jhallenworld

Rant on: I use it in-browser on Linux also: Microsoft's security system is horrendous. I have to log into Teams-based client meetings using incognito browser windows because Teams keeps getting into mystery login loops with regular windows. It was working for a while, but I needed to log into Intel for Altera FPGA information. Well Intel uses Microsoft Azure for identity management (I had to log into Microsoft to log into Intel). After I did this, I couldn't attend client meetings anymore. I'm pretty sure the root of the issue is that I have multiple Microsoft identities, and their security model does not handle this case well, or at least it's incomprehensible to me and I don't want to waste any more time on it. Now Microsoft also knows me via github which is screwing it up further. It tries to tie your identity to your phone number, but it will not allow multiple account on a single number. It's a nightmare. Oh yeah, Microsoft also insists on having me install a phone app for identity management, but it doesn't solve any issues (I can't log in) other than wasting my time. This is the only thing I have to use Microsoft for, and the experience sucks. F* Microsoft.

Say what you will about Google, at least I never had these issues.

Edit: I'm not the only who constantly has this issue, see: https://techcommunity.microsoft.com/t5/microsoft-teams/teams...

https://answers.microsoft.com/en-us/msteams/forum/all/new-te...

So this is one argument- another is that I'm impressed that they got their own LLM running and integrated into facebook messenger.

I ran across this interesting graphic recently:

https://www.theverge.com/2023/12/4/23987953/the-gpu-haves-an...

Suddenly facebook is useful as a search engine..

Also interesting is that Meta AI is much faster than ChatGPT from end-user's point of view, but results not quite as good. Here is a comparison:

https://www.youtube.com/watch?v=1vLvLN5wxS0

Yes but PDP-10 is pretty good: there is only one instruction format, so the instruction set is very uniform.

It's, shall we say, delightfully weird in its support of bytes and half-words (addresses).

Also: having 256K words is much better than 64 KB.

The same argument can be used for Tesla full self driving: basically it has to be (nearly) perfect, and after years of development, it's not there yet. What's different about LLMs?

Hah yes, fixed.

What a shit job, I'm amazed we (well they) were unified enough to do it, even with Pearl Harbor.

https://www.nationalww2museum.org/war/articles/american-nazi...

https://www.nationalww2museum.org/war/articles/great-debate

News today (from Nvidia!) was that Tesla isn't getting them, Elon redirected them to his new AI company.

The new AI has the same problem as AI for self driving- you can't trust it.

I like Tag's new Edge-Connect- we tried them, they are awesome:

https://www.tag-connect.com/product/ec06-ctx-20

These take much less board space (especially when considering locking holes) and work much better than Tag-Connect with and especially without the locking holes (those pin lock things are available, but they suck)- I mean they are easy to insert and stay in place, very nice for firmware people to use. They seem ambiguous about which side of the board you can plug them into, but they are actually (barely) polarized.

The main disadvantage is the expense of the connector and expense of castellated holes on the edge of your board. If you can afford the BOM cost use a surface mount 10-pin 50-mil header, or if you can afford the footprint space and some holes, use the SKEDD connector.

BTW, for SPI-flash chips: an option is to use a socket for them.

The descriptions are fun. So it needs the "Bounty Bob Stikes Back!" effect: birds deliver each letter, then stand proudly on top of the sign:

https://www.youtube.com/watch?v=18n-MBlVIvw

Interesting that there is no cache (instead it's using a prefetch queue).. "Implement multi-cycle to reduce area / power", I guess it's not one-cycle per instruction, so maybe not needed.

Well, where is the superscalar variant? :-) At least on an FPGA, there would be very diminishing returns for doing it. Even bypassing has bad effects on fMAX.

Yeah, hopeless.

Normally nuclear accidents lead to a much more significant destruction of property value, vs. actual loss of life- but on Long Island?

Forgetting that for a moment: Fukushima cleanup is estimated to cost $470 to $660 billion (https://en.wikipedia.org/wiki/Fukushima_disaster_cleanup). Just the aggregate single family home total property value (Suffolk and Nassau) is something like $650 billion (median times number of homes- I'm sure this is a massive underestimate considering the mansions on the Gold Coast and the Hamptons).

Existing insurance is not enough:

https://www.nrc.gov/reading-rm/doc-collections/fact-sheets/n...

Same story as Shoreham Long Island

https://en.wikipedia.org/wiki/Shoreham_Nuclear_Power_Plant

Built and powered up, but never used..

"...the state taking over the plant and then attaching a 3 percent surcharge to Long Island electric bills for 30 years to pay off the $6 billion price tag"

"Had the Shoreham Nuclear Power Station gone into operation as planned, it would have prevented the emission of an estimated three million tons of carbon dioxide per year"

So here's the scenario: you're using SystemVerilog, so Icarus does not work since it doesn't fully support it. You could use xsim (part of Xilinx Vivado)- it does work, but yuck, it's part of Vivado. So you end up using ModelSim- only now it's QuestaSim, and Siemens has eliminated the low cost tier (also, I'm not a huge fan the ModelSim GUI). I'm mean you might as well use ncsim or vcs..

So Verilator is the answer! Except..

- It's a two-state simulator, so some reset bugs are not found.

- Your Verilog-style testbench will not work.

- No encrypted IP support.

On the other hand:

- Verilator is fast. This is a big deal for fuzzing tests.

- The linting is awesome- it's good enough that you don't have to run the synthesis tool.

- gtkwave works well with it (you must use the FST format "--trace-fst", not VCD), then you get enums and structures. Counterintuitively, structures show up as modules not signals, but you get used to it.

You can also get free ModelSim as part of Lattice (or Questa as part of Altera) if you don't need the Xilinx IP.

Ban oil exports. Why are we selling our oil overseas?

I would also be OK with tariffs on exports, but they are not legal (it's in the constitution: concession for southern states, protecting their cotton exports).

Also: fix Venezuela.

I'm surprised that there is no mention of Operations Research in this article.

I'm actually curious if any business use Operations Research for real, vs. managers just guessing. I mean do something like Linear Programming to optimize profits based on various constraints.

3 is my desire, but sadly you are right: https://en.wikipedia.org/wiki/Assassination_of_Yitzhak_Rabin

Hah, you should make this comment on reddit, because I'm curious about the response you'll get.

https://www.reddit.com/r/litrpg/comments/tw44yu/sentient_vs_...

Sadly the technology for these scams is also very useful for legitimate new self-published authors. How much gatekeeping do we really want? I would think the review system would weed out the junk.

So here is a recent example of modern publishing:

I've been reading things on r/hfy.. which I don't think anyone would argue is high quality literature, but the stories are fun and the premiss of the entire genre works as background structure. One of the better stories is "Nature of Predators" by SpacePaladin15.

https://www.reddit.com/r/HFY/comments/u19xpa/the_nature_of_p...

Ok, so the author wants to publish on Amazon, no big deal these days, here it is:

https://www.amazon.com/Nature-Predators-Book-One/dp/B0CQ5QNW...

(I would argue that the reddit experience is actually better, because there are reader comments after each chapter. Where did the cover art come from? I don't know..)

What about the passive income? Well he's making $6K a month on Patreon. Maybe not passive since he's working on the sequel..

Well, how about an audiobook? Well patreon.com/Adastra650 has made it into an excellent audiobook on youtube:

https://www.youtube.com/playlist?list=PLm8OjwhOz-FZG_M_A1u9d...

Some of the voices are AI generated, but they are not bad.. it has me curious what software is used. The thumbnails are also AI.

All significant books require a page on TV Tropes: https://tvtropes.org/pmwiki/pmwiki.php/Literature/TheNatureO...

Naturally the story has grown out of HFY, has its own subreddit for fans and fanfiction:

https://www.reddit.com/r/NatureofPredators/

Naturally there is a NSFW version:

https://www.reddit.com/r/NatureOfPredatorsNSFW/

The Silo series (or Wool Omnibus) had similar beginnings:

https://en.wikipedia.org/wiki/Silo_(series)

People forget the simtel20 archives, with its 00_INDEX.TXT and AAAREADME.TXT

Billionair Jeffrey Yass is famous for getting his start from gambling: https://www.forbes.com/sites/antoinegara/2021/04/06/how-trad...

He's invested in TikTok, recently visited Mar-a-Lago, now suddenly Trump is against the TikTok ban. Getting off tanget, but the Palentir billionairs are anti-TikTok for some reason.. we'll see how the fight goes..

Long ago I wrote an FPGA image to convert I2C to serial for long-term logging, to catch that rare, once a week event:

https://github.com/jhallen/i2cmon

You capture the serial using minicom log to file or similar, then you can peruse the log at your leisure.

I used LFSRs all the time in early FPGAs (XC2000 and XC3000 series). These did not have fast carry chains, so LFSR made a much faster counter than binary. Used them for video timing generators.

Here's the book:

https://www.amazon.com/Human-Machine-Team-Artificial-Intelli...

I'm in the camp that thinks the two-state solution is dead.. which means we are left with an eventual one state solution. Which means they are killing their own future voters.

Perhaps it performs sentiment analysis of your social media posts.

My guess: facial recognition. It's easy, if you're a male of fighting age you're a Hamas terrorist.

The social media input is terrifying: show any Palestinian sympathies (sentiment analysis) in your posts and you're on the list.

This should be asked explicitly.

I imagine you get to tune the probability window of "person is >90% likely a Hamas terrorist" and choose how many innocent people you kill. Who set the window?

"Hamas terrorist" criteria: a male of fighting age, give higher weight to those congregating with others of fighting age. Basically take out a generation of Palestinian men and you're all set. Lovely.

>This sentence is horrifically dystopian... "in order to save time and enable the mass production of human targets without hindrances"

Reminds me of similar industrial thinking of a certain previous fascist government.

Albums released on floppy:

https://www.youtube.com/watch?v=tRm2_Uy_7bk

Sony invented the 3.5" floppy:

https://www.youtube.com/watch?v=djsyVgTGaRk

I recently went through my large box of old 3.5" disks and discovered that most of them are failing. The older 5 1/4" disks seem better, maybe because the density is lower.

Read about Mother Jones: https://en.wikipedia.org/wiki/Mother_Jones