denismi

Those "price per task" benchmarks seem to pre-suppose the existence well-defined requirements and acceptance criteria.

When that isn't available, and the task involves "understanding the problem", "deciding what I'm actually after" etc. iterations, I find price-per-token to still be a better predictor of cost.

No phone, no TV, only a PC that they sit at logged in to Twitter all day refreshing their feed in case a local emergency comes up?

What amount of bitcoin buys "generate highly accurate empirical measurements for the transmissibility rates of this novel virus across realistic human populations"?

How about mortality rates?

The loop required to make language models useful through iterative test and evaluation needs to be able to both operate and escape premature detection, which seems... implausible?

I have not used an adblocker for many years.

Ad servers seem to generally just nope out of serving me ads on their own when I have Firefox's tracking protection cranked up.

Between that and Kagi down-ranking shitty ad-infested domains, it just ... isn't really a problem for me.

> This is like saying my Intel Pentium 3 matches Apple M5 in perf/watt

... if it was within orders of magnitude of being true, perhaps.

On OpenRouter it seemed to only affect a couple of specific providers. Ignoring them has made it s non issue for me.

Didn't Neo4J pivot away from being a boring embedded DB which you point at a path and then traverse through Node objects, and decide to become some kind of paid platform with a client-server protocol and proprietary query DSL?

I remembered it from a uni course (early 10s?) a few years ago for a use-case we didn't end up pursuing, but I wasn't hugely comfortable with investing effort into what I saw.

> Ask Fable to conduct an adversarial /ultareview of their PR and send the same wall of text back to them.

Not necessary. Use Haiku.

The response doesn't need to be good, it just needs to be substantial. Presumably the goal here is basically DoS of the problematic colleague through token limits.

Coming from Australia, which has a more sensible electoral system, I instinctively think about "preference flows" of the minority candidates when I look at these numbers.

Trump was 316,571 short of majority, Harris was 2,601,538.

Neither party gained an absolute majority because 2.9 million out of 155.2 million voters picked third-party candidates: 862,049 Green, 756,393 RFKJr, 650,126 Libertarian, 171,786 PSL, and 477,755 "other".

One of the many failings of the US electoral system is that it effectively ignores these 2,918,109 million voters. But, if forced to state a preference between the two majority candidates, how do you genuinely think they would've gone?

I don't think there's any question that, even just among the 756,393 people who voted for a guy actively campaigning for Trump, there are 316,571 people who would prefer a Trump presidency over a Harris presidency.

That in itself - if 100% of Greens and 100% of Libertarians, and 100% of PSL, and 100% of "other", and the other 58.14% of RFKJr voters ALL prefer Harris - is enough to grant an absolute majority to Trump.

If you plug more realistic guesstimates into preference distributions for the minor candidates, I don't think there is a plausible run-off result below 50.5-49.5 Trump out of those numbers.

> He didn’t win a majority of the votes

Donald Trump plus candidates who withdrew from the race and endorsed Donald Trump DID get a majority of votes though.