ajdecon
@ajdecon.org
Building supercomputers! Former materials physicist, recovering SRE, now mostly herding cats. Perpetually a bit confused. He/him.
Currently at NVIDIA, formerly FB and LANL. Opinions mine as always.
🏡: Denver, CO
🔗 : https://www.ajdecon.org
Currently at NVIDIA, formerly FB and LANL. Opinions mine as always.
🏡: Denver, CO
🔗 : https://www.ajdecon.org
Pinned
ajdecon
@ajdecon.org
· 14d
Silly thread for a Saturday: some of the #HPC clusters I’ve worked on over the years.
First up is Cielo, a Cray XE6 I worked on at LANL! Which might actually be the prettiest supercomputer I’ve worked on.
First up is Cielo, a Cray XE6 I worked on at LANL! Which might actually be the prettiest supercomputer I’ve worked on.
Given all my recent interactions with medical professionals, personally or with family members, this doesn’t shock me at all
One consistent medical finding since 2023 has been that patients rated GPT-4 as significantly more empathetic than human doctors. (And more recent AI models have higher apparent empathy, but haven’t seen many RCTs on them yet) academic.oup.com/bmb/article/...
November 29, 2025 at 5:41 PM
Given all my recent interactions with medical professionals, personally or with family members, this doesn’t shock me at all
This was really thoughtful, and while the QP links the blog post, I’d also encourage listening to the recorded talk. The Q&A at the end was also thought-provoking.
I’m also a sucker for fluid dynamics analogies, and for highlighting the complexity of path-dependent processes.
I’m also a sucker for fluid dynamics analogies, and for highlighting the complexity of path-dependent processes.
November 29, 2025 at 4:31 PM
This was really thoughtful, and while the QP links the blog post, I’d also encourage listening to the recorded talk. The Q&A at the end was also thought-provoking.
I’m also a sucker for fluid dynamics analogies, and for highlighting the complexity of path-dependent processes.
I’m also a sucker for fluid dynamics analogies, and for highlighting the complexity of path-dependent processes.
Waking up to snow this morning was a lovely surprise
It’s officially Denver’s first snow of the season!
A whopping 0.2” at DIA = we’re on the board. First measurable snow since April 18 (225 days).
❄️ 2nd-latest first snow of the season (12/10/21).
❄️ 225 snowless days = T-3rd-longest snowless on record for Denver.
#COwx
A whopping 0.2” at DIA = we’re on the board. First measurable snow since April 18 (225 days).
❄️ 2nd-latest first snow of the season (12/10/21).
❄️ 225 snowless days = T-3rd-longest snowless on record for Denver.
#COwx
November 29, 2025 at 3:19 PM
Waking up to snow this morning was a lovely surprise
Reposted by ajdecon
The biggest technological revolution of our lifetime is underway and it's not AI:
Official data is in this week, and US solar continues to grow rapidly, with generation up 30% compared to last year!
November 29, 2025 at 12:21 AM
The biggest technological revolution of our lifetime is underway and it's not AI:
A bratty cat who was just caught licking the butter on the counter
November 27, 2025 at 4:03 PM
A bratty cat who was just caught licking the butter on the counter
Reposted by ajdecon
Silly TTRPG idea, which I may never do anything with and offer to the Internet.
Posit a world where high-power compute is highly regulated internationally due to AI concerns.
Put the players in as international inspectors going into a rogue nation that has stopped doing mandatory reporting.
Posit a world where high-power compute is highly regulated internationally due to AI concerns.
Put the players in as international inspectors going into a rogue nation that has stopped doing mandatory reporting.
November 27, 2025 at 1:26 AM
Silly TTRPG idea, which I may never do anything with and offer to the Internet.
Posit a world where high-power compute is highly regulated internationally due to AI concerns.
Put the players in as international inspectors going into a rogue nation that has stopped doing mandatory reporting.
Posit a world where high-power compute is highly regulated internationally due to AI concerns.
Put the players in as international inspectors going into a rogue nation that has stopped doing mandatory reporting.
Reposted by ajdecon
I don't quote this to dunk, that's not my intent, but one of the most critical things to internalize is that chatbots are for toolmakers. They're toolkits for building the environment *you* need. You have to demonstrate understanding of your own needs, and some taste.
okay, but that's a you problem.
and, tbh: jira-mcp is right there, and I developed a claude skill for interacting with jira via jira-cli (to save context space) in 45 minutes. these aren't tools; they're toolkits. to use them effectively, having them help you build tools is necessary.
and, tbh: jira-mcp is right there, and I developed a claude skill for interacting with jira via jira-cli (to save context space) in 45 minutes. these aren't tools; they're toolkits. to use them effectively, having them help you build tools is necessary.
November 26, 2025 at 10:26 PM
I don't quote this to dunk, that's not my intent, but one of the most critical things to internalize is that chatbots are for toolmakers. They're toolkits for building the environment *you* need. You have to demonstrate understanding of your own needs, and some taste.
Continuum was so good
I think if Canada wants to have a strong cultural identity we need to go back to doing what we're best at, making the weirdest fucking TV shows you've ever seen
November 26, 2025 at 9:35 PM
Continuum was so good
Yes, we put the Christmas tree up before Thanksgiving. We couldn’t help it, Phryne and Percy love it so much.
November 26, 2025 at 3:59 AM
Yes, we put the Christmas tree up before Thanksgiving. We couldn’t help it, Phryne and Percy love it so much.
This is the right answer
bring back NBC's Kings
If you were a despotic president, what movie would you force Hollywood to make? I want to see Quentin's Star Trek movie or maybe Kill Bill Vol 3.
November 25, 2025 at 9:36 PM
This is the right answer
Reposted by ajdecon
bring back NBC's Kings
If you were a despotic president, what movie would you force Hollywood to make? I want to see Quentin's Star Trek movie or maybe Kill Bill Vol 3.
November 25, 2025 at 8:44 PM
bring back NBC's Kings
This post from Sean Goedecke does a great job of capturing one of the key roles of senior/staff engineers: providing clear information and recommendations on technical topics.
How I provide technical clarity to non-technical leaders
My mission as a staff engineer is to provide technical clarity to the organization. Of course, I do other stuff too. I run projects, I ship code, I review PRs…
www.seangoedecke.com
November 25, 2025 at 5:18 PM
This post from Sean Goedecke does a great job of capturing one of the key roles of senior/staff engineers: providing clear information and recommendations on technical topics.
Catching up on my RSS feeds, and finally read this good post from @norootcause.surfingcomplexity.com
The problem of identifying great engineers in hiring is real. In addition to Lorin’s mentions of online presence and Github activity, this is also why so many orgs rely *heavily* on referrals.
The problem of identifying great engineers in hiring is real. In addition to Lorin’s mentions of online presence and Github activity, this is also why so many orgs rely *heavily* on referrals.
The illegible nature of software development talent
Here’s another blog post on gathering some common threads from reading recent posts. Today’s topic is about the unassuming nature of talented software engineers. The first thread was a …
surfingcomplexity.blog
November 25, 2025 at 4:56 PM
Catching up on my RSS feeds, and finally read this good post from @norootcause.surfingcomplexity.com
The problem of identifying great engineers in hiring is real. In addition to Lorin’s mentions of online presence and Github activity, this is also why so many orgs rely *heavily* on referrals.
The problem of identifying great engineers in hiring is real. In addition to Lorin’s mentions of online presence and Github activity, this is also why so many orgs rely *heavily* on referrals.
This is a good exercise, but I’d encourage computing folks in particular to do this not just for your work but also your non-tech hobbies or knowledge.
The LLMs mostly do pretty well on computing questions, actually! But that’s not exactly representative of their performance across the board.
The LLMs mostly do pretty well on computing questions, actually! But that’s not exactly representative of their performance across the board.
Saw an interesting piece of advice recently: ask the AI search engines for an answer in a field in which you are very knowledgeable.
The results will make it clear just how much they get wrong, and how brazen they are about it.
Now, apply that to every subject in which you are NOT an expert.
The results will make it clear just how much they get wrong, and how brazen they are about it.
Now, apply that to every subject in which you are NOT an expert.
Looked up “what mobile parking app do I need for <this town>” - AI summary on Google confidently gave me the wrong answer, which I ignored because **it is wrong a third of the time, about ANYTHING AND EVERYTHING** (according to the last study I read.)
The shit just doesn’t work.
The shit just doesn’t work.
November 25, 2025 at 4:23 PM
This is a good exercise, but I’d encourage computing folks in particular to do this not just for your work but also your non-tech hobbies or knowledge.
The LLMs mostly do pretty well on computing questions, actually! But that’s not exactly representative of their performance across the board.
The LLMs mostly do pretty well on computing questions, actually! But that’s not exactly representative of their performance across the board.
Ok, I intensely dislike Google’s AI overviews and many of their recent product changes.
But also… I really don’t think you could transplant the old Google search algorithm to today’s Internet and have it do nearly as well. The information environment itself has gotten a lot harder to search.
But also… I really don’t think you could transplant the old Google search algorithm to today’s Internet and have it do nearly as well. The information environment itself has gotten a lot harder to search.
Google at its peak was basically the best information retrieval system in human history and they and every competitor decided going from there to “you didn’t want answers you wanted half-assed auto-complete 80%-wrong hallucinations” in a few years was the right idea
November 25, 2025 at 4:18 PM
Ok, I intensely dislike Google’s AI overviews and many of their recent product changes.
But also… I really don’t think you could transplant the old Google search algorithm to today’s Internet and have it do nearly as well. The information environment itself has gotten a lot harder to search.
But also… I really don’t think you could transplant the old Google search algorithm to today’s Internet and have it do nearly as well. The information environment itself has gotten a lot harder to search.
Liberal Currents has produced a lot of the best pieces I’ve read this year. I’ve been a paying subscriber for a while, and very much encourage folks to subscribe or donate to the current fund if you’ve appreciated their work.
Yesterday we raised over $35,000. Today, we are excited to announce a donor who has generously offered to match the second $35,000. If you help us reach $35,000 today he will match you dollar for dollar. Please help us get there! gofund.me/638697766
November 25, 2025 at 4:11 PM
Liberal Currents has produced a lot of the best pieces I’ve read this year. I’ve been a paying subscriber for a while, and very much encourage folks to subscribe or donate to the current fund if you’ve appreciated their work.
Reposted by ajdecon
one of the things i objected to most at at GOOG over the years, from the labor perspective, is the slow decay of the blue-collar routes up into corp. you used to encounter guys who started out swapping drives in 120 degree heat in the hot aisles and who were now SREs, and that's gone.
Yes. I probably could have put it better in the original post and tried to clarify in the next one
i don't want to discuss the ways in which this is wrong, BUT there are a subset of tech jobs -- like, if you are the only sysadmin at a small baked bean company -- which have the same cultural valence as "HVAC contractor," as in "highly-paid WWC job," and have the same politics.
November 25, 2025 at 1:05 AM
one of the things i objected to most at at GOOG over the years, from the labor perspective, is the slow decay of the blue-collar routes up into corp. you used to encounter guys who started out swapping drives in 120 degree heat in the hot aisles and who were now SREs, and that's gone.
This post is buried in the thread and too perfect not to quote. But you should open the top of the thread and read the whole thing and replies, it’s amazing
The damned in hell are not more powerful for the fact that the fires do not consume them; it is part and parcel of their torment.
November 24, 2025 at 1:56 PM
This post is buried in the thread and too perfect not to quote. But you should open the top of the thread and read the whole thing and replies, it’s amazing
Reposted by ajdecon
Really important to stress that the Crown Jewels of the US higher education system were never the Ivies or elite SLACs (other countries have equivalents of these) but the well-funded, large, cheap, and excellently staffed public state university systems bringing high quality education to the masses.
One of the bragging rights that the US ed system had in the 20th century is that we didn't have education tracks. Essentially, any kid could go to a CC or state school & major in whatever they wanted to (obviously an oversimplification). I fear this aspect of the American dream is dying.
November 23, 2025 at 5:18 PM
Really important to stress that the Crown Jewels of the US higher education system were never the Ivies or elite SLACs (other countries have equivalents of these) but the well-funded, large, cheap, and excellently staffed public state university systems bringing high quality education to the masses.
Apropos of nothing, I find it difficult to take folks seriously if they make criticism their whole career.
I don’t even have a problem with criticism coming from outside the field it’s aimed at!
I just roll my eyes at critics who do not *also* participate in some field in a constructive fashion.
I don’t even have a problem with criticism coming from outside the field it’s aimed at!
I just roll my eyes at critics who do not *also* participate in some field in a constructive fashion.
November 23, 2025 at 9:24 PM
Apropos of nothing, I find it difficult to take folks seriously if they make criticism their whole career.
I don’t even have a problem with criticism coming from outside the field it’s aimed at!
I just roll my eyes at critics who do not *also* participate in some field in a constructive fashion.
I don’t even have a problem with criticism coming from outside the field it’s aimed at!
I just roll my eyes at critics who do not *also* participate in some field in a constructive fashion.
Reposted by ajdecon
nothing about this surprises me tbh.
LLMs should not "connect with" users. they are plumbing for words
LLMs should not "connect with" users. they are plumbing for words
The company essentially turned a dial that made ChatGPT more appealing and made people use it more, but sent some of them into delusional spirals.
OpenAI has since made the chatbot safer, but that comes with a tradeoff: less usage.
OpenAI has since made the chatbot safer, but that comes with a tradeoff: less usage.
November 23, 2025 at 8:47 PM
nothing about this surprises me tbh.
LLMs should not "connect with" users. they are plumbing for words
LLMs should not "connect with" users. they are plumbing for words
FWIW, Malwarebytes has also issued a correction in the original article.
Gmail reads your inbox to provide features like spam filtering, filtering promotions, and providing summary cards.
It does not, apparently, use this to train GenAI models.
Gmail reads your inbox to provide features like spam filtering, filtering promotions, and providing summary cards.
It does not, apparently, use this to train GenAI models.
November 22, 2025 at 9:54 PM
FWIW, Malwarebytes has also issued a correction in the original article.
Gmail reads your inbox to provide features like spam filtering, filtering promotions, and providing summary cards.
It does not, apparently, use this to train GenAI models.
Gmail reads your inbox to provide features like spam filtering, filtering promotions, and providing summary cards.
It does not, apparently, use this to train GenAI models.
I stayed up way too late the other night reading the new edition of @qntm.org’s excellent “There Is No Antimemetics Division”.
If you read the original, I highly recommend checking out the new version too, it’s really well-done.
If you read the original, I highly recommend checking out the new version too, it’s really well-done.
November 22, 2025 at 7:15 PM
I stayed up way too late the other night reading the new edition of @qntm.org’s excellent “There Is No Antimemetics Division”.
If you read the original, I highly recommend checking out the new version too, it’s really well-done.
If you read the original, I highly recommend checking out the new version too, it’s really well-done.
Well this seems bad
I don't ever wanna hear someone from the UK talking shit about the US court system again, lol www.theguardian.com/law/2025/nov...
MoJ to remove right to trial by jury for thousands of cases in controversial overhaul
Exclusive: Courts minister says change needed to stop criminals opting for juries to delay cases, sometimes by years, and clear huge backlog
www.theguardian.com
November 21, 2025 at 8:01 PM
Well this seems bad