acradcliffe
@acradcliffe.bsky.social
Engineer turned software developer. Into cooking, cats, board games, and reading (sff and history primarily).
As a developer I'm always slightly baffled by how many people in the profession go "ah yes this text prediction system is the ideal tool to perform some logical operation that has minimal relation to text prediction". The implicit belief seem to be that it's actually AI not glorified auto predict.
December 19, 2025 at 6:09 PM
As a developer I'm always slightly baffled by how many people in the profession go "ah yes this text prediction system is the ideal tool to perform some logical operation that has minimal relation to text prediction". The implicit belief seem to be that it's actually AI not glorified auto predict.
If we assume that each impeachment and government shutdown is roughly equivalent to loosing a confidence vote, Trump's career would have crashed and burned several times by now under such a system.
December 10, 2025 at 1:35 AM
If we assume that each impeachment and government shutdown is roughly equivalent to loosing a confidence vote, Trump's career would have crashed and burned several times by now under such a system.
Westminster style parliamentary democracy really does seem to have structurally stronger guardrails against this sort of executive capture. It's always slightly odd as a Canadian looking at American politics to see how hard it is for you to rein in the executive branch.
December 10, 2025 at 1:35 AM
Westminster style parliamentary democracy really does seem to have structurally stronger guardrails against this sort of executive capture. It's always slightly odd as a Canadian looking at American politics to see how hard it is for you to rein in the executive branch.
I haven't looked into the source code deeply to confirm, but at initial reading of release notes and PR commentary, this potentially seems like a fairly responsible implementation (Off be default, local model support available), even of it is a kind of stupid feature.
December 8, 2025 at 5:16 PM
I haven't looked into the source code deeply to confirm, but at initial reading of release notes and PR commentary, this potentially seems like a fairly responsible implementation (Off be default, local model support available), even of it is a kind of stupid feature.
To be fair to Calibre, this seems to be off by default and doesn't sound like it does anything until you go in and hook up a model to it.
December 8, 2025 at 5:16 PM
To be fair to Calibre, this seems to be off by default and doesn't sound like it does anything until you go in and hook up a model to it.
While it's very much in the beginning stages, with improving filtration/ventilation/UVC etc... technologies, we're plausibility looking at the last century in which airborne disease is a thing. (At least in wealthy countries). Basically the air equivalent of water treatment.
September 12, 2025 at 3:52 AM
While it's very much in the beginning stages, with improving filtration/ventilation/UVC etc... technologies, we're plausibility looking at the last century in which airborne disease is a thing. (At least in wealthy countries). Basically the air equivalent of water treatment.
The movement of engineering societies towards better clean air standards in light of Covid is also likely to be a big thing looking forward.
September 12, 2025 at 3:52 AM
The movement of engineering societies towards better clean air standards in light of Covid is also likely to be a big thing looking forward.
From a very quick back of the envelope calculation onions (based on rough yearly yield/calorie density) look to produce about a third the calories/per area as potatoes.
July 1, 2025 at 4:25 AM
From a very quick back of the envelope calculation onions (based on rough yearly yield/calorie density) look to produce about a third the calories/per area as potatoes.
Looking at that account, I'm pretty sure that's satire. Horribly plausible satire, but still satire.
June 20, 2025 at 4:24 PM
Looking at that account, I'm pretty sure that's satire. Horribly plausible satire, but still satire.
I'm guessing the claim is accelerometer data e.g.:
pmc.ncbi.nlm.nih.gov/articles/PMC...
But in practice a lot of other smartwatch data is abusable once it's in government hands. Location, body temp etc...
pmc.ncbi.nlm.nih.gov/articles/PMC...
But in practice a lot of other smartwatch data is abusable once it's in government hands. Location, body temp etc...
WearSense: Detecting Autism Stereotypic Behaviors through Smartwatches
Autism is a complex developmental disorder that affects approximately 1 in 68 children (according to the recent survey conducted by the Centers for Disease Control and Prevention—CDC) in the U.S., and...
pmc.ncbi.nlm.nih.gov
May 9, 2025 at 2:52 AM
I'm guessing the claim is accelerometer data e.g.:
pmc.ncbi.nlm.nih.gov/articles/PMC...
But in practice a lot of other smartwatch data is abusable once it's in government hands. Location, body temp etc...
pmc.ncbi.nlm.nih.gov/articles/PMC...
But in practice a lot of other smartwatch data is abusable once it's in government hands. Location, body temp etc...
I am pleased to see Kingston at the top of that list, not quite as high as it should be but a hell of a lot closer than most places.
March 3, 2025 at 3:14 AM
I am pleased to see Kingston at the top of that list, not quite as high as it should be but a hell of a lot closer than most places.
How's the battery life? The portability looks very convenient.
March 1, 2025 at 11:26 PM
How's the battery life? The portability looks very convenient.
We have clearer evidence of what the Olmecs did with cacao than we do for earlier civilizations but we know that it was domesticated and consumed in some form by the Mayo-Chinchipe culture (Modern Ecuador/parts of Peru) a couple of thousand years earlier.
January 27, 2025 at 5:33 AM
We have clearer evidence of what the Olmecs did with cacao than we do for earlier civilizations but we know that it was domesticated and consumed in some form by the Mayo-Chinchipe culture (Modern Ecuador/parts of Peru) a couple of thousand years earlier.
The most coherent bot has had significant overlap with the least coherent human for a long time now. ELIZA, one of the first chatbots from the 60s falls someplace between GPT-3.5 and GPT-4 in Turing test performance see:
arxiv.org/abs/2310.20216
Maybe this isn't the metric you want to depend on?
arxiv.org/abs/2310.20216
Maybe this isn't the metric you want to depend on?
Does GPT-4 pass the Turing test?
We evaluated GPT-4 in a public online Turing test. The best-performing GPT-4 prompt passed in 49.7% of games, outperforming ELIZA (22%) and GPT-3.5 (20%), but falling short of the baseline set by huma...
arxiv.org
January 25, 2025 at 7:43 PM
The most coherent bot has had significant overlap with the least coherent human for a long time now. ELIZA, one of the first chatbots from the 60s falls someplace between GPT-3.5 and GPT-4 in Turing test performance see:
arxiv.org/abs/2310.20216
Maybe this isn't the metric you want to depend on?
arxiv.org/abs/2310.20216
Maybe this isn't the metric you want to depend on?
Getting the political will for the changes to things like building codes/residential air quality standards etc... that will be needed to make this happen will take awhile, but we're seeing movement from engineering organizations in favor of measures like ASHREA 241 which would be a solid start.
January 25, 2025 at 4:12 AM
Getting the political will for the changes to things like building codes/residential air quality standards etc... that will be needed to make this happen will take awhile, but we're seeing movement from engineering organizations in favor of measures like ASHREA 241 which would be a solid start.
(Hell, we killed an entire strain of Flu with a relatively short period widespread mask wearing, that should demonstrate that this problem isn't particularly intractable)
January 25, 2025 at 4:12 AM
(Hell, we killed an entire strain of Flu with a relatively short period widespread mask wearing, that should demonstrate that this problem isn't particularly intractable)
Widespread adoption of air cleaning has the potential to do for airborne diseases what water treatment did for waterborne disease. We didn't completely kill them but they also aren't something that the average person encounters much. This can be done using existing tech, no more breakthroughs needed
January 25, 2025 at 4:12 AM
Widespread adoption of air cleaning has the potential to do for airborne diseases what water treatment did for waterborne disease. We didn't completely kill them but they also aren't something that the average person encounters much. This can be done using existing tech, no more breakthroughs needed
Technological we're at a point where cleaning indoor air through some combination of ventilation/filtration/upper room UV/direct UVC is increasingly affordable and likely to continue decreasing in price as some of the innovations that got boosted by covid mature.
January 25, 2025 at 4:12 AM
Technological we're at a point where cleaning indoor air through some combination of ventilation/filtration/upper room UV/direct UVC is increasingly affordable and likely to continue decreasing in price as some of the innovations that got boosted by covid mature.
While the next couple of decades are likely to be a mess with so many people suffering immune damage from covid and the prospect of human H5N1 around the corner, it also seems increasingly likely to me that we could be in the last century of airborne disease.
January 25, 2025 at 4:12 AM
While the next couple of decades are likely to be a mess with so many people suffering immune damage from covid and the prospect of human H5N1 around the corner, it also seems increasingly likely to me that we could be in the last century of airborne disease.
I will also note that Kobo has a decent ebook store and is much less annoying about DRM, they still have it but they don't fiddle with new types of it much which is a lot easier for people with 3rd party ebook readers etc...
January 25, 2025 at 3:05 AM
I will also note that Kobo has a decent ebook store and is much less annoying about DRM, they still have it but they don't fiddle with new types of it much which is a lot easier for people with 3rd party ebook readers etc...
Do you know of any actual court rulings on the legality of using a tool like Caliber with deDRM etc... for the purposes of making books readable by other programs? A naive reading of Canada's law makes me think that might fall under the interoperability exemption.
January 25, 2025 at 3:05 AM
Do you know of any actual court rulings on the legality of using a tool like Caliber with deDRM etc... for the purposes of making books readable by other programs? A naive reading of Canada's law makes me think that might fall under the interoperability exemption.
It's probably worth submitting a bug report here:
www.libreoffice.org/get-help/fee...
It doubt it'll get fixed quickly, but it sounds like a serious, if niche issue. If you can provide detailed steps to reproduce the issue, doing so would help improve the project in the long-term.
www.libreoffice.org/get-help/fee...
It doubt it'll get fixed quickly, but it sounds like a serious, if niche issue. If you can provide detailed steps to reproduce the issue, doing so would help improve the project in the long-term.
January 14, 2025 at 3:04 AM
It's probably worth submitting a bug report here:
www.libreoffice.org/get-help/fee...
It doubt it'll get fixed quickly, but it sounds like a serious, if niche issue. If you can provide detailed steps to reproduce the issue, doing so would help improve the project in the long-term.
www.libreoffice.org/get-help/fee...
It doubt it'll get fixed quickly, but it sounds like a serious, if niche issue. If you can provide detailed steps to reproduce the issue, doing so would help improve the project in the long-term.
We also gave a lot of nuclear in Ontario. Something like half the province's power requirements are met by nuclear.
December 29, 2024 at 4:52 PM
We also gave a lot of nuclear in Ontario. Something like half the province's power requirements are met by nuclear.
Unless someone finds a way to make these models run in a way that actually produces a profit, the problem may turn out to be self limiting. A lot of the current worst effects depend on heavy subsidiaries so the end user doesn't bear the cost of operating the model.
December 24, 2024 at 12:37 AM
Unless someone finds a way to make these models run in a way that actually produces a profit, the problem may turn out to be self limiting. A lot of the current worst effects depend on heavy subsidiaries so the end user doesn't bear the cost of operating the model.
Ah I'd missed that. Yeah that makes perfect sense. A real photo under a fake name should make a report/ban for impersonation easier to get at least.
December 22, 2024 at 4:27 PM
Ah I'd missed that. Yeah that makes perfect sense. A real photo under a fake name should make a report/ban for impersonation easier to get at least.