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?

No you can't, I just tried it.

I deleted all of my history and "paused" history tracking. Then I reloaded https://www.youtube.com.

My expectation: A feed of the pages I'm subscribed to.

My experience: A blank page which just says "Your watch history is off / You can change your setting at any time to get the latest videos tailored to you. [Learn more] / [Update Settings]".

Hilariously, it is exactly the same - in glorious 4K - on the YouTube app on my TV.

This is deliberately fucked in order force users into the algorithm.

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.

The article you're quoting [1] concerns itself with the creation of systemic "encryption-breaking" capabilities and exploits which said law bends over backwards to expressly prohibit [2].

[1] https://fee.org/articles/australia-s-unprecedented-encryptio...

[2] https://classic.austlii.edu.au/au/legis/cth/consol_act/ta199...

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.

As someone who runs Firefox on both Linux and Android, with Enhanced Tracking Protection enabled, and tries to use web over native mobile apps wherever possible ... I really don't feel this at all?

The first unwatched video from the user's followed/subscribed channels. Chronological, reverse chronological, sorted alphabetically, by the user's channel prioritisation, by likes, by views... whatever the user chooses. And then an end of feed.

For new users? A search bar and a set of (human? AI?) curated seed recommendations that the platform is comfortable with being held liable for.

Hmm..

  pacman -Ss ollama | wc -l                                                                                                              
  16
  pacman -Ss llama.cpp | wc -l
  0
  pacman -Ss lmstudio | wc -l
  0

Maybe some day.

I am quite confident that the following was NOT LLM:

> New users were signing up but not doing anything, they weren’t creating an org, a project, or a deployment, they just left an account sitting there.

Surely the LLM version is:

> New users were signing up but not doing anything; they weren't creating an org, a project, or a deployment—they just left an account sitting there.

> the only artifact left is the diff

You also have code comments, docs in the repo, the commit message, the description and comments on the PR, the description and comments on your Issue tracker.

Providing context for a change is a solved problem, and there is relatively mature MCP for all common tooling.

"The Brutal Choice"

Is there an established name for this LLMism?

I don't need a "Reality Check" or a "Hard Truth". The thought can be concluded without this performative honesty nonsense or the emotive hyperbole.

This probably grates me more than any other.

> Not paying musicians anything at all if they don't have enough streams

... 1000 plays in a year?

We're taking a handful of people (Close friends? A proud mother? The artist themselves?) listening a few times a week.

If an artist has no following, and creates music that listeners consider substitutable for AI slop or low-effort shovelware, then they are hobbyists with no reasonable right to renumeration?

Outside of work, I'm a very sporadic coder. On some side-projects where I'm using Actions, I'll have an inspired few days of progress followed by completely idle weeks/months/quarters.

Losing free Actions doesn't particularly bother me, and I have no issue with paying what is most likely a negligible amount, but I don't really want to have a credit card on file which could be charged some unbounded amount if somebody gets into my account. I've shut down my personal AWS for similar reasons.

Is there any way of me just loading up a one-time $20? That will probably last well into 2027, and give me the peace of mind that I can just let it run. If my account's compromised, or I misconfigure something that goes wild, I am perfectly happy to write off that amount and have my incredibly-low-stakes toy projects fail to build.

This legislation left it entirely up to the service providers to determine implementation, and so far they don't seem particularly motivated to disrupt my usage by asking me to prove my age.

My suspicion is that fairly simple heuristics of age estimation, combined with social graph inspection, are probably enough to completely disrupt the network effects of "social media" for kids, and achieve the stated objectives well enough that I never have to.

Maybe it turns out that I'm wrong, but why even risk it? If the true policy goal is extending mass-surveillance, why waste so much political capital on such a round-about approach which might yield nothing, or even set back your existing capabilities.

MyID (myid.gov.au) already exists, and could easily have been mandated, or "recommended", or even offered as a means of age verification now. But it wasn't.

Australia is a Five Eyes country, with carte blanche access to data that the incumbent social media companies freely share with all the acronym deep-state authorities.

Could you elaborate further on how preventing a sizeable proportion of its citizens from communicating through these established spy-nets, causing them to disperse out to unpredictable alternatives they might not be able to control, increases mass surveillance?

So far I haven't been KYCd by anything.

Aside from YouTube I don't particularly engage with any of these often, but my Google, Facebook, Discord, Twitter, Bluesky, (current) Reddit, Slack, Telegram accounts all seem to be BAU without new requirements.

If the 80% of us currently holding unambiguously-over-16 accounts are exempt, and it only affects future over-16 users as they're onboarded, then it is a very blunt and very slow form of data harvesting which won't yield useful results until years/decades after all of the relevant decision-makers have moved on, retired and/or died. So this seems unlikely?

> Out of the five options available, only one is European (the one I am using). What I don't like is how I cannot add my own custom endpoint. What if I run Mistral locally (with Ollama, for example) and want to use that?

Set up your preferred self-hosted web interface (OpenWebUI or whatever, I haven't looked into this for a while), point it at ollama, and then configure it in Firefox:

browser.ml.chat.provider = http://localhost:3000/

At home I point this at Kagi Assistant, at work I point it to our internal GenAI platform's chat endpoint.

I got different IDs in regular browsing vs my first incognito window vs my second incognito window.

It is fucking wild that we need to resort to putting pig kidneys into humans to squeeze out a few more months of life, while tens of millions of perfectly good human organs are burned or left to rot in the ground each year.

I recently moved away from pass after a decade or so.

Two main reasons:

1. This laptop up was set up with flatpak versions of all GUI applications, including Firefox, and the browser plugin just doesn't work. I persisted with the work-around of `pass -c <path>` from the run command prompt for a while to paste into the browser, but its not ideal.

2. I realised that the Android app was archived. There's at least one fork, but who knows how that will be maintained going forward. https://github.com/android-password-store/Android-Password-S...

For now I'm content with hosting vaultwarden and using various Bitwarden clients.

This pre-election BBC summary - https://www.bbc.com/news/articles/cwy343z53l1o - pretty clearly spells out what has eventuated, describing it as a "central campaign pledge":

> Trump has made tariffs a central campaign pledge in order to protect US industry. He has proposed new 10-20% tariffs on most imported foreign goods, and much higher ones on those from China.

The em-dash everywhere. The constant bullet-lists with bolded titles. The emoji-annotated asides for emphasis.

Convenient timing!

I've moved my credentials over from pass to Vaultwarden about a month ago (after discovering the pass Android app was abandoned and pulled off app stores), and spent the last two weeks since discovering Pocket ID migrating a few self-hosted services to OIDC.