@dave_andersen@hachyderm.io avatar

dave_andersen

@dave_andersen@hachyderm.io

Computer science professor at Carnegie Mellon University. Co-founder & CTO, Enriched Ag. he/him.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

Driving a new grounding rod by hand is not the most fun task I've ever undertaken.

dave_andersen,
@dave_andersen@hachyderm.io avatar

Woo!
Dave: 1, grounding rod: 0

But lest I declare victory prematurely, this is where I discovered that the spot where I could get the grounding rod to go in is farther than 25 ft from the thing I was trying to ground, and of course, I bought 25ft of copper based upon the original location. 🤣

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

Radio is wild. I was poking around with my SDR tonight and saw a transmission on 455.355Mhz (kinda near the pittsburgh police radios). I tuned it and heard... the audio from a local TV station. Quite clearly. And then after about 3 minutes it cut off (at the end of the news segment). And then resumed a few minutes later.

My best guess is I was hearing a neighbor's TV/stereo system that was inadvertently radiating the audio signal. Not sure why that frequency, though.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@mattblaze Ooooh. Yes, during a news broadcast, and it's been cutting out during what I suspect are commercial breaks, and then resuming. (Unfortunately, I lack a broadcast-capable TV so I can't actually check what's happening).

Thank you!

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

We've entered bunnyseason.

danderson, to random
@danderson@hachyderm.io avatar

The poor original maintainer of xz is on it now, and has already found another "fun" thing: https://git.tukaani.org/?p=xz.git;a=commitdiff;h=f9cf4c05edd14dedfe63833f8ccbe41b55823b00 . The configure check for enabling the Landlock sandboxing facility was subtly broken, so that Landlock support would never get enabled. The original malicious commit landed around the same timeframe as the main backdoor, also at an abnormal time of day compared to the new maintainer's historical activity pattern.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@danderson that one is deliciously clever. I didn't see it when I looked at the diff despite having been primed to look for something evil.

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

I love it. The manufacturer info page for the contactless card for the lock system where I'm staying now:

> For security-relevant applications, customers should refer to our MIFARE® DESFire® and MIFARE Plus® product families.

(It's using mifare classic, which seems to have more security holes than sendmail.)

The reason I checked it:
https://www.wired.com/story/saflok-hotel-lock-unsaflok-hack-technique/

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

Shit that goes through my head: "would it be more efficient to go outside and pee on the compost pile, or would the heat loss from opening and closing the door twice at 36F outweigh the water savings?"

dave_andersen,
@dave_andersen@hachyderm.io avatar

I don't have a problem, I can stop optimizing any time I want.

danluu, to random
@danluu@mastodon.social avatar

Coming up on a decade since https://danluu.com/cpu-bugs/ and the prediction seems to be holding up: https://www.radgametools.com/oodleintel.htm.

I wonder if we'll ever go back to the level of verification effort Intel used to put in. I suspect not in my lifetime. Historically, AMD was much worse than Intel w.r.t. serious stability bugs and, rather than AMD catching Intel, Intel seems to have caught AMD.

Intel has not caught up to Nvidia in bugginess, but maybe that's something to look forward to in the next decade.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@danluu Is this necessarily an Intel bug, though? Or could it be a "your motherboard is overclocking things behind your back" issue?

drewharwell, to random
@drewharwell@mastodon.social avatar

New: The U.S. government's subsidies for semiconductor factories are really kicking off now. Billions for companies that make the little digital brains inside cars, missiles, dishwashers, satellites, phones and fighter jets https://www.washingtonpost.com/technology/2024/02/19/biden-billion-chip-grant-new-york/

dave_andersen,
@dave_andersen@hachyderm.io avatar

@drewharwell er: does Intel no longer exist? (I assume the article meant to say Intel and not globalfoundries there, as globalfoundries is certainly not one of the top five anymore, or are you counting only foundry revenue?)

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

Listening to 6yo call out "mom!" in about seventeen different tones and volumes to try to see if one will magically summon her after she said "I can't come now, I'm busy". I guess.. experimentation and perseverance are positives?
"mom. Mom. MOOOOM. Mommy! Mooooooooooooooom. MOM!" (I'm missing some).

dave_andersen, to random
@dave_andersen@hachyderm.io avatar

Ok, duckdb is really handy sometimes.

Calculate the integral of power production, grouped by day, from both the ingested-into-duckdb component of the data, and also the streaming logged json output from the logger. That's nifty.

(Logging directly into duckdb, alas, doesn't seem to be as good an idea, but this works quite well and is awfully easy to debug/monitor with a text/json log to stare at in realtime.)

dave_andersen,
@dave_andersen@hachyderm.io avatar

@bitprophet Every "now and then" (soon to be cron), I run "copy powerlog from 'log.json'" to import the data, and then archive log.json.

Or, more specifically:
mv log.json import.json
copy powerlog from import.json
mv import.json old/timestamp.json
zstd old/timestamp.json

I've found that b/c of duck's single-process model, having lower-frequency bulk imports works better than having my logging process trying to keep the db open, because there's no single-writer+multiple readers option.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@bitprophet That's kind of why I'm poking at it too. :-) Postgres remains better for "lots of clients, db handles the concurrency".

duck is astoundingly fast for analytics but you need to let its single-process model (or just a lot of read-only readers with no writing) own the data. The fact that it can handle json and csv so well natively makes it a nice swiss army knife. And now it can integrate with postgres and sqlite to pull data from them also.

molly0xfff, to random
@molly0xfff@hachyderm.io avatar

fuck i love blogs. if i had nothing but time i would just read blogs all day.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@molly0xfff https://da-data.blogspot.com/

I should post more. Hmmm. Thanks for the implicit nudge. 😁

johncarlosbaez, (edited ) to random
@johncarlosbaez@mathstodon.xyz avatar

Can a PI on an NSF grant proposal commit some percentage of their time to the project but not collect any money? I'm trying to work for free, and someone just told me it's against the rules. I have trouble understanding the rule below.

Here's a potentially important subtlety: I'm an emeritus professor, I get a pension but no salary - but I'm "on recall" and a Professor of the Graduate Division, which is a position that allows me to apply for grants! Indeed this position was set up by my university precisely to allow emeritus faculty to apply for grants (thus giving the university overhead).

If something is against the rules, maybe there's some other way to accomplish the desired result? I have co-PIs who need money to do their part in this project, but I want to work for free.

Please only give advice if you have experience with this issue or are an expert on NSF grants. Of course I'm trying to get advice from grant experts at my university, but so far I'm getting some conflicting information.

Of course you're all encouraged to gripe and bemoan the situation.

𝗘𝗗𝗜𝗧: good news, I've found a way around the problem that seems to be widely used. I have to make sure I'm not mentioned in the budget - not even my name. But I should describe my work, and the time I plan to spend doing it, in the section of the proposal called "Facilities, Equipment or Other Resources". And I can still be a PI.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@johncarlosbaez You can't be a PI on it without an explicit time commitment and corresponding summer salary. But there's no hard and fast rule about how much time you have to commit -- it just has to be believable at proposal review. (And there's also, of course, a corresponding upper limit of your time commitment, namely, you can't commit more than 100%. 🙂

Do you have to be a co-PI? You can certainly be a collaborator, write a letter of support that you're helping work on the project, etc.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@johncarlosbaez Argh, fat fingered 2nd part. Anyway, for tenure track faculty, that salary only has to be summer salary, if your AY salary is covered from teaching. So if you say you're devoting 1/8th of you to the award, that could be as little as 1/4 month of salary; with overhead, etc., that could be under $10k/year.

For example, this is from a very small proposal I submitted .. a while ago, showing salary only, but FT benefits / fringe / overhead don't quite double it.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@johncarlosbaez We've used language like this in the past to indicate unpaid collaborators, if you wanted to go that route, though my gut is that if you feel like you being a PI will substantially increase the likelihood of success, I'd try to find the minimum feasible amount and just go that way.

dave_andersen,
@dave_andersen@hachyderm.io avatar

@johncarlosbaez The difference is that the bean counters can't count uncommitted voluntary time as part of the "is this going to be effectively managed?" time count. So they're allowed to ignore your time and ask "are these other PIs going to succeed at this thing?"

dave_andersen,
@dave_andersen@hachyderm.io avatar

@johncarlosbaez oh goodness, you just zoomed right past my expertise level when you invoked the emeritus part! :-) Do you want me to check with an emeritus colleague here who might know? Cool question.

kennwhite, to random

deleted_by_author

  • Loading...
  • dave_andersen,
    @dave_andersen@hachyderm.io avatar

    @kennwhite I still remember that one of the most aggressively hiring companies at the job fair when I was an undergrad (ok this was a while ago) was... Cargill. The largest company of whom I had never heard, and certainly not one that I thought of as aggressively recruiting software engineers. But when you move $165 billion of agriculture products a year, it turns out you really want some software engineers. Quite a few of them, in fact.

    dave_andersen, to random
    @dave_andersen@hachyderm.io avatar

    Hmmm. Well, now I know that at least one of our thermometers is horribly uncalibrated, but I have no idea which, or if it's both. 🤣

    dave_andersen,
    @dave_andersen@hachyderm.io avatar

    Finally identified the culprit. Glad second thermometer is right based upon new one. With bonus shot of the inside... Wonder what j6 interfaces.

    image/jpeg

    dave_andersen,
    @dave_andersen@hachyderm.io avatar

    Little R5F2LA67ANFP 16 bit MCU: 20mhz, 48KB flash. Yup, my thermometer has a better CPU than the first computers I had as a kid.

    Might be able to run Doom. Fortunately, that's not really my gig.

    dave_andersen,
    @dave_andersen@hachyderm.io avatar

    @darryl_ramm I don't have debug clips that tiny at home,. fortunately. We'll neglect to remember I work for a major cs/ece university with some extremely well equipped hardware labs, because, fortunately, it only has 3.5KB of RAM and that seems too small for doom. 😁

  • All
  • Subscribed
  • Moderated
  • Favorites
  • relationshipadvice
  • thenastyranch
  • rosin
  • mdbf
  • DreamBathrooms
  • khanakhh
  • magazineikmin
  • Youngstown
  • slotface
  • everett
  • Durango
  • osvaldo12
  • ethstaker
  • kavyap
  • bokunoheroacademia
  • tester
  • GTA5RPClips
  • anitta
  • normalnudes
  • InstantRegret
  • cubers
  • modclub
  • tacticalgear
  • Leos
  • HellsKitchen
  • cisconetworking
  • lostlight
  • sketchdaily
  • All magazines