The data is probably small enough and infrequently updated such that almost anything will work. If you split the items then you would have a separate episodes table with one row per episode joined to the podcasts table. If you store the whole XML in a row then you can overwrite the row when the feed changes. Doing the parsing and splitting into episodes in DB sounds better to me because your clients will otherwise constantly be repeating that work.
The US is always in some fucking war killing people from other countries, now threatening to invade 3 other countries, has killed somewhere between 2000-3000 civilians already in the Iran War and driven countless people in poor Asian countries to suffering and sometimes death because of the oil price shocks. As someone who lives in neither the US nor China, my primary concern is their impact on the rest of the world. You want to bring in wars China supported, then we have bring in all the wars the US has started or supported too [1]. Not to mention Jimmy Carter also supported Pol Pot during Khmer Rouge [2].
China hasn't started any wars, invaded any countries or kidnapped any heads of state recently. By any measure of recent history, there is no absolutely no question at all on which country has killed the most people outside its borders, its the US. Again, as someone who lives outside US & China, I don't care what goes on inside their borders I care about what they do to the rest of the world.
China has killed maybe 50 people outside its borders since 2000. The US has killed 100,000-250,000 in that time period. I know which one Id be more worried about as a third party.
life is way too short to deal with this nightmare of a language and its 50000 footguns for anything longer than a 2 line script, especially in the age of LLMs. Just write a python/TS/any real language script instead. Bash is great for the command line, it should be limited to use there.
"Security" is just a scourge on software at this point. It means 2FA on every trivial site, being logged out every few hours for no good reason, having to fuck with settings and type "disable sandbox" to run an agent in YOLO mode which still won't work over mobile, being unable to install an unsigned extension at all in firefox (not behind a setting, literally impossible - you have to get Firefox Developer Edition), sites spamming me to get passkeys which will no doubt be declared insecure and replaced by some more moronic thing when users find a way to get hacked with those too, 80 layers of access control/service identities/IAM/Oauth to host an S3 bucket, OAuth everywhere that won't even work on a headless device, banking websites that want their own special snowflake app as 2FA instead of using TOTP, banning VPNs and slowly rolling out completely real identity surveillance on every corner of the internet to "protect the children", ban open-weight models because the numbers are going to send your data to China, and it just goes on and on and on..
When is this insanity going to stop? I really think the IT security industry ought to be ashamed of itself. Security has become a totalizing value that trumps every other value - convenience, user-friendliness, privacy, hackability, openness, just anything at all in the name of MORE SECURITY.
All of this macro only ever obfuscates the real picture. The rich are getting much, much richer. The houses are more expensive because people obstruct construction. The salaries haven’t kept up with CEOs salaries or productivity because the employers have more power. None of this is some mysterious inevitable macro, it’s a conscious choice of an oligarchic system to crush people and launder the blame.
If the law was applied uniformly, I would support its continued uniform application. In the last 10 years, I don't see it being applied fairly at all, I see an oligarchy, a criminal and corrupt government and rich and powerful entities getting away with anything. The most minimally competent legal system would ask the AI companies, show us the list of all the data you've used to train and lets hash out the copyright - instead we have to pray someone leaks one tiny piece of what they trained on and then sue for that. Open-weights models are the closest thing we have to justice in the world where the legal system no longer provides justice, because at least the model trained on all of our data is given back to all of us.
All of the models stole the entirety of written knowledge on the internet to train. They are being sued for the few cases where we have some proof of what they did because of some whistleblowers, all the rest will just go unpunished. They breached Github TOS, robot.txt's, copyright, patents every form of IP protection under the sun from a billion sources. It's just ridiculous for the thieves to cry about someone else stealing from them.
You're not gonna believe what Paul Graham's blog or the discussion board related to it looks like.. straight out of the 90s. Forget LaTeX it doesn't even support images!
Don't forget AOE4! It's really a great evolution of AOE2 - more unique playstyles per civ, has about 1/2 the playerbase of AOE2 but still plenty to play online and support its own pro scene (was also in the Wololo tournament).
I wonder if this is a response to the dawning realization by all the mega rich AI Lab bagholders that these companies might not be as profitable as they want.
Believing that AI played a very small role while we know this problem was open for decades with at least a few people taking serious cracks at it is just not a coherent logical position.
Amazing, i've been looking for something like this and ended up doing transcription + diarization on a local server for now. Are you looking for contributions? Have you tried this one for diarization - https://huggingface.co/pyannote/speaker-diarization-communit... - it performed much better than Sortformer for me.
Just based on working at big companies and knowing how much goes on that isn’t visible outside, i would guess they definitely have tried this especially if it only takes a few days on the cluster. It’s very likely they have experiments with weird ideas and architectures running nonstop for years now and the grokking idea is pretty well known.
It’s good but will ultimately end up like the GPDR popups. Everything will say rent, won’t change the fact that you can’t actually own anything. The law should instead give us real ownership.
Wow this is great, it explains what I've seen. My Codex sub seems to last way longer than Claude in real codebases, and Claude eats up a ton of context in the initial read. I thought the harness might be the cause but it seems like tokenization is probably the bulk of it.
UPI is more like a routing/reconciliation layer like VISA that can work across any bank/wallet, and that core is run by a non-profit government org. It doesn't take custody of the funds and doesn't centralize the information into a private corporation. The fee is also minimal and flat-rate with no profit incentive, instead of paying a percentage of the entire countries transaction volume to some private company. I would guess that adds up to a huge amount of money saved overall.
I'm probably going to called a lunatic but I'm convinced this kind of telemetry is somehow linked to and behind the huge coordinated advertising push for VPNs in the last few years. More and more, the "invisible hand of the market" seems like literally the hands of a few very large conglomerations of power and capital that shape the economics of the entire market to effectively control it - they shape the gradient and make sure companies optimize loss. VPNs are either directly spyware that increases tracking capability, or are being offered now because they don't need your IP to track you anymore so they might as well make money off your fears while still tracking you. More broadly, I don't see much of a free market or democracy left anymore, now every government is doing a coordinated push to eliminate privacy as well.
Try actually playing a modern Doom game and then a modern UE5 game or look at some benchmarks. UE games mostly run like shit, whereas Doom/idTech games are the smoothest in the entire industry.
So this is tailored towards kind of a "reader view" for models right? Can it handle images, tables, shadow DOMs too? Like there are 3 use cases I have now - one is a simple text view for models to understand it, one is a "web clip" mode which would ideally preserve images and media, and one is to extract tabular data from web pages. Which ones is this good at?
Windows is leading the pack here as usual. They've introduced a new "all your RAM are belong to us" feature where all 48GB of my RAM slowly drips into a black hole over time, ne'er to be released again until reboot.
I not only rebooted but upgraded my server from 22.04 to 24.04 and every single package on it this weekend with AI help - took a few hours of talking to the thing, even though the server had accumulated some ugly hacks (like installed KDE on top of Pop OS, had a newer version of glibc than supported) and runs a ton of containers. The AI figured out all the apt and dependency hell bullshit that I would never have gotten through. Its a new world!
With all the work around AI sandboxes, microVMs, browser sandboxes, device attestation, secure boot etc. I feel like we should be able to construct a proper software sandbox that works on most PCs and guarantees that e.g nothing outside of the word processor runs now. Like the OS would need to guarantee that nothing outside some narrow well-defined qemu VM runs for some time an the VM takes care of the rest.
This is bullshit and basically this kind of justification is part of the moral and ethical rot of most institutions in the US now. You do have a choice, you just want to pretend you don't to get away with it. Besides, no one outside of a few stuffy finance/quant shops ever even asked what my GPA was in college, they don't care.
There is no such thing as "tech circles", tech is now a huge huge industry with 10M people in the US alone and no uniform political beliefs. There are the beliefs of some of the vocal billionaires and VCs which media likes to equate to "tech", but they are a tiny fraction of tech. Presumably he said it because he's a narcissist moron, I don't know.
If I had to describe it, the zeitgeist of VC/Founder tech now is kind of originated in "It's Time to Build" [1]. The diagnosis is that the nation is incapable of doing anything - too hamstrung by bureaucracy, laws, congress gridlock etc. The same kind of things that MAGA uses to justify authoritarianism. The movement now basically prescribes action, agency, urgency, conquest void of any reflection of what the goals should be, any question of morality or values etc.. just build (what you're being told/funded to build).
Yes lol same! I was using Sunshine/Moonlight before, it does work great but then I got a 4K 120Hz HDR TV that only has 100Mbps ethernet, a weak video decoder that can't really handle AV1 at decent FPS, weak network stack that can only hit like 80Mbps, and a "Game Mode" that only triggers with external cables.. unfortunate. I also looked into Steam Decks and other docks but none of them actually support the full bandwidth either network or video that it would take to get those specs either. Even with a direct ethernet connection between the TV and PC and lower quality, it still has a noticeable input lag.
So I ended up getting an long HDMI 2.1 fibre optic cable and a long USB-extender for the controller dongle.. works great. It just sidesteps a whoole lot of annoying problems and limitations all over the stack.
He was bragging about working on weekends and comparing his shitty little insurance company to the Manhattan Project a while back. Somewhere he claimed this company/industry is the most important application of AI in the world. I have no doubt they ripped it off, this guy is not trustworthy to say the least..