Amazing tech demo. I would have liked to see more discussion of the cold-gas thrusters in the earlier Hacker News article. Some people seemed confused and made claims that the demo was a separate "rocket" version that used a combustion technology.
Well it is a separate rocket version, it just doesn't take jet fuel. "Invite only" for the rocket version means probably less than 5 will ever be made, if that.
I strongly suspect that the invite-only version is the tech that was described in the demo, and differs from the $250K version by the size of the air tank and compression motor.
I doubt it is actual air. When compressing gas to 10,000+ psi and then running it through a rocket motor, clean/dry nitrogen is the way to go. Normal air has moisture, which can form ice, and oxygen which gets extremely reactive when heated.
And many a rocket scientist would be insulted by the comparison of a nozzled rocket motor to the simple pipe of an air rifle.
A rocket engine doesn't require combustion, it just requires expelling a gas to produce thrust, and that the engine not be air breathing. Most rocket engines use combustion to produce the high speed gases, but there are plenty that don't (water rockets, steam rockets, plasma rockets, and... cold gas thrusters as demonstrated by Tesla)
If you take the 1998 NAEP 8th grade reading scores by race, re-weight them by the new 2024 demographic makeup of American 8th graders, you will find a predicted point decline of 4.6, rather close to the actual 4 point fall over the period.
Over the same period, reading scores actually improved somewhat for each of the 1998 poorly scoring groups (and quite a bit for Asians), but the poorly scoring groups simply make up so much more of the population now that it drags everything else down.
| Group | 1998 NAEP | 2024 NAEP | Change |
|---|---:|---:|---:|
| White | *268* | *266* | −2 |
| Black | *242* | *243* | +1 |
| Hispanic | *241* | *245* | +4 |
| Asian/Pacific Islander | *261* | *280* | +19 |
| All public-school students | *261* | *257* | −4 |
I reformatted the table that was provided in the parent comment.
Your additional table with the additional columns for 2013 absolutely helps with the shape of the data. The data for years 2013 to 2024 erases all the gains of 1998 to 2013 (and often a bit more) and shows a worrying drop (that the pessimist in me worries is just the start of the slope).
1998? I wanted a quarter century of Grade 8 and that gives me 1998 to 2024 with NAEP. 2000 reading assessment was Grade 4 only. You can do it with any pair of years.
The article says that scores began drifting down around 2012. Gemini is telling me that NAEP peaked in 2013. Elsewhere in this thread there's a claim of an inflection point around 2018 in PISA, which eyeballing Gemini's numbers lines up with NAEP.
Does this correlate with the racial proportion data?
If you graph actual NAEP since 1998 versus a 1998 group-fixed score, you'll see increases (1998-2002, 2005-2013, 2015-2017) and multiple declines twice taking the line back down to the 1998 linear trend prediction (2002-2005, 2017-2024).
It is wrong to get overly exact here though. We are talking about a overall score variance of just 10pts in the face of an achievement gap of nearly 30pts, with an absolutely massive 20% decline in white population share over the 26 year period.
If you're seeing trends in your residuals that's a violation of the independence assumption of linear models (assuming you're using about a linear model since you're talking about "linear predictions").
I'm in a different field but an arbitrary 27 years (1998-2024) of data and the autocorrelation both would get flagged in a review for me. Not to give statistics homework but you should test and correct the autocorrelation issue you're describing in the model and if the data goes farther back I would go farther back, too (with how you're describing the errors going off pattern then back on it sounds like the inference here would be at least somewhat unstable depending on year chosen).
Edit: these are the assumptions and basics of how to correct for violations, in particular you're describing a violation of assumption 2 but you should test for all of them - https://www.statology.org/linear-regression-assumptions/
I was checking if there was a response and upon reading your comments again, looks like your method is less a model and more just some ad hoc procedure. Based on what your conclusion is it sounds like a linear model based on demographic data would work and linear models are easy to train. Though by your summary there are probably going to be issues with the model and you have the same issues with your procedure, you just can't tell because it hasn't been rigorously specified.
There's a channel on YouTube called Statquest that teaches statistics in a pretty accessible way if you're interested in analyzing this sort of data.
I learned my statistics in university, not Youtube. You are confusing yourself about residuals.
To make my point in the original post, I held the group scores fixed and extrapolated combined scores based on changing population sizes. This is straightforward. Simple even.
\(\widehat S=\sum_g p_gS_g,\)
Why did I call this "linear"? (Not a "linear model", your term, because it is not). The model function actually tracks the population function. Over the 25-year period, however, this trend is in fact linear because the White-to-Hispanic population trend is linear on that timescale. Over a longer timescale it would not be.
What you're describing is a linear model, it just has an ad hoc construction instead of something like a least squares training (https://en.wikipedia.org/wiki/Linear_model). Also linear models are about linear associations and work even with non-linear predictors. Your assumption in the model is explicitly a linear association so how the predictors vary shouldn't affect it.
I don't mean to disparage amateur statistics since often there are interesting things that experts miss that amateurs have insight into but this sort of data with few data points (the data you have doesn't sound independent) and a lot of confounding factors is not trivial to accurately model. Not that everyone needs to learn statistics but this sort of data (and also this topic) probably deserves a bit more of a rigorous approach.
As a very simple example, picking a time range for the model because the model doesn't work when you go farther back adds a lot of potential bias into the model and that along with the non-independence issue (and without looking at the data myself I don't know if there are other issues) are going to lead to overconfidence in conclusions.
The challenge here is that you're assuming performance by race should be constant over time. Maybe that's true, maybe it's not, I think a richer analysis with wealth, parental education, english as a first language cofounders might be helpful.
Also if you want to entertain the theory that it is racial/genetic, it'd be important to do that analysis on a continuum since a lot of people who are "minority X" may be more than 50% white or there may be trends in that over time.
> Doesn't it strike you that poor Hispanic reading performance might actually be largely related to Spanish vs. English-language proficiency?
Hispanic != non-native English speaker
There is pretty clearly something going on that causes the results upthread, but there is nowhere near enough data in this thread to figure out what it is.
Pew Research Center says that over this period, 2000 to 2024, 78% (2000) to 68% (2024) of Hispanics spoke Spanish at home. To me, that sounds more than sufficient to explain a 20pt NAEP 8th-grade reading gap.
FWIW, until 2024, the US treated the state of being Hispanic differently from race, which seems like it could be a confounding factor. (Not that I believe that it explains any of the educational outcome data, but it’s still bizarre and IMO messy.) Here’s the updated guidance:
That's real. Asians show a massive (compared to anything else that moved) NAEP score increase over the 25-year period, which is presumably tied to changing English-language competency as a group. It matters less for the overall average score as their population share is much smaller.
If you want more data on something or other, you should state the hypothesis that you are trying to test. Mine is pretty simple and was easy to test like I did in the first post: in a world of mostly static in-group NAEP 8th-grade reading scores (like the one we live in) large shifts in population shares (like we see) dominate the overall average.
"roma aeterna" is also the sort of alias that's associated with a particular set of political views. (Not conclusive! But a rational thing for someone to notice.)
He's trying to imply that you're a crypto-fascist because you presented some hate-facts.
It's sad that there can be no straightforward discussion of issues like this. It doesn't help anyone except people who are paid from ineffectual educational and social programs.
If you don't think that what happened and is happening to black people affects our average performance on tests, you haven't eliminated racist assumptions, you've eliminated the entire black experience in America and any reason for a grudge or complaint.
Grouping a list of test performances by race shouldn't lead to an assumption of natural stupidity any more than grouping a list of incomes or wealth by race should lead to an assumption of natural laziness.
And the fact that you jump to that conclusion, and it makes you nervous, so you think it's bad to talk about the facts that remind you of that conclusion, does not help black people.
You could say the same thing about Hispanic economic migrants. You don't think that economic migrants from countries with good universities and healthy middle classes are going to score lower on tests? The ones that fly in are a different story, they're often elites, but the ones that walk in? They're bottom caste where they come from, and looking for a chance to change things.
In September the Russians accused the Ukrainians of having repairing downed Geran drone airframes at the Lviv LORTA plant, transferring them to Yavoriv, and launching them unarmed into Poland. They presented various evidence like serial numbers, photographs of silver-tape repairs, claims that a Ukrainian SIM-card was used, lack of impact destruction, etc.
What's true though? I don't believe that I have a way to find out.
Usually you can find out. There are probably radar records showing them coming from Russia. But people aren't that interested and it's hard to keep up with the firehose of falsehood coming out of Russia constantly.
There's no real solution other than long-term tariffs. However the President doesn't have the power to do it and Congress doesn't have the willpower. 6-month flash-in-the-pan efforts can only cause chaos.
I don't like the title. The article actually describes the process of proving that dqlite does not have the same bug as SQLite, using a TLA+ specification. The SQLite bug fix was entirely separate from what is described here.
Are they gatekeeping to protect against AI "slop" or are they gatekeeping to protect an inefficient and non-scalable review process against velocity?
I've seen policies like this in various places and they do not generally seem to be based quantitative measures of code functionality, security, and efficiency.
There are other approaches to take to deal with large numbers of incoming PRs: improved CI, AI-readable standards for AI code, better static testing, AI first-pass review, etc.
It's fine to enshrine hobbyism into your code review policy to keep things fun and human-scale. On the other hand, where projects actually matter, it is necessary to think about code review as part of an industrial process with inputs and outputs, one where this sort of thing has no place.
In my experience success in a project isn't about the amount of code that can be produced, it's more about the thoughtfulness behind features and fixes. A lot of the work going into reviewing pull requests is understanding it in the context of the project's broader goals.
This gets a lot harder when there is a giant wall of text that the pull requester doesn't understand and can't really discuss outside of asking AI (and probably getting "The reviewer is right to push back - I was too aggressive in my blah blah blah" as an answer).
A lot of (but maybe not all) pull requests are fundamentally a human to human task, even if there is a lot of technical details involved.
Part of this is an argument for size limits on PRs.
Another part is "how do we communicate project goals and ideals and standards?" The answer to that, I'd argue, is not simple, and will inherently run into scale issues. It's something that needs to be tackled as you move from a bespoke craft project to an industrial project.
Now, maybe you never move. It's okay and good for hobby projects to exist. But let's be clear about the choice there.
You can review that user’s trade history. It’s a bit weird for sure, but they didn’t just stop in to bet on those. They’ve bet huge sums on every prediction relating to Iran, Israel, sp500, recessions, etc.
They’ve been betting for 2 years but only recently starting profiting a lot.
If they're smart they'll have multiple accounts with bets each way netting to ~0 on all kinds of topics, and just omit the losing trade on the information they actually have.
That will always be true for some individuals. But the sector isn't going away. Everyone is retooling right now and it's hard to tell what this generation of AI is even going to do to jobs. I strongly suspect that the companies that manage the retooling most successfully will be hiring more people not less.
I'd be interested in more details about the 17mph collision as well. Was it a dead-center collision with a pole after hard braking? Was it a mirror clip or a curb clip or something similar? There seem to be a wide range of possibilities.
reply