Hacker Newsnew | past | comments | ask | show | jobs | submit | jph00's commentslogin

Not only is that not true, but in fact FastMail predates AWS by some years.

Source: I founded FastMail.


Great username :-) I'm a happy longtime Fastmail customer and I'm migrating to the new EU area.


The range is provided in the title. Presumably it's just regularly sampled over that range.


Wow mucosal innervation was around half in long covid patients - that's super worrying and would have nasty symptoms in practice.

The findings also support the hypothesis that SARS-CoV-2 may cause structural nerve damage, which is perhaps the even bigger worry. :(


What kind of symptoms? I have insane stomach issues since early 2021 that only this year started to largely subside.


Any Long Covid patient can tell you its damaged their nerves, they can feel it. Nerve pain all over and various weird zaps and shakes etc, the nervous system is clearly damaged/being damaged across the entire body.


This is I believe one of the markers for small fiber neuropathy which is now increasingly being found in association with long covid and related syndromes.


All the B12 and gabapentin in the world won't reverse the damage, but at least there's hope.


Heaven forbid we let poor people use software that helps their kids read.

If they can't afford a tutor, they deserve nothing.

(Am I doing this right?)


Your reaction is based on 3 unproven assertions:

* An AI tutor is a net positive in learning for the subject matter it covers.

* An AI tutor does not cause other harms.

* An AI tutor is going to be cheap enough that someone who cannot afford a human tutor will still be able to afford an AI tutor.

I'm mostly willing to give the benefit of the doubt on the first point, but the third point seems unlikely, and history has given us no shortage of reasons to distrust tech companies on the second point, even if we assume this company can be trusted now.


The second one feels like a subset of the first. If it’s net positive, the other harms are worth it.

And I don’t think you’re really engaging with what they said, which is that it might be better than nothing, which is what is realistically available to many kids. Do you really think something like this is worse than nothing?


I limited the first point to specifically deal with learning the subject matter, but I added the second point to account for any negative externalities which may fall outside that.

A small (contrived) example: a child uses an AI tutor because they're falling behind in mathematics. The tutor does an excellent job at getting the child back up to speed and improving his/her grades in math, but the child then starts becoming dependent on AI tutors for all subjects because they've either lost interest or capability for self-learning.

We've already seen a study or two indicate extended use of AI coding agents causes some people to lose their own coding ability over time, making them dependent upon the agents. It doesn't seem far-fetched that children could become dependent upon AI tutors in a similar fashion.

That's not to mention that AI tutors require children to have computer access, which would ultimately cement in place school laptop programs I was under the impression were neutral to net negative in and of themselves.

So yes, in short, I think AI tutors have the potential to be worse than nothing. I'm not asserting they are, but I'd like to see some studies done on the topic before rolling them out for children en masse.


Your reaction is based on 3 unproven assertions:

* An AI tutor is a net negative in learning for the subject matter it covers.

* An AI tutor does cause other harms.

* An AI tutor is going to be more expensive that someone who cannot afford a human tutor will still be able to afford an AI tutor.

Seriously, my first reaction to reading this headline was a knee jerk "are you insane", but this whole thread is just people arguing out of their arse while claiming authority. As of today, it was never attempted. It is also possible that the kids gain an advantage by learning to use llms to teach themselves things, which would be positive for their future.

I don't think it's going to happen, but that's just my opinion based on essentially prejudice, not a fact


there's a big universe of solutions between getting nothing versus getting an AI 'educational' software that could do more harm than good.


[flagged]


Pie in the sky ideas that ignore the huge collective action problems, like “provide more education funding”.


Won't somebody think of the poor people's kids?!

Seriously though, it's great they're going to donate all this compute & tokens to the poor (who by definition can't afford to pay for it).


We could encourage parents to take an active role in tutoring their kids but I guess that's just entirely out of the question?

Nah. Let's have AI do it


Yeah tell them to change things.. why didn't we think of that? Provide opportunities to make parenting less stressful like here so that they can involve more.. this reflexive anti AI luddite attitude isn't productive as it's just less signal and more noise..


A reflexive anti-AI Luddite attitude is the healthy default until there is overwhelming evidence otherwise.

We should be wary of all new technology until it’s proven to be a benefit to society.


> We should be wary of all new technology until it’s proven to be a benefit to society.

Nope. Any nation that thinks like this will be outcompeted by more tech-positive nations in the long-term. It's on the luddites to demonstrate evidence of harm, if they want some use of technology banned.


>It's on the luddites to demonstrate evidence of harm, if they want some use of technology banned.

Here's one study from last year (1000 students) [1]. They showed that improved outcomes with AI are short term, and likely a result of AI doing the work for the student. Their methodology involved a test without the use of AI at a later date, and one of the groups who used AI tested much worse than the control group who did not use AI during learning.

Moreover, a tertiary group that used a specially trained AI for tutoring showed no improvement in testing outcomes over the control group.

[1] Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakçı, Ö., & Mariman, R. (2025). "Generative AI Without Guardrails Can Harm Learning: Evidence from High School Mathematics." PNAS, 122. PNAS


> It's on the luddites to demonstrate evidence of harm, if they want some use of technology banned.

But then we're told we can't put the genie back in the bottle, it's too late, so we just have to suck it up and deal with it.

There's nothing preventing us from taking a more measured approach to new technology other than choice.


It's not reflexive, I've spent a lot of time and effort considering why I think AI is bad for society.


[flagged]


In many ways electricity is bad for society. In this particular case I think the benefits outweigh the costs. I do not the think the benefits of AI outweigh the costs.


yes, an using it as an educational tool for 5-year olds is an unproven use we don't need to implement.


hmm bad attitude ZAP


"reflexive anti AI luddite attitude"

That's awfully reductive there, champ. Most critiques of AI are based on some combination of observed failings of the technology, observed failings of the tech industry writ large, and healthy skepticism in the face of Yet Another Tech Industry Hype Typhoon. Anyway, the burden of proof isn't on skeptics, it's on the technology and it's proponents, so let's see some receipts before we agree to squander limited public resources on unproven systems yeah?


>Most critiques of AI are based on some combination of observed failings of the technology

Most critiques of AI are based on experience with the cheapest AI available a year ago, by midwits unable to recognize that AI's performance on the Putnam, IMO and bar exam makes it far smarter than they are by almost every metric that was traditionally used to measure intelligence.


> Most critiques of AI are based on experience with the cheapest AI available a year ago, by midwits

You don't even need to use AI at all to have valid critiques of it

Plenty of other people are showing its flaws daily by replacing their entire personalities with AI slop

Oh it can do some impressive stuff sure. It is still bad for society


To make my earlier critique of your argument clearer, some applications of AI might be bad for society, but it does not follow that the tutoring application is also bad for society.


A responsible society would make sure that the applications have good results with adults before even thinking about applying it to children

And despite the hype, the jury is still out for how the results are for adults using AI


Nobody has tried to systematically apply AI tutoring for adults. AI tutoring for children is easier because there are more people who understand the domains that need to be taught to children and can evaluate if the children have grasped the concepts appropriately.

It is clear that a personal human tutor can achieve incredible results, and almost everybody who has revolutionized any field you might think of is a product of such a system. Scaling that would be immensely valuable to society.


> AI tutoring for children is easier

We shouldn't do questionable shit just because it's easier


We should do it because of the huge benefits to society. The fact that it is easier than doing something with marginal benefit is why we should do it first.


You have to demonstrate benefits to society before claiming you are providing them


The aforementioned people who revolutionized their fields are the demonstrated benefits. If we're going to stop a technology that could multiply that before it even gets to testing, it shouldn't be done on vibes.


Educational outcomes tank when computers are introduced to classrooms. It absolutely follows that AI as a tutoring application is bad for society.

https://fortune.com/2026/03/14/america-math-and-reading-scor...


It does not follow that anything produced by an electronic computational device is bad for children or society. I've personally seen computers teach children barely two years old to read.


Anecdote < Data. Studies consistently show declines in every trackable metric associated with education when computational devices are introduced to the process. Until actual data is presented that contradicts these findings there's no grounds for rational debate.


> Studies consistently show declines in every trackable metric associated with education when computational devices are introduced to the process.

They're doing it wrong, just as the incentives in the classroom are wrong (to increase a number showing how many children meet some artificially low standard of competency). The rational debate comes from the hundreds of people who are doing it right on their own. These days, all the top chess players practice with an engine. All the top mathletes use AoPS tools. It's the same across every academic field, and the level at which the top is performing is far beyond what it was before computers were introduced. The amount of effort it would take to teach a barely two year old to read is far beyond what most parents have to give, but with an electronic device patiently saying letter sounds whenever an infant interacts with letters, it's trivial.


I look forward to your nomination for Secretary of Education since you appear to have a handle on something trained professionals on two continents backed by billion dollar industries have failed to master.


I already told you that their incentives are wrong. Meanwhile, the people training the elite in any field use computers extensively and have been doing so for years.


> Educational outcomes tank when computers are introduced to classrooms.

Correlation does not necessarily mean causation.


Fair, but when presented with a situation where one side of an argument has evidence and the other has baseless assertions the choice is rather plain.


> You don't even need to use AI at all to have valid critiques of it

You cannot rationally criticize something if you have no up-to-date knowledge about it.


The industry hype mill and it's attendant horde of touts are working overtime to make sure the dude that mows my lawn is up to date on AI, so I'm not sure how you're going to advance the argument that anyone routinely posting to the epicenter of the AI hype typhoon is some how poorly informed on the topic.


That's like saying you can't know if a stove is hot without putting your hand on it


It takes a lot of digging to find their cache pricing - it's $0.50, which is unusually expensive.

The vast majority of input tokens are normally cached, so this is actually the price that matters. I wonder why it's so high.


Huh so that's why it's hard to find. They probably haven't properly optimized their caching, or they are just trying to make more money from there.


I'm a big fan of sqlite-utils, but I really don't like how Python (particularly 3.12+) changes how sqlite's transactions work -- the native behavior explained in the sqlite docs is much better IMO. I understand why Python had to change it (to be compatible with other databases) but I don't think it's a good model for sqlite.

Therefore, I created apsw-utils, a port of sqlite-utils to the amazingly-awesome apsw lib -- which is a really idiomatic sqlite lib for python. It's here: https://answerdotai.github.io/apswutils/

I've used it in lots of projects including in significant production stuff, and it's always worked great for me. IMO if you're serious about doing sqlite in python, at some point you'll probably want to check out apsw.


> changes how sqlite's transactions work

What specifically are you referring to? The apswutils website also does not explain.


They're probably talking about the addition of the autocommit flag, which hides more fine-grained transaction control in favor of more uniform behavior across multiple databases:

https://docs.python.org/3/library/sqlite3.html#sqlite3.Conne...

You can still use previous behavior with "legacy" mode that lets you control when transactions are opened in which isolation level.


> hides more fine-grained transaction control

In what way does having autocommit=False hide more fine-grained transaction control?

autocommit=False gives full control to the programmer to do whatever they want.


From the link in my previous post[0]:

> False: Select PEP 249-compliant transaction behaviour, implying that sqlite3 ensures a transaction is always open.

This means you don't get to control the isolation level of the transaction, because [1]:

> sqlite3 uses BEGIN DEFERRED statements when opening transactions.

If you want to use `IMMEDIATE` or `EXCLUSIVE` isolation level[2] for your sqlite transaction using the new flag, you have to set `autocommit=True` to be able to open the transaction yourself with `.execute("BEGIN IMMEDIATE")`.

However, with `autocommit=True`, the connection's `.commit()` and `.rollback()` methods will *silently do nothing* and you have to execute the respective raw SQL yourself to commit or abort your manually-opened transaction. This also concerns the context-manager behavior of the connection object, which will not commit or abort manual transactions on context exit in this case.

So, the autocommit flag becomes a little complicated and foot-gunny if you want more precise control over when exactly other readers or writers should get blocked by sqlite.

[0] https://docs.python.org/3/library/sqlite3.html#sqlite3.Conne...

[1] https://docs.python.org/3/library/sqlite3.html#sqlite3-trans...

[2] https://sqlite.org/lang_transaction.html#deferred_immediate_...


Thank you!

The link had some of the details, but not the intelligent thought linking them into a narrative, much appreciated! This is something super important I would have been foot-gunned on at some point.


Pythons options are here: https://docs.python.org/3/library/sqlite3.html

SQLite behavior is here: https://sqlite.org/lang_transaction.html . The regular implicit transactions there plus explicit where needed aren’t supported in any python mode.


> The regular implicit transactions there plus explicit where needed aren’t supported in any python mode

Specific examples would be extremely useful. You've done some work learning and deducing this stuff, others could learn if you would share and explain it.


It's both an illogical question and comes from a place of ignorance about a topic which one would expect folks nowadays to have some basic competence in.

First, the illogical part: the statement was about "inequality" to which you asked about wealth generation. These are two separate issues, and inequality is not, logically, necessarily tied to wealth generation. So "How does creating wealth hurt others?" is, at best, a non sequitur.

The ignorance part: there is a lot of empirical research over many decades showing the negative impact of wealth inequality on societies. With Google and AI there's really no good reason to ask such ignorant questions when in a moment you could educate yourself, and then ask an informed and thoughtful question instead.


You seem to have a rather limited understanding of what kinds of concurrency exist and how those needs are best met. Whether something is a server or not is not very relevant to this discussion.

SQLite is an excellent production db for many real world workloads, as has been widely documented. It is very different to Postgres, so requires learning a whole new skill set.

One way to think about it is that SQLite can work well for the parts of your system where there is naturally strong partitioning.


> SQLite can work well for the parts of your system where there is naturally strong partitioning.

Or the parts of your system that don't have big data and no need for massively concurrent writes. And that's the vast majority of systems!


You can do big data in SQLite. Concurrent writes, sure, I'd recommend something else.

If you think the majority of systems require massively concurrent writes, I think you need to look a bit harder. SQLite is, after all, the most widely deployed database system, ever.


We recently just partitioned the data into many SQLite databases and got away with it. It's telemetry data from IoT devices: one device, one database. Backups are an easy rsync job now instead of streaming a multi gigabyte database with compression that take hours. Reporting will just open each database and aggregate multi device data into another database (Duckdb, SQLite or something else, we'll see). Duckdb is not readable when locked so it's probably also going to be SQLite. Even it it's going to spit out JSON it will go into SQLite rows instead of many files.


Check out Quack for DuckDB.


I saw that but it's a bit too much friction. We'd rather attach a SQLite database and copy the results into that. Also no worries about version changes and compatibility, the types are not too restrictive. But we'll see.


Internet Explorer 6 was the most widely deployed awesome piece of software. Those that hated it need to look a bit harder.


It was not really “deployed” by a lot of people in the same sense.

It was forced upon most of us(not me, I used BeOS then Debian then FreeBSD).

I deployed phoenix.


The reason SQLite is the most deployed is that it's used by Android.


…and iOS, and Windows, and Mac OS, and Boeing, and Sony, and Firefox and Chrome and Safari…


Yes, which goes in line with the argument that claiming that it's "the most deployed" as proof of superiority or suitability for any use case is equivalent to claiming the same for Internet Explorer. It's the most deployed because it's bundled in a lot of systems, not because people are purposefully using it as a DBMS.


But it doesn't, because none of those systems are presenting SQLite to the user as something they should be using; they don't even make SQLite available to the user at all. Those systems all use SQLite internally to manage data.


It’s widely deployed as a local DB for local apps like phones, desktops, and web browsers. But it’s not the most used for distributed, concurrent web apps, which db servers were designed for. Maybe people are talking past each other, but that’s the debate I see.


Not sure why there’s a debate at all. The discussion is on using SQLite instead of jq and markdown files. People got lost on a tangent! :)


No it's not.. the context of other threads on this post (which mention jq) do not apply here. How poorly coded are you?


Like how poorly did Jesus code my DNA? Ask him. By him I mean ChatGPT Jesus mode.


What's massive amount concurrent writes? What's big data?


Such an exercise is left for the reader.


What additional skill set do you need to "learn" for SQLite? Copying files around?


Yeah these are deeply unserious people.


I couldn't find a CPAP mask that worked for me, even after 6 months of trying across 4 different masks.

Then I read studies showing oropharyngeal exercises can help treat sleep apnea. Originally I used https://snoregym.com/ for that, but then I switched to simply chewing gum once a day (being careful to chew on both sides of my mouth roughly equally), and also chewing my food a bit longer.

It totally cured my sleep apnea.


How come none of the masks worked?

I was fully prepared to have to go through the entire gamut of masks to find one that worked, because I snore so surely I would need a mask that covers my mouth at least, right?

But even while sleeping, my brain is keeping my mouth shut so the simplest, nose-only mask works perfectly, with very little leakage.


What was your severity of sleep apnea? Do you mind sharing your AHI?


Have you tested it with a polysomnography?


q and ch both exist in pinyin, and have different sounds. The q sound requires placing the tip of the tongue on the teeth, and isn’t an initial consonant we have in English, so we don’t have a way to spell it.


Ahh, I see, so it's not "Chian", it's "Tsian"?


'q' is alveolar: https://en.wikipedia.org/wiki/Alveolar_consonant. In Wade-Giles it would be "Ch'ian".


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: