Jadon Naas
@djadonn.bsky.social
I make technology do boring and interesting things! Software dev, sysadmin, cloud, and other beep boop type things
Honestly, I'd be willing to say Occam's Razor becomes a fallacy or a trap insofar as apparent simplicity isn't the same thing as real simplicity. Sure, pick the thing that appears simpler to you, but be ready to handle any unknown complexity that you missed in making that choice
December 19, 2025 at 4:42 PM
Honestly, I'd be willing to say Occam's Razor becomes a fallacy or a trap insofar as apparent simplicity isn't the same thing as real simplicity. Sure, pick the thing that appears simpler to you, but be ready to handle any unknown complexity that you missed in making that choice
Reposted by Jadon Naas
When I see folks on BlueSky mention some personal success, I make a point to reply to encourage and congratulate. It doesn't matter to me that they are strangers and I may never hear about their life again. What matters is that for the price of one reply, someone knows they're seen and supported.
December 18, 2025 at 10:41 AM
When I see folks on BlueSky mention some personal success, I make a point to reply to encourage and congratulate. It doesn't matter to me that they are strangers and I may never hear about their life again. What matters is that for the price of one reply, someone knows they're seen and supported.
As it turns out, having 16000+ filesystem mounts on a single Linux machine is not a good idea.
No one did that on purpose. It appears to be a configuration bug in Kolla-Ansible for OpenStack that mounts a Docker volume inside of another Docker volume.
Froze up systemctl something fierce
No one did that on purpose. It appears to be a configuration bug in Kolla-Ansible for OpenStack that mounts a Docker volume inside of another Docker volume.
Froze up systemctl something fierce
December 18, 2025 at 2:36 PM
As it turns out, having 16000+ filesystem mounts on a single Linux machine is not a good idea.
No one did that on purpose. It appears to be a configuration bug in Kolla-Ansible for OpenStack that mounts a Docker volume inside of another Docker volume.
Froze up systemctl something fierce
No one did that on purpose. It appears to be a configuration bug in Kolla-Ansible for OpenStack that mounts a Docker volume inside of another Docker volume.
Froze up systemctl something fierce
I saw someone on here talking about how metacognition is the way to save yourself from falling for LLM hallucinations. I think Baldur Bjarnason made a compelling case how that can precisely make you *more* vulnerable rather than keep you safe:
www.baldurbjarnason.com/2025/trustin...
www.baldurbjarnason.com/2025/trustin...
Trusting your own judgement on ‘AI’ is a huge risk
Writing at the end of the world, from Hveragerði, Iceland
www.baldurbjarnason.com
December 18, 2025 at 2:31 PM
I saw someone on here talking about how metacognition is the way to save yourself from falling for LLM hallucinations. I think Baldur Bjarnason made a compelling case how that can precisely make you *more* vulnerable rather than keep you safe:
www.baldurbjarnason.com/2025/trustin...
www.baldurbjarnason.com/2025/trustin...
Reposted by Jadon Naas
Remember, hit www.eff.org/age for a full set of resources about age verification: Why it doesn't (and can't) work, why it hurts everyone's privacy — including kids, and what we can all do to push back.
Age Verification and Age Gating: Resource Hub
Age verification (or age-gating) laws generally require online services to check, estimate, or verify all users’ ages—often through invasive tools like ID checks, biometric scans, or other dubious “ag...
www.eff.org
December 17, 2025 at 1:53 AM
Remember, hit www.eff.org/age for a full set of resources about age verification: Why it doesn't (and can't) work, why it hurts everyone's privacy — including kids, and what we can all do to push back.
I saw the post about the top 50 Devin user finding AI code harder to review. I think one important loss from using AI to generate your code for you is missing out on constructing your own mental model of how the code fits the problem you are trying to solve and how the code fits together
December 15, 2025 at 2:50 PM
I saw the post about the top 50 Devin user finding AI code harder to review. I think one important loss from using AI to generate your code for you is missing out on constructing your own mental model of how the code fits the problem you are trying to solve and how the code fits together
Reposted by Jadon Naas
From top AI coding user globally to completely stop using it because it's so much harder to review AI code than write it myself. And they say all software developers jobs will be replaced by LLM. Life comes fast at you. 🤣
December 13, 2025 at 1:01 PM
From top AI coding user globally to completely stop using it because it's so much harder to review AI code than write it myself. And they say all software developers jobs will be replaced by LLM. Life comes fast at you. 🤣
Today I learned you can use the date command from GNU coreutils to do things "date --date='366 days ago'" to easily get the date 366 days ago, among other date math things. I was using this to write a short shell script to delete OpenSearch indices that are older than one year to run in a cron job
December 10, 2025 at 10:53 PM
Today I learned you can use the date command from GNU coreutils to do things "date --date='366 days ago'" to easily get the date 366 days ago, among other date math things. I was using this to write a short shell script to delete OpenSearch indices that are older than one year to run in a cron job
Reposted by Jadon Naas
I hope folks who think investing in their IT infrastructure is a waste of money/a cost center remember those words when the next big whatever is due in a week, everything breaks, and dozens of highly paid engineers can't do work because the team that keeps everything running was eliminated
December 10, 2025 at 9:59 PM
I hope folks who think investing in their IT infrastructure is a waste of money/a cost center remember those words when the next big whatever is due in a week, everything breaks, and dozens of highly paid engineers can't do work because the team that keeps everything running was eliminated
Apparently, Atlassian has partnered with some racing company, and, to celebrate, they are adding a "fun little racing game" as a treat in Confluence.
I'm not sure why someone thought *that's* what will get folks excited about sort of meh productivity software
I'm not sure why someone thought *that's* what will get folks excited about sort of meh productivity software
December 5, 2025 at 1:28 PM
Apparently, Atlassian has partnered with some racing company, and, to celebrate, they are adding a "fun little racing game" as a treat in Confluence.
I'm not sure why someone thought *that's* what will get folks excited about sort of meh productivity software
I'm not sure why someone thought *that's* what will get folks excited about sort of meh productivity software
Reposted by Jadon Naas
I see people asking "I have UI skills, how do I get better at UX?" My answer is: learn content design.
You can't hide behind whizbang styling when your medium is a sentence with hyperlinks to other sentences. Starting with the content also breaks you out of seeing work as a screen to be "laid out".
You can't hide behind whizbang styling when your medium is a sentence with hyperlinks to other sentences. Starting with the content also breaks you out of seeing work as a screen to be "laid out".
December 3, 2025 at 2:03 PM
I see people asking "I have UI skills, how do I get better at UX?" My answer is: learn content design.
You can't hide behind whizbang styling when your medium is a sentence with hyperlinks to other sentences. Starting with the content also breaks you out of seeing work as a screen to be "laid out".
You can't hide behind whizbang styling when your medium is a sentence with hyperlinks to other sentences. Starting with the content also breaks you out of seeing work as a screen to be "laid out".
I was reading the Wikipedia article about Context Collapse (en.wikipedia.org/wiki/Context...) to understand the term better. The article used the phrase "a surfeit of different audiences", and I would like to consider "a surfeit of audiences" as a collective noun for audiences henceforth
Context collapse - Wikipedia
en.wikipedia.org
December 2, 2025 at 9:10 PM
I was reading the Wikipedia article about Context Collapse (en.wikipedia.org/wiki/Context...) to understand the term better. The article used the phrase "a surfeit of different audiences", and I would like to consider "a surfeit of audiences" as a collective noun for audiences henceforth
I do wonder how much of OpenAI's plans to delay ads is because they can't figure out a way to actually do ads well enough to make money.
I don't have any proof for this, I grant, but declaring a "Code Red" emergency before pivoting back to what you are comfortable doing sounds like a failure mode
I don't have any proof for this, I grant, but declaring a "Code Red" emergency before pivoting back to what you are comfortable doing sounds like a failure mode
December 2, 2025 at 8:46 PM
I do wonder how much of OpenAI's plans to delay ads is because they can't figure out a way to actually do ads well enough to make money.
I don't have any proof for this, I grant, but declaring a "Code Red" emergency before pivoting back to what you are comfortable doing sounds like a failure mode
I don't have any proof for this, I grant, but declaring a "Code Red" emergency before pivoting back to what you are comfortable doing sounds like a failure mode
I once failed to convince a CEO with whom I worked for several years of this exact phenomenon. Meetings with this CEO and other folks in the company often were really frustrating b/c CEO would make some off-the-cuff remark that immediately turned into a multi-team initiative and waste of time
November 26, 2025 at 6:59 PM
I once failed to convince a CEO with whom I worked for several years of this exact phenomenon. Meetings with this CEO and other folks in the company often were really frustrating b/c CEO would make some off-the-cuff remark that immediately turned into a multi-team initiative and waste of time
After weeks of terrible customer support calls with the appliance maker Bosch, I wonder if decades of scripted customer service/support calls primed people to be more likely to accept the outputs from LLMs/generative text. There are uncanny similarities between the call scripts and AI slop text
November 12, 2025 at 3:53 PM
After weeks of terrible customer support calls with the appliance maker Bosch, I wonder if decades of scripted customer service/support calls primed people to be more likely to accept the outputs from LLMs/generative text. There are uncanny similarities between the call scripts and AI slop text
Reposted by Jadon Naas
I only knew the modern meaning of "Luddite" until recently. Turns out, the Luddites were badass, actually. So you get a thread! And now you'll know too!
October 29, 2025 at 12:38 AM
I only knew the modern meaning of "Luddite" until recently. Turns out, the Luddites were badass, actually. So you get a thread! And now you'll know too!
Reposted by Jadon Naas
Zohran Mamdani caught lying that he understood Plato’s allegory of the cave, yet when asked to explain it, he seemed to instead explain Baudrillard’s theory of Simulacra and Simulation where reality has been replaced by symbols and signs, which seems similar to Plato’s theory but is not the same
October 29, 2025 at 3:28 AM
Zohran Mamdani caught lying that he understood Plato’s allegory of the cave, yet when asked to explain it, he seemed to instead explain Baudrillard’s theory of Simulacra and Simulation where reality has been replaced by symbols and signs, which seems similar to Plato’s theory but is not the same
Reposted by Jadon Naas
TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10/NSF-...
🧵
🧵
The official home of the Python Programming Language
www.python.org
October 27, 2025 at 2:47 PM
TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10/NSF-...
🧵
🧵
Reposted by Jadon Naas
I don't care how good LLMs *might* be getting at some things, *nothing* compares to how good it feels to *learn* something new and to acquire the knowledge you need to do something you've always wanted to do but didn't know how. 💪🏻
October 2, 2025 at 1:52 PM
I don't care how good LLMs *might* be getting at some things, *nothing* compares to how good it feels to *learn* something new and to acquire the knowledge you need to do something you've always wanted to do but didn't know how. 💪🏻
Reposted by Jadon Naas
you cannot get good with a piece of software unless you kinda hate it
September 24, 2025 at 8:07 AM
you cannot get good with a piece of software unless you kinda hate it
I saw a car commercial showing vehicles going through extreme environments to testify the quality/endurance of the vehicles, and I laughed at their idea of "endurance". Don't give me a car that can drive up a sand dune.
September 8, 2025 at 2:57 PM
I saw a car commercial showing vehicles going through extreme environments to testify the quality/endurance of the vehicles, and I laughed at their idea of "endurance". Don't give me a car that can drive up a sand dune.
Reposted by Jadon Naas
Linguistics lesson for the day: Accountability is crucial to interpreting language. We make sense of something based on who we believe said it. Synthetic text extruded by LLMs was not said by anyone -- so a crucial step in the chain is broken.
>>
>>
LLMs are not a suitable technology for information access. Here is a quick summary of why not:
buttondown.com/maiht3k/arch...
But to take the SIFT framework, LLMs cannot be a source. They are synthetic text extruding machines, that's all. Text without accountability.
buttondown.com/maiht3k/arch...
But to take the SIFT framework, LLMs cannot be a source. They are synthetic text extruding machines, that's all. Text without accountability.
Information literacy and chatbots as search
By Emily This post started off as a thread I wrote and posted across social media on Sunday evening. I'm reproducing the thread (lightly edited) first and...
buttondown.com
September 7, 2025 at 5:57 PM
Linguistics lesson for the day: Accountability is crucial to interpreting language. We make sense of something based on who we believe said it. Synthetic text extruded by LLMs was not said by anyone -- so a crucial step in the chain is broken.
>>
>>
Reposted by Jadon Naas
Interesting take on rewrites 👇
I used to be more cynical about this… I thought most rewrites were ill-justified.
Promo projects, over-engineered contraptions suffering from 2.0 syndrome, etc…
Promo projects, over-engineered contraptions suffering from 2.0 syndrome, etc…
September 6, 2025 at 4:39 AM
Interesting take on rewrites 👇
Reposted by Jadon Naas
I put away my phone for the day and instead of doomscrolling, I cleaned my room. Much better use of time!
13/12, highly recommend
13/12, highly recommend
August 27, 2025 at 3:34 AM
I put away my phone for the day and instead of doomscrolling, I cleaned my room. Much better use of time!
13/12, highly recommend
13/12, highly recommend