin. The value of the dollar is not set by the cost of paper. So the "self-referential" nature of stake value and attack value just means that asset value is not pegged to the consensus mechanism in as strong a way as in PoW. As long as asset value is driven by other factors (e.g. utility) that is not really a major concern.

In practice social networks form a cornerstone of all of the unstated assumption of all consensus mechanisms. I'm more worried about supply chain compromise in wallet code than I am about an eclipse attack on a new node. At that point we know our models are too simple to make real world security comparisons.

I don't particularly care to rebut the author point by point, but "asset valuation" is an extremely common term that anybody discussing the properties of a novel currency should understand: https://www.investopedia.com/terms/a/assetvaluation.asp

In relation to that I was specifically referring to the misunderstandings present in "Nothing at stake".

You say:

"There are blockchain projects raising literally Billions of dollars on this false guarantee so it is valid to criticize them."

Are you not presupposing the correctness of the author's argument by calling it false? Have you already made up your mind?

Why do unsophisticated, redundant, vitriolic takes like this get upvoted on HN so much? Is there some common ax to grind here?

The strongest point here is the strawman presentation of the altered security model that PoS can be proven to form consensus under. Reading the source he cites is far more informative: https://blog.ethereum.org/2014/11/25/proof-stake-learned-lov...

The majority of the article frames distributed consensus mechanisms in an extremely sophomoric understanding of asset value and the PoW security model. All of these topics (including valid ETH criticisms) are discussed in much better ways in many other places.

The problem with literal conspiracies is that the probability of public disclosure increases exponentially with the number of people who are "in" on the conspiracy. This is why a priori one typically assigns low likelihood to them (especially if they require a large number of conspirators).

Do you believe animals literally conspired to transmit to humans?

I believe what you're describing is rather "the coordinated malevolent animal transmission theory".

If taxes affect upside for investors (especially in large gain outcomes) then it would definitely reduce rational risk appetite.

Say it's rational for some untaxed investor to take a chance on 1% on no less than $10M return. If after taxes somewhere that turns into a $5M outcome then you would need a 2% chance for equivalent expected value.

If the lab leak was real, covering it up to the extent they did requires a literal conspiracy (not to mention hypothetical function gain research). In the absence of concrete evidence, it is quite literally a conspiracy theory.

Are you implying that startup success in America requires exploitation of workers? I don't think most people would consider the management of early employees at successful startups to be "exploitative". Most of them are heavily incentivized.

There are specific issues with the worker models of Uber et al, but that's an orthogonal point to the general formation of startups.

1) Less upside for entrepreneurs/investors/early employees due to tax policy,

2) greater hiring/firing friction due to stronger employee protections,

3) greater fixed operating costs/overhead all around due to increased incorporation costs/bureaucratic requirements.

I don't think these things are necessarily negative, but in a world with competition between markets with high incorporation friction and low incorporation friction you'd expect entrepreneurship to concentrate in lower friction markets.

If this perspective of "startup failure is a deep moral sin insufficiently penalized by the government" is common in Europe then the lack of entrepreneurship is not very surprising.

I understand the argument that medical associations intentionally limit supply to prop up their own salaries. I don't think there's sufficient evidence to demonstrate that's a critical factor. Any form of licensing "friction" introduced could be countered by a compensation increase - unless there's evidence that they're imposing a specific cap on licensing numbers then it seems implausible.

I think the thesis should actually be the opposite. Computer science tells us that determining properties of the result of executing arbitrary code without actually executing it is extremely difficult (e.g. halting problem, static analyses etc.). So there's no reason to think "DNA is code" implies "we can predict phenotype from simple analyses of DNA inferred statistically".

(1) If there is an insufficient number of doctors then it must be that they're insufficiently compensated (factoring in cost of education) - if there's money to be made, people will do it. Education will (and does) grow to support demand, and America has no shortage of educatable young adults.

If doctors aren't being compensated well enough while health care costs are high (controlled for confounders e.g. population disease prevalence) it's hard to blame anything other than the industries drawing profit off those interactions (insurance/hospitals).

(2) Stop using "socialism" to define any form of government intervention you do not like.

You probably want to compress your figures - I think your line plots are stored in some vector format. The paper is 30MB and rendering chokes on those ultra dense figures (and the data resolution is not buying you any information). If the figures are vector format you should convert to png/jpeg etc.