aubanel

"well preserved tools" said the ad -> I bought some, surprisingly expensive for a hammer -> it's a mishap and inform piece of wood -> straight to dump

Yet the carbone intensity of energy production in Germany is among the worst in Europe.

And France (nuclear powered, no particular huge investment in a green transition) beats them easily in both price and carbon.

https://app.electricitymaps.com/map/live/fifteen_minutes

The success of the Energiewende ? Just take a look at average electricity price and carbon footprint in Germany Vs France.

https://app.electricitymaps.com/map/live/fifteen_minutes

I'd advise you take a look at some of Musk's companies: - Tesla is the top seller of EVs in the US, beating century-old companies. - SpaceX has left public institutions like NASA and ESA in the dust despite their vastly bigger budgets - Although it joined late, xAI is now firmly in the top 4 of AI companies worldwide (OpenAI, Anthropic, Google, xAI)

What's the common element between these successes?

Ukraine is not and was never part of EU, FWIW

Other sources indicate that privatization was a huge success!

The number of passengers dropped after nationalizations, and exploded after privatisation (despite taxpayer money investment logically dropping)

https://commons.wikimedia.org/wiki/File:GBR_rail_passengers_...

For Europe, use ECMWF, they provide great data: https://www.ecmwf.int/en/forecasts

> If I were to steelman the Sutton perspective, it would be...

I don't find it very courteous to say that you're steelmanning someone's argument. Sutton is certainly smart enough to have steelmanned his argument himself. Steelmanning : do it in your head, don't say it!

Great article! It seems like the whole industry has overfitted to the System designs interviews of FAANG, thus focused on extreme scaling need that few companies actually have.

I love that when I opened this article i already knew some elements, from having read it months ago on HN

So now I will remember it a bit better and for longer

Hackernews is actually like Anki cards for nerd (and in this case useless) Internet stuff

> hard to disrupt with AI for two reasons: (1) mistakes, hallucinations, etc., aren't acceptable

I think AI hallucination rate is already below my own hallucination rate, especially in a boring/unknown domain

> "We are hearing from a lot of rightsholders who are very excited for this new kind of "interactive fan fiction" and think this new kind of engagement will accrue a lot of value to them, but want the ability to specify how their characters can be used (including not at all)"

Marvelous ability to convolute the simple message "rightholders told us to fuck off"

"And I tell you, you may only avoid an effort in the name of a greater one, for you must grow"

Antoine de Saint Exupéry (author of The Little Prince), in Citadelle

Thank you, interesting! I don't know that much about Fair use: if I understand well, the key is that the use should be "transformative", right? Am I correct in understanding that: - if the original "This is fine" meme was under copyright, the dog picture would be exempted from copyright by Fair use as it's a transformation - here it's not even needed since the original is not under copyright ("this is not a derived work")

That, and the dragon looking straight out of How to Train Your Dragon - I wonder if they have agreements with the right holders, or if they expect massive lawsuits to create free advertising for their launch.

Giscus is a cool comments management solution vis Github: It lets users auth through Github then posts comments as discussions under your repo. We use it here for instance: https://predibench.com/models

That title looked like clickbait until... Oh well they did actually did it

The answer to this is just "intermittent" : the higher the share of renewables, the higher the share that you have to make up for when you're at night without wind. This can be done through batteries, water storage, or interconnection, but that's the real cost of renewables.

The premise "LLMs have reached a plateau" is false IMO.

Here are the metrics by which the author defines this plateau: "limited by their inability to maintain coherent context across sessions, their lack of persistent memory, and their stochastic nature that makes them unreliable for complex multi-step reasoning."

If you try to benchmark any proxy of the points above, for instance "can models solve problems that require multi steps in agentic mode" (PlanBench, BrowseComp, I've even built custom benchmarks), the progress between models is very clear, and shows no sign of slowing down.

And this does convert to real-world tasks : yesterday, I had GPT-5 build me complex react charts in one-shot, whereas previous models needed more constant supervision.

I think we're moving goalposts too fast for LLMs, that's what can lead us to believe they've plateaued : but just try using past models for your current tasks (you can use use open models to be sure they were not updated) and see them struggle.

From experience, it's much more engineering work on the integrator's side than on OpenAI's. Basically they provide you their new model in advance, but they don't know the specifics of your system, so it's normal that you do most of the work. Thus I'm particularly impressed by Cerebras: they only have a few models supported for their extreme perf inference, it must have been huge bespoke work to integrate.

- AI competing is "wholly unfair"

- "[AI is] far away from being substantially being better than MCTs"

^ pick only one

One of the key points of the article is "I feel FPV drones to be mostly a failure because their success rate is low" Why is that a failure? If one 500$ drone has even only 10% success rate, if the target is a 1M$ equipment it's still a win!

I've read Bertrand Russell's "A History of Western Philosophy" and it's the first ever philosophy book that I didn't drop after 10 pages, because of 2 things: 1- He's logic (or at least has the same STEM kind of logic that we use), so he builds his reasoning logically and not via bullshit associations like plays on words or contrived jumps. 2- He's not afraid to tell "this philosopher said that, it was an error", which is extremely new compared to other scholars who don't feel authorized to criticise even obvious errors. Really recommend!

No it's not for a 17.1428x engineer like me!

Don't be too harsh, he added "we're telecom" somewhere

Certainly not, it's actually possible to add 3 float32 numbers with 90% precision using AI! With a recent breakthrough, the team is working on pushing that to 10, we have enough cracked engineers to hope to make it happen soon!

There's already a lot of info in there: model architecture and mechanics.

Using the model to generate synthetic data also allows to distil its reasoning power into other models that you train, which is very powerful.

On top of these, Qwen's technical reports follow model releases by some time, they're generally very information rich. For instance, check this report for Qwen Omni, it's really good: https://huggingface.co/papers/2503.20215

-> Power outage -> Wild guess that Russia has been doing it, 0 proof or even hint -> "This is a war of agression, we should respond"

Are you suggesting to attack Russia, based on absolute thin air?

If only one thing it's his series of books Jesus of Nazareth: the only theology book I've ever read, it clarified so much about the Scriptures, and was written from the point of view of a world-class scholar, much more scientific than I would have thought, taking into account the views of other scholars, Jew, Protestant or atheists alike, as long as they had interesting ideas.