Double Underscore
@doubleunderscore.bsky.social
Send 2000s web gifs directly to my inbox.
It's worth stressing that the diagnostics & translation uses aren't worth much for the bubble either. They're both just the same old ML that didn't end up changing the world.
W.r.t. Translation, the big difference is that LLMs can create more natural-looking text, at the cost of worse translation.
W.r.t. Translation, the big difference is that LLMs can create more natural-looking text, at the cost of worse translation.
February 14, 2026 at 6:58 PM
It's worth stressing that the diagnostics & translation uses aren't worth much for the bubble either. They're both just the same old ML that didn't end up changing the world.
W.r.t. Translation, the big difference is that LLMs can create more natural-looking text, at the cost of worse translation.
W.r.t. Translation, the big difference is that LLMs can create more natural-looking text, at the cost of worse translation.
The concern being, those are *vibes*, not data.
You *think* you are more productive, but you don't *know* it, and research suggests those feelings are not super reliable.
You *think* you are more productive, but you don't *know* it, and research suggests those feelings are not super reliable.
February 10, 2026 at 9:43 PM
The concern being, those are *vibes*, not data.
You *think* you are more productive, but you don't *know* it, and research suggests those feelings are not super reliable.
You *think* you are more productive, but you don't *know* it, and research suggests those feelings are not super reliable.
It'll never be politically acceptable to say but degrees in general should probably just be expanded a year or two.
They're gonna work until their mid-70s, and it used to be fairly normal & good for social development for people to graduate late anyway.
(Obv. need to get rid of student loans)
They're gonna work until their mid-70s, and it used to be fairly normal & good for social development for people to graduate late anyway.
(Obv. need to get rid of student loans)
February 10, 2026 at 8:25 PM
It'll never be politically acceptable to say but degrees in general should probably just be expanded a year or two.
They're gonna work until their mid-70s, and it used to be fairly normal & good for social development for people to graduate late anyway.
(Obv. need to get rid of student loans)
They're gonna work until their mid-70s, and it used to be fairly normal & good for social development for people to graduate late anyway.
(Obv. need to get rid of student loans)
The problem isn't "needing to understand the tool", it's that "writing the code" doesn't necessarily mean more productive overall.
In your example; The work isn't just writing the code for a GUI, it's that the GUI has to be designed *well*. That's why existing auto config-GUIs aren't used much.
In your example; The work isn't just writing the code for a GUI, it's that the GUI has to be designed *well*. That's why existing auto config-GUIs aren't used much.
February 10, 2026 at 8:12 PM
The problem isn't "needing to understand the tool", it's that "writing the code" doesn't necessarily mean more productive overall.
In your example; The work isn't just writing the code for a GUI, it's that the GUI has to be designed *well*. That's why existing auto config-GUIs aren't used much.
In your example; The work isn't just writing the code for a GUI, it's that the GUI has to be designed *well*. That's why existing auto config-GUIs aren't used much.
What doesn't help is that the AI tooling made available to developers keeps changing, so a lot of the research from mere months ago is already widely considered "outdated".
It will likely be a few years yet before we get good data.
It will likely be a few years yet before we get good data.
February 10, 2026 at 7:36 PM
What doesn't help is that the AI tooling made available to developers keeps changing, so a lot of the research from mere months ago is already widely considered "outdated".
It will likely be a few years yet before we get good data.
It will likely be a few years yet before we get good data.
The research suffers from this same measurement problem, and a lot of the research is also just garbage quality work.
The paper you mention ("Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity") has one of the better approaches, but requires a large sample size.
The paper you mention ("Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity") has one of the better approaches, but requires a large sample size.
February 10, 2026 at 7:35 PM
The research suffers from this same measurement problem, and a lot of the research is also just garbage quality work.
The paper you mention ("Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity") has one of the better approaches, but requires a large sample size.
The paper you mention ("Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity") has one of the better approaches, but requires a large sample size.
The whole claim of "AI makes software development more productive" has been extremely suspicious right from the start.
Software Developer productivity is an extremely hard to measure thing. It's an enormous struggle for every company in the field, and the ones pretending know, fool themselves.
Software Developer productivity is an extremely hard to measure thing. It's an enormous struggle for every company in the field, and the ones pretending know, fool themselves.
February 10, 2026 at 7:02 PM
The whole claim of "AI makes software development more productive" has been extremely suspicious right from the start.
Software Developer productivity is an extremely hard to measure thing. It's an enormous struggle for every company in the field, and the ones pretending know, fool themselves.
Software Developer productivity is an extremely hard to measure thing. It's an enormous struggle for every company in the field, and the ones pretending know, fool themselves.
The de-orbiting & associated replacement cost is only getting worse as Starlink is moving their satellite constellation to a lower orbit this year.
While they're making noise about moving to "less used orbits", Starlink is 85% of satellites, so it doesn't change the Kessler Syndrome problem much.
While they're making noise about moving to "less used orbits", Starlink is 85% of satellites, so it doesn't change the Kessler Syndrome problem much.
February 5, 2026 at 5:51 PM
The de-orbiting & associated replacement cost is only getting worse as Starlink is moving their satellite constellation to a lower orbit this year.
While they're making noise about moving to "less used orbits", Starlink is 85% of satellites, so it doesn't change the Kessler Syndrome problem much.
While they're making noise about moving to "less used orbits", Starlink is 85% of satellites, so it doesn't change the Kessler Syndrome problem much.
2008 was a fundamentally different beast w.r.t. bailouts.
Bankers should've gone to jail, yes.
But the theory behind the bailouts is solid. They were loans to handle liquidity problems, and were paid back. And they did stop a 2nd Great Depression.
AI is different, as it's no loan & not paid back
Bankers should've gone to jail, yes.
But the theory behind the bailouts is solid. They were loans to handle liquidity problems, and were paid back. And they did stop a 2nd Great Depression.
AI is different, as it's no loan & not paid back
January 31, 2026 at 4:50 PM
2008 was a fundamentally different beast w.r.t. bailouts.
Bankers should've gone to jail, yes.
But the theory behind the bailouts is solid. They were loans to handle liquidity problems, and were paid back. And they did stop a 2nd Great Depression.
AI is different, as it's no loan & not paid back
Bankers should've gone to jail, yes.
But the theory behind the bailouts is solid. They were loans to handle liquidity problems, and were paid back. And they did stop a 2nd Great Depression.
AI is different, as it's no loan & not paid back
There may well be no meaningful bailouts, the AI bubble has simply gotten too large.
There's so much money being poured into datacenters that to bail it all out would require bailouts the size of the covid stimulus package. *Every few years*. Even if they wanted to, they might not be able to do it.
There's so much money being poured into datacenters that to bail it all out would require bailouts the size of the covid stimulus package. *Every few years*. Even if they wanted to, they might not be able to do it.
January 31, 2026 at 4:12 PM
There may well be no meaningful bailouts, the AI bubble has simply gotten too large.
There's so much money being poured into datacenters that to bail it all out would require bailouts the size of the covid stimulus package. *Every few years*. Even if they wanted to, they might not be able to do it.
There's so much money being poured into datacenters that to bail it all out would require bailouts the size of the covid stimulus package. *Every few years*. Even if they wanted to, they might not be able to do it.
The answer then lies in "Manufacturing Consent"
Politicians need to talk about the problem and direness of the situation. To force into the media.
And at first, to accept they'll lose some votes to populist trash in favour of long term gains. And that's the hard part right now. They're all cowards
Politicians need to talk about the problem and direness of the situation. To force into the media.
And at first, to accept they'll lose some votes to populist trash in favour of long term gains. And that's the hard part right now. They're all cowards
January 29, 2026 at 10:28 PM
The answer then lies in "Manufacturing Consent"
Politicians need to talk about the problem and direness of the situation. To force into the media.
And at first, to accept they'll lose some votes to populist trash in favour of long term gains. And that's the hard part right now. They're all cowards
Politicians need to talk about the problem and direness of the situation. To force into the media.
And at first, to accept they'll lose some votes to populist trash in favour of long term gains. And that's the hard part right now. They're all cowards
If I had a *good* solution I'd be prime minister of half of Europe 🙃
One key observation is that most homeowners are not selfish shitbags. They care about their own interest, but not excessively so. The problem is the many of them that are mostly ignorant of the problem as they stay out of politics
One key observation is that most homeowners are not selfish shitbags. They care about their own interest, but not excessively so. The problem is the many of them that are mostly ignorant of the problem as they stay out of politics
January 29, 2026 at 10:28 PM
If I had a *good* solution I'd be prime minister of half of Europe 🙃
One key observation is that most homeowners are not selfish shitbags. They care about their own interest, but not excessively so. The problem is the many of them that are mostly ignorant of the problem as they stay out of politics
One key observation is that most homeowners are not selfish shitbags. They care about their own interest, but not excessively so. The problem is the many of them that are mostly ignorant of the problem as they stay out of politics
That is nominally true, and many homeowners love to proclaim as such, but they won't walk the walk.
Most people expect to grow old enough that they will downsize or sell their home. At that point they will cash in.
For many in the US, it is also a future buffer for healthcare expenses in old age
Most people expect to grow old enough that they will downsize or sell their home. At that point they will cash in.
For many in the US, it is also a future buffer for healthcare expenses in old age
January 29, 2026 at 10:03 PM
That is nominally true, and many homeowners love to proclaim as such, but they won't walk the walk.
Most people expect to grow old enough that they will downsize or sell their home. At that point they will cash in.
For many in the US, it is also a future buffer for healthcare expenses in old age
Most people expect to grow old enough that they will downsize or sell their home. At that point they will cash in.
For many in the US, it is also a future buffer for healthcare expenses in old age
With most of the other "boomer wealth" there's possibilities to redistribute wealth; Take pensions & tax benefits from the wealthiest boomers and keep them for the poorest.
Housing reform will hit *everyone*, there are no tax/welfare tricks to make a big house and a small house worth the same.
Housing reform will hit *everyone*, there are no tax/welfare tricks to make a big house and a small house worth the same.
January 29, 2026 at 9:51 PM
With most of the other "boomer wealth" there's possibilities to redistribute wealth; Take pensions & tax benefits from the wealthiest boomers and keep them for the poorest.
Housing reform will hit *everyone*, there are no tax/welfare tricks to make a big house and a small house worth the same.
Housing reform will hit *everyone*, there are no tax/welfare tricks to make a big house and a small house worth the same.
What makes the housing crisis tricky is that it's a crisis of politics itself
Concretely the question is: "How do you get homeowners, nearly all of them, to agree to give up 25% of their net worth for the greater good"
They're >2/3rds of the electorate. Even pissing off a few % loses the election.
Concretely the question is: "How do you get homeowners, nearly all of them, to agree to give up 25% of their net worth for the greater good"
They're >2/3rds of the electorate. Even pissing off a few % loses the election.
January 29, 2026 at 9:51 PM
What makes the housing crisis tricky is that it's a crisis of politics itself
Concretely the question is: "How do you get homeowners, nearly all of them, to agree to give up 25% of their net worth for the greater good"
They're >2/3rds of the electorate. Even pissing off a few % loses the election.
Concretely the question is: "How do you get homeowners, nearly all of them, to agree to give up 25% of their net worth for the greater good"
They're >2/3rds of the electorate. Even pissing off a few % loses the election.
They won't dare, and maybe they would even be correct in that judgement this time. 2/3rds of US households are homeowners.
Trump's saying the quiet part out loud because he's an idiot, but even in "commie" Europe it's aid for first time buyers only and exclusively in ways that also drive up prices.
Trump's saying the quiet part out loud because he's an idiot, but even in "commie" Europe it's aid for first time buyers only and exclusively in ways that also drive up prices.
January 29, 2026 at 9:39 PM
They won't dare, and maybe they would even be correct in that judgement this time. 2/3rds of US households are homeowners.
Trump's saying the quiet part out loud because he's an idiot, but even in "commie" Europe it's aid for first time buyers only and exclusively in ways that also drive up prices.
Trump's saying the quiet part out loud because he's an idiot, but even in "commie" Europe it's aid for first time buyers only and exclusively in ways that also drive up prices.
Nothing stopping them.
Bluesky is just weird and annoying about this stuff. They don't want to be running a social media, they just want to be a bunch of programmer nerds.
Nothing to done but drag them kicking and screaming into having a better platform and having better Trust & Safety.
Bluesky is just weird and annoying about this stuff. They don't want to be running a social media, they just want to be a bunch of programmer nerds.
Nothing to done but drag them kicking and screaming into having a better platform and having better Trust & Safety.
January 16, 2026 at 10:49 AM
Nothing stopping them.
Bluesky is just weird and annoying about this stuff. They don't want to be running a social media, they just want to be a bunch of programmer nerds.
Nothing to done but drag them kicking and screaming into having a better platform and having better Trust & Safety.
Bluesky is just weird and annoying about this stuff. They don't want to be running a social media, they just want to be a bunch of programmer nerds.
Nothing to done but drag them kicking and screaming into having a better platform and having better Trust & Safety.
Correct, DMs are private and secure, as exception to the rest of Bsky. (Character limit on prev post, sorry. 🙃)
But the DMs are not 'decentralized' in the way Bsky wants to be. Nothing about private accounts is impossible, Bsky'd just need to make the posts work like DMs. They don't want to.
But the DMs are not 'decentralized' in the way Bsky wants to be. Nothing about private accounts is impossible, Bsky'd just need to make the posts work like DMs. They don't want to.
January 16, 2026 at 3:16 AM
Correct, DMs are private and secure, as exception to the rest of Bsky. (Character limit on prev post, sorry. 🙃)
But the DMs are not 'decentralized' in the way Bsky wants to be. Nothing about private accounts is impossible, Bsky'd just need to make the posts work like DMs. They don't want to.
But the DMs are not 'decentralized' in the way Bsky wants to be. Nothing about private accounts is impossible, Bsky'd just need to make the posts work like DMs. They don't want to.
Your intuition is correct.
Boring story short, Bluesky's own design choices has made this hard for them. Mostly a consequence of them being weirdos insisting on making a "decentralized"/"federated" platform which is hard (boring reasons) to do without making everything public on the backend.
Boring story short, Bluesky's own design choices has made this hard for them. Mostly a consequence of them being weirdos insisting on making a "decentralized"/"federated" platform which is hard (boring reasons) to do without making everything public on the backend.
January 16, 2026 at 2:59 AM
Your intuition is correct.
Boring story short, Bluesky's own design choices has made this hard for them. Mostly a consequence of them being weirdos insisting on making a "decentralized"/"federated" platform which is hard (boring reasons) to do without making everything public on the backend.
Boring story short, Bluesky's own design choices has made this hard for them. Mostly a consequence of them being weirdos insisting on making a "decentralized"/"federated" platform which is hard (boring reasons) to do without making everything public on the backend.
(This also does not include risks that the Annoying Orange is going to respond to the recession by doing something stupid that wipes out the bonds at the same time as the equity. It would be very fair to include that scenario as a major risk, I just left it out as it's not historically comparable)
January 14, 2026 at 11:29 PM
(This also does not include risks that the Annoying Orange is going to respond to the recession by doing something stupid that wipes out the bonds at the same time as the equity. It would be very fair to include that scenario as a major risk, I just left it out as it's not historically comparable)
The catch being of course, I can't find very concrete data on how many such schemes are "proper".
OECD data (source->alt) shows that *collectively* the US has a healthy mix of equity and non-equity, but how that is distributed... 😬 The 401k system requiring active involvement does not bode well.
OECD data (source->alt) shows that *collectively* the US has a healthy mix of equity and non-equity, but how that is distributed... 😬 The 401k system requiring active involvement does not bode well.
January 14, 2026 at 11:25 PM
The catch being of course, I can't find very concrete data on how many such schemes are "proper".
OECD data (source->alt) shows that *collectively* the US has a healthy mix of equity and non-equity, but how that is distributed... 😬 The 401k system requiring active involvement does not bode well.
OECD data (source->alt) shows that *collectively* the US has a healthy mix of equity and non-equity, but how that is distributed... 😬 The 401k system requiring active involvement does not bode well.
Vanguard is going to eat an enormous 'L' on the AI bubble. But for their plans that are properly hedged; i.e. having shifted wealth to non-equity assets to hedge against precisely such a recession, those already/near retirement aren't going to suffer the kind of catastrophic loses.
January 14, 2026 at 11:25 PM
Vanguard is going to eat an enormous 'L' on the AI bubble. But for their plans that are properly hedged; i.e. having shifted wealth to non-equity assets to hedge against precisely such a recession, those already/near retirement aren't going to suffer the kind of catastrophic loses.
That is true but a different failure mode; What I was referring to is the "Employer goes bankrupt -> uh oh underfunded the pension / some other nonsense -> their employees specifically are screwed"
The Vanguard 401ks are close enough to a conventional pension (mostly because those too are Vanguard)
The Vanguard 401ks are close enough to a conventional pension (mostly because those too are Vanguard)
January 14, 2026 at 11:25 PM
That is true but a different failure mode; What I was referring to is the "Employer goes bankrupt -> uh oh underfunded the pension / some other nonsense -> their employees specifically are screwed"
The Vanguard 401ks are close enough to a conventional pension (mostly because those too are Vanguard)
The Vanguard 401ks are close enough to a conventional pension (mostly because those too are Vanguard)
Another thing is that even where the 'dotcom bubble' did cause problems, a fair amount of lessons were learnt.
Pension reform has finally gotten rid of a lot of the schemes reliant on the employer not going bankrupt. Far fewer people are going to lose huge chunks of their pension over this.
Pension reform has finally gotten rid of a lot of the schemes reliant on the employer not going bankrupt. Far fewer people are going to lose huge chunks of their pension over this.
January 14, 2026 at 10:24 PM
Another thing is that even where the 'dotcom bubble' did cause problems, a fair amount of lessons were learnt.
Pension reform has finally gotten rid of a lot of the schemes reliant on the employer not going bankrupt. Far fewer people are going to lose huge chunks of their pension over this.
Pension reform has finally gotten rid of a lot of the schemes reliant on the employer not going bankrupt. Far fewer people are going to lose huge chunks of their pension over this.
I could be incorrect, I found both styles in Ukrainian text and english->Ukrainian dictionaires seem to give differing options as "canonical".
(Soz for re-post; I tried to copy words to get the accents correct, but the clipboard scuffed up the accents)
(Soz for re-post; I tried to copy words to get the accents correct, but the clipboard scuffed up the accents)
January 11, 2026 at 1:55 AM
I could be incorrect, I found both styles in Ukrainian text and english->Ukrainian dictionaires seem to give differing options as "canonical".
(Soz for re-post; I tried to copy words to get the accents correct, but the clipboard scuffed up the accents)
(Soz for re-post; I tried to copy words to get the accents correct, but the clipboard scuffed up the accents)