Alejandro Rivero (@arivero) header
Alejandro Rivero (@arivero) profile picture

Linux User #40306. Única persona que sigue usando &c. para abreviar etcetera.

Zaragoza, España
Joined November 2007
Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)Photo by Alejandro Rivero (@arivero)
Alejandro Rivero retweeted
i hate sf halloween, what do you mean you’re a human in the loop
40
66
7
2,427
50,952
Alejandro Rivero retweeted
Lo último que confiaría a los agentes es hablar con mis clientes. En una empresa de servicios el principal propósito de los agentes debería ser liberar tiempo para hablar con los clientes.
8
14
6
90
3,641
Alejandro Rivero retweeted
Interesting. Revolut reportedly only started seeing meaningful AI productivity gains after moving away from expensive OpenAI and Anthropic models to self-hosted GLM. Decagon reportedly runs ~90% of its workloads on open weight models. Why? Because when you're running AI agents in production for e.g. customer service, cost and latency matter enormously. You don't need a frontier model with PhD-level reasoning or even a model that knows the capital of Finland to handle every customer interaction. We're seeing similar dynamics across hypergrowth AI startups and increasingly the large enterprises we work with. But most importantly, the future isn't binary. When use cases are new, you want the smartest general-purpose model available. But once workflows mature and become predictable, smaller, faster, specialized models increasingly win. I wouldn't be surprised if 80–90% of routine enterprise AI workflows eventually run on open-weight and smaller models. This doesn't mean frontier labs lose. It means the market is splitting into two massive opportunities: frontier intelligence and specialized execution at scale.
50
74
27
801
77,402
Alejandro Rivero retweeted
Replying to @mdp_sec
it's just an effort change to xhigh for one turn. any subagents spawned in that turn will also run at xhigh.
4
2
1
59
5,373
Alejandro Rivero retweeted
The smarter a model is, the more subagents it spawns I noticed this before but the correlation is stronger than I expected
12
3
3
124
10,979
Alejandro Rivero retweeted
Customer: "Do you do bones? I've been told to make broth for the collagen." Butcher: "Who told you?" Customer: "The woman in the health shop. She's just sold me a tub of the powder as well. Twenty-eight quid." Butcher: "What's it say on the tub?" Customer: "Bovine collagen. Hydrolysed." Butcher: "What's bovine?" Customer: "...Cow." Butcher: "Which bit of the cow?" Customer: "It doesn't say." Butcher: "The hide. They take the skin off him, boil it down, dry it to a powder and sell it back to you by the gram. That's the whole industry. It's him, in a tub, at about forty times what I'd get for the skin." Customer: "So it does work?" Butcher: "It's gelatine. Same as a trifle. Your grandmother got it off the Sunday joint and called it gravy that set." Customer: "What should I buy, then?" Butcher: "Shin. Oxtail. A trotter if you're brave. The bits with the gristle in, that's the collagen you've just paid £28 for. Three hours, lid on, and the pan sets solid in the fridge overnight." Customer: "Solid?" Butcher: "Like a jelly. That's the test. If the broth doesn't set, there was no collagen in it. The pouches in the supermarket don't set. Half of them have gelatine added so they look like they did." Customer: "And the tub?" Butcher: "The tub's a trotter that went to university. Keep it if you like. I'll do you a kilo of shin for a fiver and you can have the bones for nothing, same as my dad gave them to the dog."
54
390
8
3,043
177,867
Alejandro Rivero retweeted
À titre de comparaison, Mistral a entraîné Mistral Large 4 sur 3800 Blackwell on manque CRUELLEMENT de compute
Replying to @XFreeze
There will be a total of around 1.2M Blackwells at Colossus 2 by end of year
10
15
5
332
11,982
Alejandro Rivero retweeted
Daughter of a uranium miner, sets off Geiger counters with just her hand and she’s absolutely beautiful. The perfect woman. ☢️💗
A Riverton woman who spent a summer in Wyoming's Gas Hills playing in the raw yellow uranium with her best friend and a poodle still sets off Geiger counters. "At least I don't glow in the dark," she said about other side-effects decades later. cowboystatedaily.com/2026/11…
15
36
5
623
13,493
la capoladora – at Saragossa, Spain
one of the very few things I remember well from my childhood in Moscow is having spent many an hour disassembling then reassembling the manual meat grinder by the kitchen window my signature threat, when I was very angry, was to "grind everyone up in the meatgrinder"
1
2
94
Alejandro Rivero retweeted
This post is a reverse CAPTCHA. Look at the replies, AIs love talking about the Oxyrhynchus Papyri, and I would be willing to guess that only a very small proportion of humans have, and most of them will spell it wrong.
A bunch of funded projects to unblock some long-standing backlogs in translations using AI would be interesting; especially when AI models can do a reasonable first pass at interpretation & significance, not just translation. There are major repositories of untranslated ancient work in many places, like the Oxyrhynchus Papyri oxyrhynchus.web.ox.ac.uk/hom…
79
10
6
319
31,214
—Mi hermano y yo hemos heredado un piso. Me lo quedo y le pago su mitad. —Si lo formalizáis como compraventa de su mitad, pagas Transmisiones Patrimoniales. Si extinguís el condominio y le compensas en dinero, pagas Actos Jurídicos Documentados.
2
7
67
21,117
Alejandro Rivero retweeted
tranquilos soy yo, que desplazo la media
Switzerland is the one country that combines both very high AI adoption and high average spend:
3
2
93
4,251
los modelos que han aprendido a hacer matemáticas a nivel de investigación son tambien los que han tenido que retrasar por problemas de alignment. Y esta es mi charla ted
1
1
229
esto mismo lo tenemos claro, lo que es util es que el csic por ejemplo renga un servicio de inferencia local, que va a estar corriendo 24/7. – at Saragossa, Spain
No dijo eso en ningún momento pero bueno Sam Altman no es santo de mi devoción pero tampoco hace falta manipular, eh? Dijo que LA INFERENCIA LOCAL es un desperdicio, INFERENCIA!! En ningún momento habló de PCs gaming ni de ordenadores de oficina, INFERENCIA, tener IA en tu casa
1
3
171
Alejandro Rivero retweeted
My video on the works of Ashoke Sen becomes my second video to reach 100 k views. Thanks for watching. #physics #scicomm
1
21
644
Alejandro Rivero retweeted
Let me get this straight. A hacker says he found Claude Fable 5's weights sitting in a misconfigured AWS bucket. He grabs one 7GB shard. Says the full model is 640 shards, 8TB. Puts it up for sale. People check his proof file. It's byte-for-byte identical to a DeepSeek model you can download for free. Imagine getting scammed with a file that was on Hugging Face the whole time 😭
🚨🇺🇸 Anthropic model weights allegedly exposed through misconfigured AWS S3 bucket An actor using the handle "localizer" claims to have obtained a 7 GB shard of Anthropic's alleged "Claude Fable 5" model weights from a misconfigured AWS S3 bucket associated with Amazon Bedrock. The actor claims the complete model consists of 640 shards totaling approximately 8 TB, but says only one shard was retrieved before access was restricted. Screenshots showing a .safetensors file, model metadata, and a hexadecimal dump have been published as alleged proof. The actor is now offering the material for sale. Claim is currently unverified. The authenticity of the files and their connection to Anthropic have not been established. 💥 No delays. No guessing. No redactions. Get the intel threat actors see, the moment they post it. darkwebinformer.com/pricing
Readers added context they thought people might want to know
The model shown here is not a Claude model, but DeepSeek V4.1 Flash. This can be verified by the fact that the file sizes match those of DS V4.1 1017. The screenshot also includes "eu-north-1," even though it does not host Fable. x.com/torisetxd/stat… x.com/YipingDeng5/st… x.com/olddonkeyblog/…
2
4
1
135
72,702
I think the excuse was to make him interested in Lattice QCD.
Replying to @TheMCUChronicle
Simpler explaination, the writers realized smart hulk sucked and needed an excuse to revert him to savage.
46
cuando comenzo la privatiza6de la internet una cuestión de fondo es si era un fenómeno reversible, dada la suavidad con la que ocurrio – at Saragossa, Spain
Como era de esperar, Alphabet admite que el AI Mode está contribuyendo a servir menos anuncios, su principal fuente de ingresos. Llevamos veinte años financiando internet a base de clics y el modelo de negocio se ha ganado la necesidad de que alguien lo reinvente. Hablaba el otro día al respecto de Instinct: quien crea que se va a financiar con publicidad, se equivoca…
79
Alejandro Rivero retweeted
Para los que de 30 y pico tacos para arriba, recordaréis la película AI de Steven Spielberg, no? :)
1
1
7
415
La invención de problemas (problem posing) es de las actividades más ricas y "económicas" que se pueden hacer en clase de matemáticas. Eso sí, también requiere algo de planificación. Gracias Chus por compartir todo esto.
Y después de eso llegará el momento más interesante: ellos inventarán sus propios patrones y los propondrán al resto de la clase para que los resuelvan. Ya no son solo resolutores. Pasan a ser creadores de problemas.
6
21
1,176