prirun

I'd love to have speech-to-text on my iPhone that actually worked without needing to proofread and edit.

> I don't think the creators of any of those technologies were warning society that their own creations might wipe out humanity.

They are only doing this now IMO so that they can be the ones writing the regulations to keep us all "safe", just like we need identity verification to "save the children". They want to keep competition at bay so they can have a $2T IPO, and unregulated, "unsafe", Chinese, open-weight, and local AI threatens that IPO.

From https://www.globalstar.com/en-us/resource-center/articles/ba...

Band n53 adds premium midrange spectrum to the premium performance of XCOM RAN. As midband spectrum, it provides a balance of speed and coverage for 5G with download speeds up to 250 Mbps and upload speeds up to 100 Mbps, wider coverage than millimeter-wave frequencies and greater efficiency than lower bands for better throughput. If you run the wireless networks that run the world, the combination of band n53 and XCOM RAN is a marriage made in heaven. Learn more in the Terrestrial Wireless section at www.globalstar.com.

> Clustered indices would be so, so nice for Postgres to have.

Doesn't that just trade slower inserts for faster deletes?

I worked with Prime minicomputers when I was a late teen. They introduced RingNet, a token ring product, in 1979.

Fun fact: each system had an IN and an OUT coax port, and the coax cable was about as thick as a thumb. If a system died, a small hardware box short circuited, tying the IN and OUT coaxes together. This kept the ring intact, but it would also make that segment of the network longer and there were strict distance limitations between nodes. If the distance got too big, the entire ring broke. Also, if any coax cable got damaged, that also lead to a ring break. Prime had a special diagnostic program they could use on multiple nodes to try to find ring breaks.

One of the most "fun" things with Fortran is the EQUIVALENCE stmt. It specifies that two variables share the same memory, even if they are different types. You can also do array equivalences, so:

  INTEGER ARR1(10), ARR2(5)
  EQUIVALENCE (ARR1(6), ARR2)
  ARR2(1) = 10

will also set ARR1(6) to 10.

I had a 1TB Samsung 9xx Pro fail on me in a Windows 10 PC. One day the PC wouldn't boot, said there wasn't a hard drive there. I put the drive in an external USB 3 case and with Linux, I could dd from the drive, but it was running really hot and had an unusual access pattern: it would run for about 5 seconds and then pause for 30 seconds. I think the pauses are what screwed up the Windows boot because it wasn't waiting long enough for the drive to respond. I eventually did get the drive copied w/o data loss, though it exceeded even Linux's timeouts a few times during the copy and I had to restart dd with an offset to get the whole thing copied.

Who cares? They will still be all lies.

> "If AI fulfills its promise, we will encounter more beautiful and consequential ideas than we have ever seen."

It will be as it always has been: these ideas will further enrich the rich, at the expense of everyone else. We'll have our first quadrillionaire, while the masses are debating whether the minimum wage of $7.25/hr should be bumped up.

I have used Google's jump hash. As I recall, one of the main differences is that jump hash doesn't have a mechanism to remove targets, eg, a server dies and you don't want to route requests to it. Traditional consistent hashing can do that. I guess if you had 4 servers, server #4 dies, then you can go back to 3 servers by just changing num_buckets from 4 to 3. But if server 1 dies, you can't.

Jump hash does allow adding more targets and preserves the property that most request targets stay the same when adding a new target, so if you had 3 targets and add a fourth, ~8% of the requests that would have been sent to targets 1-3 are sent to target 4, evenly chosen from servers 1-3.

Or, like me, who has never been a runner, a heart rate in the 30's can mean that you have heart block, an electrical problem with your heart where beats are gradually spaced further apart and then there is a longish delay before a reset. In a recent 2-week heart monitor test, my longest delay was 3 seconds and I had 4 of these. There are different types of heart block: mine is type 1 Mobitz/Wenckebach. Not really dangerous - I've had it all my life and known about it for ~35 years. It shows up on all of the modern automatic heart/blood pressure machines because the delays cause unusually low readings whereas a nurse checking your pulse may miss it.

My average heart rate on the test was 58, lowest was 21 (for 3 seconds), and highest was 250 (for 6 beats). A few minutes after getting in bed, my heart rate regularly is in the 40s and sometimes low 30s.

I don't mind a checkbox. I don't even mind the "proof of work" types that take 10 seconds extra. But identifying traffic-y things over and over is way too much for me. I did do one screen, thinking it would let me in, but it just gave me another.

I do use Firefox. And I couldn't cancel my account myself: had to request it via email since I couldn't login. They were good about doing it right away and said they issued a refund for the balance.

I read that Internet Archive keeps 2 copies of all their data. Storj, the decentralized storage service, is currently in Chap 11 bankruptcy, mostly IMO because of incompetent management. They pay their node operators, the guys that actually maintain the hard drive space, $1.35/TB/mo to store data. (However it costs Storj 1.8x that because of erasure coding). They still have way more available capacity going unused, so it's likely they could lower the payout rate even further and have talked about doing that.

Internet Archive could buy Storj, probably for not much, keep 1 copy of their archive on the IA custom server setup they have now, keep the other copy on Storj, and free up half of the IA disk space. Storj automagically uploads data to 54 nodes, any 29 of which can reconstruct the original data (54/29 is where the 1.8 expansion factor occurs).

I've used vultr.com for years with no problems. Recently it wouldn't let me login without solving a Google captcha. I tried private windows, etc and ended up having to cancel my Vultr account and ask for a refund. Fortunately I only used them to spin up test VMs and didn't have any production stuff running there.

I refuse to pay a company for service and then be required to identify motorcycles and traffic lights every time I sign in. I went a few rounds with Vultr customer service and they said (paraphrasing) "It's not something we can fix, you have to talk to Google about it". Right... Google forced you to put their captcha on your web site.

OrderedTable in Nim appears to be the same and still have O(n) performance. On a delete, an entirely new table is constructed.

https://github.com/nim-lang/Nim/blob/devel/lib/pure/collecti...

I didn't know about Nimony and just took a look.

https://github.com/nim-lang/nimony/blob/master/lib/std/table...

OrderedTable is now a synonym for Table, ie, both are ordered. To delete a key, the key is looked up to get an index position x in a Seq, the Seq entries after x are shifted left, the Seq is trimmed of the last element, and then the entire table is rehashed because shifting the Seq invalidates the previous hash codes. The performance is better than Nim OrderedTables because at least an entirely new table isn't created, and it probably works fine for small tables, but it still isn't O(1). The performance would be similar to removing an item from a single-threaded sorted list: most painful to remove the 1st element, least painful to remove the last element, but then you still have to rehash the entire table.

> its culture is not very collaborative.

Second this. Their BDFL's personality is not suited for the job IMHO.

Example: Nim's hash table data structures (Python dict, Go map) are called tables. There are several variants of these, including OrderedTable. Deleting a key from an OrderedTable had O(n) performance because Nim built an entirely new OrderedTable, filtering out the key to be deleted. There were a couple of reasons for this:

- the internal list that preserved element order was only forward threaded, making an O(1) delete impossible

- Nim tables allowed the same key to be inserted multiple times, each with different values. I thought this was a bit crazy compared to other languages.

I tried to changed OrderedTable to use a doubly-linked list to allow O(1) deletes, and to get the multiple key feature removed.

What I didn't realize is that Araq, the BDFL, used ordered tables a lot in the Nim compiler, used the multiple value feature, and didn't want to add the memory overhead of a 2nd list to OrderedTable. His main reason was "deletes don't happen too often". At that point it became impossible to convince him that for a hash table to have O(n) delete performance was ridiculous and would be unexpected for anyone using OrderedTable. I gave up, left, and haven't been back.

I just asked Google AI how LLMs translate COBOL code containing GOTOs into Java, a language that doesn't have GOTO. It gave several examples, one of them being this:

COBOL code:

  PROCESS-DATA.
      ADD 1 TO COUNTER.
      DISPLAY COUNTER.
      IF COUNTER < 10 GO TO PROCESS-DATA.

Java translation:

  while (counter < 10) {
    counter++;
    System.out.println(counter);
  }

If "counter" is 10 on entry, the COBOL code prints 11 while the Java code prints nothing. So not only keeping old bugs, but apparently introducing new ones too!

I wrote COBOL code for a few years at a job when I was a teenager. What makes legacy COBOL code difficult IMO is it can sometimes be very hard to maintain a mental execution state when examining the code, for several reasons:

1. all variables are global, aka, WORKING-STORAGE. You list all the variables used in the program and they are accessible to the entire program.

2. programs are divided into paragraphs. Control normally flows sequentially top to bottom through paragraphs, one executing after another. Except that the PERFORM statement can drastically alter this normal flow control, and you can't tell by looking at a paragraph how it will be executed. To do that, you have to look at all PERFORM statements that mention this paragraph or any paragraph physically before it, because in COBOL you can say PERFORM PARA1 THROUGH PARA27. If PARA13 is physically between PARA1 and PARA27, it's potentially going to get executed.

3. In true legacy COBOL, before structured COBOL was a thing (circa 1985), the main control flow statement in addition to PERFORM was GOTO. Lots of flag setting, and lots of GOTOs. So in the previous example, you can't tell if PARA13 is going to get executed because any prior statement might be a GOTO PARA14, skipping execution of PARA13. But even worse, you are still under the influence of the PERFORM THRU, so after PARA27 is executed, control returns to the statement following the PERFORM THRU, wherever that was. But if you GOTO PARA27, without being under a PERFORM THRU, then PARA27 is executed followed by the next sequential paragraph. Trying to figure this out statically by looking at the program can be very difficult, especially considering PERFORMs that are nested at runtime but may not be anywhere near each other in a code listing.

Isn't this exactly what every mobile app does when the db is on a phone, and every app where the db is kept on the local machine? I do it with HashBackup and have done 35 db migrations over 17 years without much trouble. There have been 1 or 2 migrations that had a bug, but you fix that by doing another migration.

SQLite has transactional DDL: you can start a transaction, do a bunch of create tables, copy data from old tables into the new tables. If an error occurs during this and a rollback occurs, everything will be just like it was before the transaction started. If a commit occurs, the migration succeeds and everyone sees the new schema on their next transaction.

In rollback mode, only the migration thread can be active because it's a write transaction, but I'm guessing in WAL mode, readers can continue to read during the migration, as with any other write transaction. I don't use WAL mode much because for my application (HashBackup), I don't need db concurrency.

My sister uses Windows 10 in her one-person accounting business. The performance of everything she does is fine as is. The main thing she needs is stability, and the last thing she wants is any kind of upheaval of her computing environment, the processes she uses every day, the software she uses every day, etc. She has all kinds of workarounds for flaky software she has to use every day, and figuring out new workarounds for new flaky software is not something she wants at all.

Microsoft would do much better IMO if it focused on reducing problems and increasing their ability to support users rather than adding features and constantly changing the way people have to interact with their computer. A lot of people simply don't care about that crap.

Seems to me the purpose of all these releases, credits, pricing changes, harness changes, unpredictable token usages for the same task, etc. is to keep customers completely befuddled so that it's impossible to compare AI products. It's like hiring a consultant who sends invoices every month that aren't related to hours worked or project progress, but are whatever the consultant feels like billing, and you're expected to keep quiet and and keep paying.

I compare online marketplaces to ISPs: we don't expect ISPs to filter the traffic going over their network to make sure it complies with all laws, because that would be impossible. Having actually run a marketplace in a previous life, it's a similar situation. How is a marketplace with millions of products supposed to vet each individual product? About the only thing a marketplace can do is provide some kind of customer feedback mechanism, make sure customers can get refunds easily and quickly, and kick off merchants and buyers that have too many complaints (though they will probably just register again under a different name).

To limit inequality, cap the CEO's total compensation to 100x (or whatever) the lowest-paid workers' compensation.

Google says: "at the 100 publicly traded U.S. companies with the lowest median worker pay, this ratio (CEO to worker pay) is much higher, averaging 632 to 1.

Also: "From 1978 to 2024, inflation-adjusted CEO pay grew over 1,094%, while typical worker compensation increased by only 26%"

> hook it up to your browser and have it yank out all advertisements live

AI: "I'm sorry, a security guardrail prevents me from performing this operation."

Consistent hashing helps solve this. You don't divide into shards based on the id, but on the consistent hash of the id. Then when you have to add shards, it's easier to distribute data to the new shards. If you have 3 shards and expand to 4, only 1/4th of the data from each shard gets moved to the new shard.

IMO colleges and other educational institutions should back the loans, not the government/taxpayers.

It says that raising the config setting still causes things to be inappropriately deleted.

And our illustrious leader doesn't tariff Chinese RAM and allows its sale in the US.

I dunno... if we gave an AI all of these medical records as training data, wouldn't it be trained to give the same answers as the doctors already gave, without knowing whether those diagnoses were correct or not?

'Mythos “broke into almost all of our classified systems, not in weeks, but in hours.”'

And the government's response was to limit access to US citizens? I don't believe this for a minute. If Mythos could actually break into all these systems, the government would declare it a national security risk and it would never see the light of day for anyone outside government staff with security clearance.