At the point where I choose an expensive model like Kimi 3.1, speed doesn't matter, at all.
Usually, it's because all else failed, and I have to resort to burning Tokens with more expensive/better models, so a few minutes here and there really should not matter.
But his point is that the faster models could _not_ complete the task satisfactorily.
I've seen this myself too: DeepSeek 4.1 Flash is really good and incredibly cheap at constrained implementations, searches, etc. However, frequently the implementations don't take into account all possible cases they should cover. Then I'm either forced to step in and steer it very precisely, or have a heavier model review and provide detailed feedback. I've had good success with this mixed workflow using GPT 6.1 Sol as reviewer. Then my work focuses mostly at the beginning (architecture, goals, constraints), and at the end (final review or very quick intermediate reviews). I'm not aware of benchmarks that simulate this kind of mixed workflow with different models.
Obviously it was not a global "everyone" like today with WhatsApp. But I would guess that perhaps 20 out of 25 in my high school class was on ICQ back in 1999 or so. Most of us only had dial up, so we wouldn't be allowed to be online and block the phone for more than a 30 minutes or an hour per day.
Anyway, ICQ was not some elite hacker thing. Anyone with enough reading comprehension to tell milk and orange juice apart from reading a label could operate it just fine.
This is type mismatch + incorrect assumptions. In 2001 ICQ went over 100 mln registered UINs. Which is not the same as "active users", it's all registrations from the inception (minus 10000 reserved). Losing UINs was a very common thing. Password recovery required email, which wasn't obligatory during registration, even if entered wasn't verified, back then few people knew what "password manager" is, and finally emails weren't particularly stable identity before gmail. But even if we would, for the sake of argument, assume the number as active users count, regular internet users by that time were numbered at more than half a billion. So even with absolutely unrealistically high estimation, ICQ share would be less than 20% of then internet population. And more realistic is probably 10-15%.
I never had ICQ because I was opposed to replacing open internet protocols with proprietary walled gardens governed by unaccountable private corporations imposing clickwrap contracts of adhesion on helpless users.
AOL instant messenger was bigger in USA, it was so big that AOL remained a major chat service even into the broadband years (though AIM clearly was under decline by that point)
My 90s and early 00s social circle was AIM, Yahoo Games, and Neopets.
Pidgin also supported XMPP: it was a multi-protocol client. It was just a client, not a messaging system of its own. They'd abstracted the send/receive from the actual messages and built a bunch of backends that could handle pretty much any messaging protocol at the time.
ICQ was released 1996; Pidgin (as GAIM) was November 1998; Jabber was released in 1999. Those were good times.
Funny, that's always the impression I had of ICQ back in the day...
Surely we can agree that no one's experience is universal, and both regional and subcultural differences are likely to have materially shaped our views of the various chat services in the '90s and early '00s...?
Surely we all recognize that our own experience is not The Default Everyone Should Understand?
do you mean Trillian, the software that offered a (very themable) frontend to AIM, ICQ, yahoo and (i think) at least 3 more chat services?
the awesomeness of Trillian made it obvious how useless it is for so many different services to coexist while they all offer the same value. still true today
yes, I used AIM from 1999 until whenever it was shutdown (2017?)
At several companies, we used AIM for what many use Slack for today. We even coordinated prod deployments in a private AIM chat room!
Poor Elon put in so much effort into his Nazi Salute, he'd be pretty upset about you downplaying his best attempt, if he could read what you wrote here about him.
I don't know which bubble you're in, but since Elon took over, Twitter became absolutely unusable.
The decline happened with bots flooding the platform, with the "creator reward" program rewarding meaningless engagement bait, and with people who I respect in the Tech bubble leaving the platform, which led to an even faster decline.
That was just not the case before the takeover; back then, you would enjoy scrolling through it, engage in authentic conversations, and learn a thing or two.
The Greens were not in Office when the nuclear exit was decided 2011.
The Greens actually prolonged their life time during Covid, after RWE, EON and EnBW argued that it does not make sense economically to let them run longer.
No insurer wanted to insure nuclear, and no energy company wanted to keep losing money on nuclear.
Since nuclear is the most expensive form of energy generation, it makes sense to switch it off first.
> No insurer wanted to insure nuclear, and no energy company wanted to keep losing money on nuclear.
The German government(s) helped kickstart the global solar price drop by their subsidy programs in the early 2000s that allowed for economies of scale, so I'm surprised that—if they're really worried about climate change—they would not subsidize the alleged expenses of running the low-carbon generation of nuclear.
> Since nuclear is the most expensive form of energy generation, it makes sense to switch it off first.
It's so expensive that prices in France, which generates >60% of power with nuclear, are at the same levels (or lower) as Germany, while producing less carbon per kWh than Germany (a quarter from some articles I've found).
See perhaps:
> […] Alternatively, Germany could have kept the existing nuclear power in 2002 and possibly invest in new nuclear capacity. The analysis of these two alternatives shows that Germany could have reached its climate gas emission target by achieving a 73% cut in emissions on top of the achievements in 2022 and simultaneously cut the spending in half compared to Energiewende. Thus, Germany should have adopted an energy policy based on keeping and expanding nuclear power.
Politically that would have been done how? I think in restropect removing nuclear powerplants that make awful targets in a conflict doesn't strike me as a bad choice either.
> Politically that would have been done how? I think in restropect removing nuclear powerplants that make awful targets in a conflict doesn't strike me as a bad choice either.
The only people causing conflict is Russia, and Germany had to be dragged kicking and screaming in shutting down Russian supply pipelines, so don't go talking about "security" when the actions from Berlin showed there was zero consideration for it:
Actually no, or unlikely, since it’s a massive escalation. Even in Ukraine Russia does not strike the nuclear plants. They strike all the other power plants though.
The chief of the International Atomic Energy Agency on Saturday expressed "serious concern" after Moscow accused Ukrainian forces of launching a drone attack on the Russian-occupied Zaporizhzhia Nuclear Power Plant in southern Ukraine
> It's so expensive that prices in France, which generates >60% of power with nuclear, are at the same levels (or lower) as Germany, while producing less carbon per kWh than Germany (a quarter from some articles I've found).
Maybe you should take a closer look how those prices come to be.
Every single nuclear watt is being subsidised and has been for decades.
Last time they wanted to play the free market game, they had to bring the company back into the forgiving hands of the French taxpayer.
I have rather Germany invest into future tech then waste it on rotting reactors which are obviously unfit for the current state of global warming already.
Nuclear was clogging up the grid with it's expensive energy. It is gone now and the grid is diversifying while it follows the path of complete decarbonisation.
There is absolutely no reason to look back for Germany and only those whose ideology is "looking back" seriously do that currently.
> Every single nuclear watt is being subsidised and has been for decades.
This is just plain false. In the 1990s and early 2000s, EDF was generating a whole lot of profit, which it wanted to reinvest. The governments of the time (both left and right) said: "Nope! We’ll take that, thank you." As a result EDF stopped making new nuclear plants and, most importantly, new nuclear engineers and technicians. When the government suddenly decided to have EDF build a new plant 15 years later, the 1980s generation was all but gone, the industrial know-how had vanished, and thi construction costs very logically skyrocketed. Oh, and in the meantime EDF had been (and still is) forced to sell electricity to its competitors under production cost.
The difference is that Germany's investments in renewables at a time when they were still expensive helped to create an economy of scale, which brought down prices for everyone and helped trigger a huge amount of investments worldwide. France build an unsustainable nuclear industry that was forced to sell electricity below cost and had to be renationalized to survive.
I agree with everything you say, but the initial "Atomausstieg" was decided in 2000/put into law in 2002, when the Greens were in office. The exit was delayed in 2010 (and then accelerated again in 2011) by the CDU/CSU/FDP, but the initial decision for the exit was very with involvement of the Green party
Can coal plants buy insurance against global warming? Can a hydroelectric dam buy insurance against downstream flooding?
Very few things in the power grid has insurance against major accidents. It is even doubtful for example if a forest fires caused by transmission lines will be repaid in full by the insurance of the power company, and to the degree that will pay all costs to property and land owners
Usually, it's because all else failed, and I have to resort to burning Tokens with more expensive/better models, so a few minutes here and there really should not matter.
reply