They turned the lights off in 2025. The post is simply not accurate to today, they even say so, stating that apparently no one has proof of better models in 2026 improving codebases over time. I've seen it happen but the author hasn't apparently, so it's a game of "you said I said", not anything concrete.
Truly, people have been saying this since 2017 and the Apple Neural Engine has proven them wrong time immemorial.
Apple's own desktops, with the fastest Apple Silicon GPUs and TDPs 20x higher than an iPhone still can't compete for real-world datacenter use even with RDNA clustering. Apple's GPGPU architecture is behind AMD at this point, there's a reason why Apple Intelligence is critically reliant on Nvidia and Google to provide inference backends.
Indeed! Materialized views won't work here as PG doesn't support "always updated" / "auto-refreshing" materialized views natively (although you can get something similar with extensions like TimescaleDB). Left joins are the thing we're often trying to avoid though, especially when the join conditions are involved, as we've seen the planner just make poor decisions in the past at unpredictable times. That's exactly the sort of situation where you might reach for this sort of trick
Awesome! It's pretty amazing all the concepts that need to come together for even these very old engines. I found I built a better understanding of what was going on as I built out the animations as well.
Not in this case, but if you aren't building it to put humans in it at all, you can build it with higher requirements in mind. Traditionally there's no point building a plane that can withstand much over 9g.
Where are we with scripting Minecraft worlds? I don't mean in a programmatic sense, but in the sense of saying to an AI "make me a worlds with a clearing in a giant forest where skeletons are attacking creepers" and it would make the world?
> Open Source AI is challenging. Even tiny models require tremendous resources to produce, so all the things we know about "open source" for software, like a random person in Nebraska can produce a critical piece of the world's infrastructure in their spare time, don't apply.
No, but a small university team with dark time on the state school system's cluster can build a basic, functional LLM.
It will be a few years (or more) until true "open source" LLMs are available and high quality, for sure. But it's not as impenetrable as it seems, IMO.
I went down the physical-writing-digital-capture system over a 10 month period. Bought the right stylus tips, the paper-like screen cover, etc.
Ultimately went back to my moleskine that fits in my back pants pocket and I don't have to worry about battery life.
Perhaps the biggest reason why my experiment failed was that I found it difficult to look back at digital written notes beyond a week. There is no physical flip...
Equally important though, I didn't retain information quite as well when writing against a screen. I can't explain it. I wanted the system to work, but it didn't. Maybe I tried to write too verbose, more carefully than I would a paper page... hoping my more careful writing would be captured by OCR better. I don't know but it was different. Intended focus matters.
Less effort = more retention to me.
Digital writing, at least at that time, was more effort to me it seems
His rights were already being violated. He was falsely arrested and denied his rights to an attorney.
They can prosecute him (and probably will) but he's got a pretty strong civil rights claim against the federal government for a false arrest and illegal search and seizure.
You can't arrest someone without a warrant or a reasonable suspicion of a crime. These thugs had neither when they arrested him, just that he's "a terrorist". That's why they were trying to get him to confess or commit a crime. That's why they didn't want a lawyer to get involved.
MP3s have been on the way out for a long time, being replaced by music subscriptions. And "MP3 player" is a feature of other devices, not its own device. Both my Apple Watch and Garmin watch can store and play MP3s without a phone; both also support music subscriptions.
The only market is in the "enthusiast audiophile flac player thing that costs way too much", just like it is currently for vinyls and DVD players.
Chinese models are open-source for purposes of inference: you know exactly what the model architecture is and the code to run it is released under an open source license. The “binary blob” is a useful artifact that has many techniques available to use in the creation of future binary blobs. And it is fixed and reliable.
Contrast this with crap like Claude, where you have no idea how many/what size/shape models you are interacting with. They can silently inject outputs from other systems, apply steering vectors to sabotage a particular user’s work, etc.
I agree that 100% free-range organic open source models are lovely, but they also are unlikely ever to attain parity with the frontier because the training data sources are open and expose them to legal liability that closed model providers can thwart.
Essentially, yes. The points of any edge that crosses the near plane escape to the infinity and if you naively just w-divide the endpoints, the image of the lineseg is not the points between the endpoints, but actually all the other points on the line that "wrap around" through infinity. This is weird to say the least.
Moreover, all edges that are fully behind the camera will actually get projected in front of the camera because they have z<0, w<0 so the signs cancel out. I think clipping against the back plane would get rid of these, but it's still weird.
> "Both sides are the same" is the last bastion of denial.
This is hyperpartisan bullshit that you obviously wouldn't say if someone were choosing between a red shirt and a blue shirt. If you support some political candidate, defend them on the merits of their positions and capabilities, not by repeating content-free jingles.
Or even worse, based on unexplored counterfactuals: "X" did this, don't you wish you picked "Y"? As if the fact that the red shirt turned out to be flammable means that the blue shirt wouldn't have been. Then when somebody points out that the labels are the same, "whataboutism!"
Even "the last bastion" is just a cynical flourish to associate this sentiment with a saying that has value to people on its own merits.
> cope
And this is twitter/online addiction. With an opposition like this, Trump will be president until the heart attack kills him. Addicted to self-regard.
Companies are run by people, 100 billion dollar companies are ran by people who want to stay billionaires and will street accordingly. But I'm not sure what we're disagreeing on, yes the talent will migrate, AI industry will eventually settle on some none bubble equilibrium, but that doesn't mean current AI economics is sensible, or inflated hardware costs beyond yield is not danger indicator. Like yeah, risky bets are burning, most people are going to move and carry their technical expertise on instead of unalive themselves or flip burgers, but that's independent of whether business models and broader economy is going to explode.
I don't know that you can be absolutist about these things, we cede a lot of power to the government because it actually does make more sense to pay for and manage some things collectively. I'm deeply against surveillance like this but I don't think practically speaking we can get rid of it entirely. I think we should do our best to dismantle it as much as we can but enforcing the great responsibility part of great power on what's left is also a useful strategy.
All new content will be behind paywalls. This will be the new way the Internet works and I've said this for the last several years. There's no value in writing unique content just to have AI steal it and distribute it without you getting any clicks. The only way it works if it there's a licensing agreement and if they don't want to pay, then who cares, you weren't going to make any money from them anyway.
We don't know that Claude doesn't feel dread when it looks at my code. To be fair, these emotional circuits are the result of evolving in various environments. AI might well have completely new emotions that we just can't comprehend.
The current generation cannot bear the accumulated weight of all past generations. Loss and compression are inevitable.
99% of an individual's life experiences are lost when they pass. 99.99% of what's left moves from distinct memory to amalgamated collective experience within a generation.
Even for those of us who are blessed with loving friends and family, nearly everything will fade away. The adolescent me harbored mild delusions that history might care about my notes. The mature me is no longer burdened by those expectations, and my notes are better for only having a single audience.
Microsoft is a big place. I recommend focusing on the parts they do well rather than getting tribal. All of the following products and services are fantastic in 2026:
Windows Server
SQL Server
.NET/C#/F#/Nuget
Visual Studio [code]
SSMS
AAD/EntraID
Defender
DirectX
Excel
I agree that Teams still kind of sucks and that Azure is a complete dumpster fire of a cloud, on average. But that doesn't mean we should throw the proverbial baby out with the bathwater. We should employ a multi-stage filtration system and extract whatever value we can.
I read two books recently that are out of print. One of them (Georges Rey's Contemporary Philosophy of Mind) is $150 on Amazon, and out of stock. The other (Philosophical Grounds of Rationality: Intentions, Categories, and Ends) isn't on Amazon at all. Used bookstores are the only viable option.
Books are treasures. The second book is an enormously valuable source of wisdom. I am glad nobody vandalized that wisdom with marginalia and highlighting.
Why would they end up in a landfill? You can also set them up inside your garden if you end up in a regular house, or mount them to the outside wall, or sell them to the next renter of your old apartment, or sell them on eBay, or...
The panel is just a regular rooftop solar panel in a slightly different form factor. The RoI is so low because you are skipping the expensive installation phase. That's literally all there is to it. They'll last a year or 20, and you can install them anywhere there is some form of sunlight.
Energy-wise you're recouped the production cost after as little as 1 to 3 years. You have to try really hard to have a solar panel be a net negative. Just don't place it indoors or in your basement or something and it'll work out just fine.
Used to write all meeting notes by hand to help me focus on the meeting and actually retain information. Fell out of the habit but must get back into it! Definitely helps especially if it's not a topic that's naturally gripping to me.