anordin95

The CEO of this startup was also the CEO of Cruise during the scandal in which Cruise hit a pedestrian, lied and heavily misrepresented evidence to CA investigators. Ethically jarring upgraded to just ethically dubious?

I initially found this odd too. However, I think the catastrophic failure probability is the same as the prior system, and presumably this new design offers improvements elsewhere.

Under the 3-voting scheme, if 2 machines have the same identical failure -- catastrophe. Under the 4 distinct systems sampled from a priority queue, if the 2 machines in the sampled system have the same identical failure -- catastrophe. In either case the odds are roughly P(bit-flip) * P(exact same bit-flip).

The article only hints at the improvements of such a system with the phrasing: " simplifies the complex task", and I'm guessing this may reduce synchronization overhead or improve parallelizability. But this is a pretty big guess to be fair.

I should have mentioned that the government foots the legal bill! That is, you don't need to hire your own costly lawyer. The case will eventually be USA vs. (your old company), and you're a potential beneficiary.

I'm an American who was retaliated against in the past for collective bargaining efforts. Luckily, that's illegal here as codified by the National Labor Relations Act of 1935 (it probably is in the UK too, I'm just not as familiar with their laws). I filed a complaint with the National Labor Relations Board (NLRB) and eventually won my case, receiving compensation with interest; the Company also had to inform all employees of their collective bargaining rights digitally and physically.

Once the government shutdown ends, I highly recommend the affected American individuals file a complaint with the NLRB via their website: https://www.nlrb.gov/

I know it and do like it! Even for those that follow Diataxis, in my experience the "Explanation" sections are often lacking, especially compared to the "How-To" or "Tutorial" ones.

Python is one example that comes to mind. They do have explanations here: https://docs.python.org/3/howto/index.html. And, to be fair, they are generally excellent in my opinion! But they're far from front and center and there's much less overall content compared to the other Diataxis types I think (granted, I haven't rigorously checked).

I am a huge fan of the "Explanations" page of their docs: "These explanation pages are oriented towards deepening your understanding of Hypothesis, including its design philosophy."

I feel much more software documentation could greatly benefit from this approach. Describing the "why" and design tradeoffs helps me grok a system far better than the typical quickstart or tutorials which show snippets but offer little understanding. That is, they rarely help me answer: is this going to solve my problem and how?

I would have liked more insight into the control group (so called "matched control") hospitals. Typically, PE firms only step in when companies (in this case, hospitals) are struggling. If the PE hospitals were in dire financial straits pre-acquisition, these results don't seem too surprising.

Ideally, the control would be a set of hospitals that PE firms otherwise wanted to acquire but were blocked for reasons unrelated to financials & performance of that hospital, e.g. regulatory. Granted, I expect that might be quite rare.

To be clear, I think private equity firms have had quantifiable negative impacts in many other aspects of healthcare. For example, acquiring helicopter-rescue/air-ambulance companies and sending them out for non-emergency situations.

Mhm. You need another thread to accomplish async file reads, which is basically what aiofiles does. This isn't really to the fault of asyncio. The necessary OS primitive isn't available. See the Linux documentation for the O_NONBLOCK flag and note this part: "Note that this flag has no effect for regular files" [1]. I actually originally wrote the sockets example in this article as using file i/o until I came across this bump in the road.

[1] https://man7.org/linux/man-pages/man2/open.2.html

Mmm fair point! Though, coroutines aren't really asynchronous objects in that usage, right? Since `await coroutine` would run that coroutine synchronously.

Ah, I'm so glad to hear it. And, thank you for the nit/feedback! I generally use python3.12 for my work which doesn't error out on that line. However, python3.11 and below will raise a SyntaxError on it. I've fixed the issue there and in a few other places and pushed the changes :)

I've used Python's asyncio a couple times now, but never really felt confident in my mental model of how it fundamentally works and therefore how I can best leverage it. The official docs provide decent documentation for each specific function in the package, but, in my opinion, lack a cohesive overview of the systems design and architecture. Something that could help the user understand the why and how behind the recommended patterns. And a way to help the user make informed decisions about which tool in the asyncio toolkit they ought to grab, or to recognize when asyncio is the entirely wrong toolkit. This is my attempt to fill that gap.

A high level overview of the ideas and objects that power Python's asyncio. And, some simple, illustrative examples.

Surprisingly, it appears TechCrunch's analysis regarding price ratios and trip distance is a bit off!

From the article: "This was true of the Uber and Lyft rides, too. But Obi found the shortest Waymo rides were priced 41.48% and 31.12% higher than Uber and Lyft, respectively. That gap shrunk as the rides got longer. In rides lasting between 4.3 km and 9.3 km, a Lyft cost $2.60 per km, an Uber cost $2.90 per km, and a Waymo cost $3.50 per km."

Relative increase in cost of Waymo vs. Lyft by trip-distance (inferred from the article's chart):

0.1 - 1.4km: 40%

1.4 - 2.2km: 45%

2.2 - 2.9km: 46%

2.9 - 4.3km: 39%

4.3 - 9.3km: 35%

That seems like pretty minor fluctuation and even then it doesn't entirely fit the pattern TechCrunch described.

Thanks for the suggestion. I've added a link to llamafile in the repo's README. Though, my focus was on exploring the model itself.

There are other popular ways to invoke these models, such as Ollama and Hugging-Face's general API package: transformers, but those hide the interesting details behind an API. Peel back the layers to poke, prod and understand!

Peel back the layers of the onion and other gluey-mess to gain insight into these models.

There are other popular ways to invoke these models, such as Ollama and Hugging-Face's general API package: transformers, but those hide the interesting details behind an API. I want to peel back the layers, poke, prod, understand and gain insight into these models and help you do the same.

I occasionally feel overly concerned with how much time each thing I'm doing is taking. As opposed to getting pleasantly lost in whatever it is that I'm doing. Of course it's useful to check-in on the time every so often or get calendar notifications to keep from missing events, but I realistically don't need to check the time every minute or two.

This is wonderful! I'd love something similar to this in my computer's time-display: random-noise until I click.

I rather enjoy the "cutesy" language! I view seamlessly weaving complex topics with fun and relevant asides as a mark of talent not immaturity.

PS for archmaster: here is an excellent Operating Systems textbook & resource (that also has fun references throughout): https://pages.cs.wisc.edu/~remzi/OSTEP/

Keep it up :)

I love the idea of receiving recommendations from an AI that's read every book! Great use-case. I suppose the analogous model would be a vision transformer for movies.

Though, I think the model may need some fine tuning. I can't seem to get the brilliant landmark text in philosophy and human self-understanding: "One Fish, Two Fish, Red Fish, Blue Fish" recommended. AI still has a long way to go...

I agree that some of the examples are a bit fluffy, but later in the article, they mention a much stronger study! They cite an example of high school students being split into a control and treatment group for tutoring students with a clearly observable output based on biological data.

This is a bit sad to read, but very well written. I can't help but think of schizophrenia after reading through some of the conversations.

I've been using Airflow for nearly a year in production and I'm surprised by all the positive commentary in this thread about the tool. To be fair, I've been using the GCP offered version of Airflow - Composer. I've found various components to be flaky and frustrating. For instance, large scale backfills don't seem to be well supported. I find various components of the system break when trying to large scale backfills, for instance 1M DAG runs. As another note, the scheduler also seems rather fragile and prone to crashing. My team has generally found the system good, but not rock-solid nor something to be confident in.

To be fair, it appears they fixed the math later, or at least roughly. (2 out of every 1M) x (the US population) [they cite ~311M]. That gives 622. Not far off the 620 they report in your quote.