jakderrida

If a video is less than 45 minutes, you can plug it into any gemini and have it transcribe and diarize the video. (although, I've been told it only works on aistudio anymore) Sometimes it trips up. However, plugging in videos longer than 45 minutes quickly leads to ludicrous behavior the longer it is over 45 minutes.

You can even just plug in the Youtube link with a short prompt.

Bear in mind that thinking that plugging in the transcripts will help the Gemini model in any way frequently results in mistranscriptions being included and actually compromises the diarizations.

The prompt I use is...

"Please transcribe, diarize, and speaker identify the speakers from the Youtube video."

Another thing I've learned is that long-worded prompts also interfere with the results in weird ways.

Is it really an ad hom if he doesn't know the hom?

Their reply is 100% based on the content of your post.

I looked into this because that depletion.org site makes the situation look scary as hell at first glance, and I think the honest answer is basically: *yes, this is a legitimate problem, but no, I wouldn't read it as "gas is about to be rationed and stay insanely expensive for years."*

First, the site isn't bullshit. The underlying inventory problem it's tracking is very real. The IEA's September 11 report says global observed oil inventories have fallen *507 million barrels since February*, more than *10 million barrels/day of Gulf production was still shut in* during August, and global refinery throughput was 4.2 million b/d lower than a year earlier. That's pretty freaking ugly.

However, there is a pretty important distinction between the site's data and the scary probability numbers it puts on the scenarios. The site itself says the probabilities are assigned using judgment, and it specifically warns that the model "wasn't developed by an energy analyst." So I'd regard it as a really useful stress dashboard and not interpret "50% corridor lapse" as if the IEA just announced a 50% probability of catastrophe.

What I found more concerning is that some actual energy experts are now describing basically the same physical problem, just without going nearly as far on the probabilities.

Columbia's Center on Global Energy Policy put out a discussion today estimating the world is currently short roughly *5 million barrels/day* of crude and petroleum products relative to demand. Their point was that the reason this didn't immediately turn into an enormous price spike months ago is because we had buffers everywhere - excess oil, oil sitting on tankers, strategic reserve releases, spare refinery capacity in some places, etc. We're now burning through those buffers. At some point price has to do the work.

And I think "price has to do the work" is the key distinction here.

When economists talk about *rationing through price*, they don't mean somebody is handing you a little card allowing you eight gallons of gas this week. They mean gasoline goes to $5, $6, whatever it takes until enough people decide not to take the road trip, companies consolidate deliveries, airlines cut marginal routes, factories use less diesel, weaker economies consume less, etc.

That's much more likely than literal nationwide American gasoline rationing.

The other thing that surprised me is that *diesel and refined products actually look scarier than gasoline*. This isn't just a shortage of crude. Gulf refining capacity is also disrupted, Russian products are constrained, and refinery margins have gone nuts. So you can theoretically have crude available somewhere on Earth and still have a shortage of the exact petroleum product somebody needs in the exact place they need it. The IEA specifically says the global refining system is stretched extremely tight.

That said, the EIA is still nowhere close to forecasting "welcome to Mad Max."

Their September 9 forecast has Brent averaging *$74/barrel in 2027* and U.S. regular gasoline averaging *$3.35/gallon*, with Middle Eastern production gradually recovering and getting back near pre-conflict levels around Q2 2027.

There is a catch there, though. Their forecast was actually completed September 3, so some of the latest deterioration isn't in it. That's probably why I wouldn't just shrug and say the whole thing is temporary either. The newer IEA report is substantially uglier. Still, even the IEA forecasts an enormous *8 million b/d rebound in global production in 2027*.

So if you're literally asking:

> should I trade my car in for an EV because I might not be able to buy gasoline?

I wouldn't.

If you were already going to replace the car anyway, though, I think this absolutely moves the needle towards an EV, especially if you can charge at home.

You're basically buying yourself some insurance against this entire category of bullshit. Strait gets closed? Iran attacks tankers? Saudi pipeline gets blown up? Russia stops exporting diesel? Oil hits $150? You care considerably less.

Wood Mackenzie actually published something on September 11 making essentially that broader argument - that persistent oil-price volatility could accelerate EV adoption because the advantage isn't merely a lower average fuel cost. You're also removing most of your exposure to oil-market shocks.

But if you've already got a perfectly good paid-off gasoline car, dumping it and buying a brand-new $40,000 EV solely because you're scared gasoline will be rationed seems like exactly the kind of panic trade where you somehow manage to lose more money avoiding the crisis than the crisis would have cost you.

Basically, my read is:

*Expensive and unusually volatile gasoline for a while? Very plausible.*

*Diesel/refined-product shortages getting seriously nasty? More plausible than I realized.*

*Localized shortages if things get worse? Definitely possible.*

*America running gasoline ration books for years? I couldn't find any serious institution treating that as the expected outcome.*

The part of this I'd actually keep watching isn't even the price of oil by itself. It's whether Hormuz tanker traffic recovers, whether the Saudi bypass pipeline comes back, whether Gulf production starts returning, and most importantly whether inventories finally stop falling.

If those things start improving, this probably looks like a brutal but temporary energy shock.

If another few months go by and we're still draining hundreds of millions of barrels out of inventories while the physical routes remain screwed up, then I think the depletion.org people start looking considerably less alarmist.

"couch all their agreements with caveats and provisos."

When you're a ChatGPT Projects or Claude Projects user, those caveats and provisos are your worst enemy because they'll change caveats into hard rules (either for the session or committed to memories) and you end up in absolute hell having to make it investigate to figure out why it can no longer produce anything but read-only pre-check code that never actually does anything but keeps performing stupid safety checks.

I found that it's shockingly good with R. (the only language I know and can correct for)

I doubt they even intended it to be, but it seems like I kept going from resorting to 3.5-3.8 (over time) to realizing that Claude and GPT, while great at Python, will make rudimentary mistakes with R; even when they compose giant complicated R code.

I think he's just saying the case they would make to avoid being sued for breaching their fiduciary responsibility. Not that it's the actual reason. But Idk.

Think of it more like the original poll. They originated the census weighting methodology. In statistics classes about methodologies and survey design, it ALWAYS starts with a reference to the Literary Digest poll being wrong about Alf Landon beating FDR and Gallup, a new poll, being dead right with only like a sample size of over 1000 while Literary Digest sampled all their readers.

I don't recall them doing polls for commission like almost every other poll does.

They cut down the sample size about 5-7 years ago, anyway, by like 90%. I learned this not through a press release, but by going through their metadata. This was a long time coming. Their business model just isn't very profitable. I wish, instead, they just sold it off to another company to continue the same methodology and maintain the prior data. This frequently happens. There's value in having a poll that has been running for at least 5 years. Much value in a poll that has been running for like 100 years.

I'd argue that Anthropic still has a hard edge on creativity for things like emulating people's comments.

I've fed into several models my past reddit comments (with the comments it's responding to) and asked it to duplicate the style. Claude has always been the only thing that comes even close to original responses that even I think would be exactly my response, wording and all.

GPT or Gemini will just borrow snippets from the example text and just smoosh it together to make semi-coherent points. Scratch that. They're coherent, but they're just unmistakably not from me.

What if I just modify the code to misspell things that no AI would misspell?

If I'm being honest... I expect the websites to keep returning errors and have hopes that those that employ you to at least start to understand what's going on.

I think more often you'll find it's the mediocre coders (like myself) that have trouble using AI. The software developers and CS majors just know exactly what to tell it to do and in the *exact* language it could best be understood. That's just my experience.

Also, I get caught up in multiple errors that will never go away and, since I'm stepping out of my wheelhouse with libraries or packages I'm completely unfamiliar with, I'm completely helpless but to diagnose what went wrong myself and improve upon my code prompting skills.

Don't get me wrong. AI makes possible many things for me. However, I think professional coders probably accomplish much more.

It'd be nice if they underwent tests and developed fine-tuning data that was proven to work rather than using fine-tuning data made to cater to human preferences.

When I would look at Copilot's searches, it was as juvenile as what my grandma would type right after I taught her what a search engine was.

I feel like it should instead link to a video in the sling.com domain with almost patronizing instructions on how to convert their login to Sling and promote it as "Your very own Blockbuster at home".