Pedri (@espeqtando) header
Pedri (@espeqtando) profile picture

Information Era enjoyor. Freedom maxi.

Madrid, Spain
Joined December 2011
Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)Photo by Pedri (@espeqtando)
El autodidacta se come al universitario.
116
146
36
1,499
47,418
En este país de pandereta los políticos se ponen las pilas un mes de campaña y cobran un sueldo toda la vida.
14
This week is the proof that we’re entering a cyber crisis. Every day a major exploit. Soon one of these will be nasty and stopping recursive self improvement will gain momentum.
16
With this week’s NVIDIA earnings well also get a leak o what % comes from smuggling into China
37
Lo de La Séptima es de vergüenza ajena. El último bastión de propaganda basura a merced de Sanchez.
43
4 o 5 fascistas? España entera piensa que es un hdp señor @sanchezcastejon
27
CIA sacked Jensen and will now make sure no more last gen GPUs get to china
I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
38
Iran, reinforced, has started to apply pressure on Trump but oil has dropped lately. Expect action the next few days.
99
Ive felt two big leaps in productivity with new released models. First was opus 4.6, second was 5.5
118
Opus 5.5 has been dwarfed by increasing the ratio of guesses vs verifying answers. Probably using much less compute but reliability has dropped dramatically @AnthropicAI @bcherny
84
Abascal aquí tiene un punto que es completamente razonable. @rtve no debería estar pagando semejantes contratos, no debería competir con cadenas privadas con el dinero de todos y debería mantenerse al margen del sesgo político. Broncano sabe que no necesita suerte para que su programa sobreviva al cambio de gobierno. Les sobrarán ofertas.
Amenazar sonriendo no es hacer humor, es degradar nuestra democracia. Para esto luchamos: para que España no vuelva a ser un país donde los ultraderechistas amenacen a cómicos y periodistas. Frente a la involución, movilización.
46
La realidad es que @LaRevuelta_TVE no necesita suerte. Es un programa con una comunidad leal e integrantes con talento. La "suerte" aplicaría a renovar el contrato millonario que tienen con @rtve, cuyos costes debería asumir un televisión privada cómo es lógico.
44
Feeling like a minor release. Ant might be saving a bullet for the last days before going public
Claude Fable 5.2 and Claude Mythos 5.2 are available now:
26
Ya era hora @oscar_puente_ Ahora le toca al perro @sanchezcastejon
Ábalos, Koldo, el fiscal general, Cerdán, los delegados del Gobierno de Ceuta y Madrid, Zapatero, Paco Salazar, Montse Minguez, su mujer y su hermano…. ¡Y ahora Oscar Puente! Sánchez siempre se esconde detrás de sus esbirros. Pero la responsabilidad es suya. Es el capo de la mafia. Es el cerebro de la organización criminal que nos gobierna. Su traición mata. Su corrupción mata. Solo la respuesta masiva de los españoles en las urnas puede sentarle en el banquillo, y recuperar la esperanza y la dignidad nacional. Sánchez es quien debe dimitir para ponerse a disposición judicial.
26
US attacks on election night just as dem sweep is announced. War escalates and the US enters state of war. Trump gains power while loosing elections.
68
¿Habrá robado el PSOE en Adif?
30
Hyperliquid trader up 50m on near:native has accelerated selling and has closed already 35% of longs.
38
If a bull run is really coming soon @Pumpfun will be a top beneficiary. Current setup + retail volume surge will make the price go parabolic.
28
What happened today in Spain looks odd. The trains crashed after the software that prevents this from happening went down for 30 minutes. I wont be surprised if this is the first of a few cyber related attacks on the current government.
48
El choque en Los Yébenes - un fallo informático? En el mes que queda hasta las elecciones preveo varios ‘escándalos informáticos’
59