Last Notes
Servidores do Xbox estão fora do ar nesta segunda-feira (27)
Com os servidores do Xbox fora do ar, alguns jogadores sequer conseguem abrir seus jogos nos consoles. Entenda o que está rolando!
https://flowgames.gg/servidores-do-xbox-estao-fora-ar-segunda-feira-27/
Servidores do Xbox estão fora do ar nesta segunda-feira (27)
Com os servidores do Xbox fora do ar, alguns jogadores sequer conseguem abrir seus jogos nos consoles. Entenda o que está rolando!
https://flowgames.gg/servidores-do-xbox-estao-fora-ar-segunda-feira-27/
Ricardo Couto escolhe a defensora pública Patrícia Cardoso para comandar o Desenvolvimento Social
https://agendadopoder.com.br/wp-content/uploads/2026/07/27-07-2026-PATRICIA-CARDOSO.jpg
Defensora pública e ex-chefe da DPRJ assume a secretaria em meio à reestruturação administrativa que também alcançou as áreas de Desenvolvimento Econômico e Segurança Pública
https://agendadopoder.com.br/ricardo-couto-escolhe-a-defensora-publica-patricia-cardoso-para-comandar-o-desenvolvimento-social/
Bruce W. Smith shared this drawing for “The Proud Family: Louder and Prouder” ahead of the series…
https://64.media.tumblr.com/2f90790993d4b8525e4668a3c051b780/662c382a6e9697c2-53/s640x960/78b850bb12964463953a8fb76ec7c4904fdc60bc.png
Bruce W. Smith shared this drawing for “The Proud Family: Louder and Prouder” ahead of the series final season. 💜
The caption says:
“Will he…?
Always…?
The final season of ”The Proud Family: Louder and Prouder“ premieres this Wednesday on Disney+.
https://www.instagram.com/p/DbRefXlhvJv/?utm_source=ig_embed&utm_campaign=loading
https://disneytvanimation.com/post/823292377693372416
Rare First Temple figurine mold found in Jerusalem confirms local production
A rare clay mold used to make the face of a female figurine has been found in Jerusalem. Archaeologists say the find gives the first direct proof that at least some First Temple period figurines were made inside the ancient city instead of arriving through trade. The artifact came from excavations at the Kishle site […]
https://archaeologymag.com/2026/07/first-temple-figurine-mold-found-in-jerusalem/
Is 16GB RAM enough for iMovie?
https://forums.macrumors.com/threads/is-16gb-ram-enough-for-imovie.2485938/
Lula e Xi Jinping conversam sobre acordos comerciais em meio à tensão com os EUA
Na conversa, Lula teria ressaltado a importância de um acordo Mercosul-China, aumentando o fluxo econômico entre os países
https://jovempan.com.br/mundo/lula-e-xi-jinping-conversam-sobre-acordos-comerciais-em-meio-a-tensao-com-os-eua/
Autofahrer fährt drei Mädchen um und flüchtet – nicht festgenommen
Drei Mädchen im Alter von zehn und elf Jahren waren am Samstagabend mit ihren Fahrrädern in München-Sendling unterwegs. Vor ihnen fuhr die Mutter eines der Kinder. Gegen 18.45 Uhr näherte sich auf der Lindenschmitstraße ein VW Polo, gesteuert von einem 42-jährigen Marokkaner mit Wohnsitz in München. Der Wagen erfasste die drei Kinder. Alle wurden durch
Der Beitrag https://www.tichyseinblick.de/daili-es-sentials/muenchen-autofahrer-faehrt-drei-maedchen-um-und-fluechtet/ erschien zuerst auf https://www.tichyseinblick.de.
https://www.tichyseinblick.de/daili-es-sentials/muenchen-autofahrer-faehrt-drei-maedchen-um-und-fluechtet/
I give Taiwan 7 days to clear the air. No conclusion for now.
https://www.reddit.com/r/China_irl/comments/1v7y0yk/i_give_taiwan_7_days_to_clear_the_air_no/
” Não vou aceitar essas mentiras”: Neymar nega que fez críticas aos jogadores após empate com a Chapecoense, VEJA VÍDEO
O atacante Neymar usou as redes sociais para negar que tenha feito críticas diretas aos jogadores das categorias de base do Santos após o empate da equipe contra a Chapecoense. Em stories publicados em seu perfil, o camisa 10 afirmou que informações divulgadas sobre uma suposta cobrança direcionada aos jovens atletas não são verdadeiras. A […]
https://terrabrasilnoticias.com/2026/07/nao-vou-aceitar-essas-mentiras-neymar-nega-que-fez-criticas-aos-jogadores-apos-empate-com-a-chapecoense-veja-video/
Eceng Gondok Disulap Jadi Media Jamur Merang, Buka Peluang Pendapatan Warga Bandung Barat
REPUBLIKA.CO.ID, BANDUNG BARAT - Eceng gondok yang kerap dianggap sebagai gulma atau tanaman pengganggu yang tumbuh subur di perairan Waduk Saguling, Kabupaten Bandung Barat (KBB), Jawa Barat akhirnya mulai...
https://rejabar.republika.co.id/berita/tiu07m368/eceng-gondok-disulap-jadi-media-jamur-merang-buka-peluang-pendapatan-warga-bandung-barat
IA é “grande truque de mágica” e investidor deve focar em chips, diz Kinea
https://www.infomoney.com.br/wp-content/uploads/2026/07/expert-xp-20260724-17h20-expert-session-6.jpg?fit=300%2C200&quality=70&strip=all
Ruy Alves, da Kinea, vê necessidade de máquinas mais potentes como fator que torna investimento em fabricantes de chips mais seguro
The post https://www.infomoney.com.br/onde-investir/ia-e-grande-truque-de-magica-e-investidor-deve-focar-em-chips-diz-kinea/ appeared first on https://www.infomoney.com.br.
https://www.infomoney.com.br/onde-investir/ia-e-grande-truque-de-magica-e-investidor-deve-focar-em-chips-diz-kinea/
Crise com os EUA entra na corrida presidencial em nova pesquisa AtlasIntel
https://www.infomoney.com.br/wp-content/uploads/2026/03/montagem-1120x720-1.jpg?fit=300%2C193&quality=70&strip=all
Pesquisa que será divulgada na terça-feira (28) volta a incluir a ex-primeira-dama nos cenários eleitorais
The post https://www.infomoney.com.br/politica/crise-com-os-eua-entra-na-corrida-presidencial-em-nova-pesquisa-atlasintel/ appeared first on https://www.infomoney.com.br.
https://www.infomoney.com.br/politica/crise-com-os-eua-entra-na-corrida-presidencial-em-nova-pesquisa-atlasintel/
Bitcoin ou Bitcoin Cash? Entenda as diferenças entre as duas criptomoedas
O Bitcoin Cash surgiu em agosto de 2017 após uma divergência técnica entre desenvolvedores do Bitcoin
https://investnews.com.br/investimentos/bitcoin-ou-bitcoin-cash-entenda-as-diferencas-entre-as-duas-criptomoedas/
Shootout Erupts At Seattle Festival, Killing 3; Second Gunman On The Run https://www.zerohedge.com/markets/shootout-erupts-seattle-festival-killing-3-second-gunman-run
https://nitter.poast.org/pic/card_img%2F2081707449879343104%2FUzphhZc9%3Fformat%3Dpng%26name%3D800x419
https://nitter.poast.org/zerohedge/status/2081707449401159780#m
擅闯中国驻日大使馆的24岁自卫官在东京被起诉 - RFI - 法国国际广播电台
27/07/2026 - 12:47
https://images.weserv.nl/?url=s.rfi.fr/media/display/3cbd95d2-89a8-11f1-9232-005056bfb2b6/w:1280/p:16x9/Capture-3626.jpg
就曾擅闯中国驻日大使馆的隶属于日本陆上自卫队虾野驻地（宫崎县）的三等陆尉村田晃大，东京地方检察厅周一以擅闯建筑物、违反《枪刀法》和进行威胁的罪名对其加以起诉。
据日本放送协会（NHK）报导，现年24岁的村田于今年3月被指控非法侵入并违反《枪刀法》，原因是他涉嫌携带一把刀刃长约18厘米的厨刀进入位于东京港区的中国驻日大使馆。 随后查明，该自卫官当时被大使馆相关人员当场控制，在接受问询时曾发表“日本的爱国者将代表日本的神将你们杀死”等言论，因此还被以威胁罪起诉。 东京地方检察厅为查明其刑事责任能力，在逮捕后委托专家进行了精神鉴定，并判定村田具有刑事责任能力。 报导指，另一方面，在这起事件中，该自卫官是从毗邻大使馆的一栋民用大楼翻越围墙闯入的，这凸显了在楼宇和公寓密集的市中心地区，大使馆安保工作面临的难题。事件发生后，警视厅向各国大使馆周边建筑的管理方提出了采取相应措施的要求等。 《朝日新闻》报导指，据对警视厅公安部及调查相关人员采访获悉，村田称，案发前一天，他从所属驻屯地所在的宫崎县出发，乘坐新干线和高速巴士前往东京。他在银座站附近的漫画咖啡馆住了一晚，并在附近的大型超市购买了一把厨刀。 案发当天，村田从地铁六本木站前往中国驻日大使馆。据推测，他在勘察了周边环境后，从相邻的大楼转移过去，翻越围栏，潜入了使馆院内。据悉，村田的手上留有疑似抓握带刺铁丝网时造成的细小伤痕。 村田原本藏身于使馆院内的一处灌木丛中，但看到使馆工作人员在室外吸烟后，便将厨刀藏在了那里。随后，待人流和车流都停歇后，他从灌木丛中走出，主动向碰巧遇到的使馆相关人员表示“想见大使”等。此时，距离他最初闯入使馆院内已过去约40分钟。 《读卖新闻》报导指，据称，当村田在使馆内要求会见大使而被控制时，工作人员递给他一张纸，要求他“写下想对大使说的话”，他便在这张纸上写下了威胁性言辞。 东京地方检察厅尚未透露村田是否承认或否认指控。
https://www.rfi.fr/cn/%E7%A4%BE%E4%BC%9A/20260727-%E6%93%85%E9%97%AF%E4%B8%AD%E5%9B%BD%E9%A9%BB%E6%97%A5%E5%A4%A7%E4%BD%BF%E9%A6%86%E7%9A%8424%E5%B2%81%E8%87%AA%E5%8D%AB%E5%AE%98%E8%A2%AB%E8%B5%B7%E8%AF%89
zankacore 🌧️ ate some recalled lettuce and spicy noodles earlie...
https://status.cafe/statuses/257212
Calls renewed for mandatory redress scheme for property businesses
https://i0.wp.com/jerseyeveningpost.com/wp-content/uploads/2026/07/Lesley-Horton-Chief-Property-Ombudsman-1.jpg?fit=1024%2C1024&ssl=1
CALLS to have property businesses sign up to a “mandatory” redress scheme have been renewed after a letting agent was fined for failing to resolve long-term damp and mould issues for two separate tenants. David Letto, director of PIOS Limited, appeared in the Magistrate’s Court on behalf of the property firm on 5 July after […]
The post https://jerseyeveningpost.com/news/2026/07/27/calls-renewed-for-mandatory-redress-scheme-for-property-businesses/ appeared first on https://jerseyeveningpost.com.
https://jerseyeveningpost.com/news/2026/07/27/calls-renewed-for-mandatory-redress-scheme-for-property-businesses/
Shootout Erupts At Seattle Festival, Killing 3; Second Gunman On The Run
Shootout Erupts At Seattle Festival, Killing 3; Second Gunman On The Run
Seattle police are searching for a second suspect after two gunmen allegedly exchanged fire inside the crowded Bite of Seattle festival beneath the Space Needle, killing three people and wounding four others, including a 2-year-old boy.
BREAKING: 2 killed, 5 injured in shooting at Seattle Center, Washington https://t.co/MPaA3lzO7W
— Rapid Report (@RapidReport2025) https://x.com/RapidReport2025/status/2081575773807218834?ref_src=twsrc%5Etfw
Assistant Seattle Police Chief Tyrone Davis told reporters late Sunday that investigators believe the suspects were shooting at each other when bystanders were caught in the crossfire.
Three people were killed and four others injured in a shooting at a Seattle food festival.
One suspect has been arrested, and authorities say the incident may have involved a shootout. Two firearms were recovered at the scene.
All 4 victims are in a stable condition.
Police… https://t.co/V6ynVSdmNE
— I Meme Therefore I Am 🇺🇸 (@ImMeme0) https://x.com/ImMeme0/status/2081623914040475696?ref_src=twsrc%5Etfw
One suspect surrendered at the scene, while the second remains at large.
"We're still trying to figure this out," Davis said.
Suspect Number One:
BREAKING - SEATTLE SHOOTING 🚨🚨🚨
Two people are dead, 5 others were injured after shooting at a Seattle festival.
Alleged shooter in image below. https://t.co/IkEDa0ehh1 https://t.co/DUs05mBGxV
— TERFs ‘r’ us ©️ (@Terfs_R) https://x.com/Terfs_R/status/2081656826291016188?ref_src=twsrc%5Etfw
Davis described the suspect in custody as "a young person" who was being questioned by investigators. He said that individual was the only gunman officers witnessed opening fire. He noted that police did not discharge their weapons during the confrontation.
A Seattle Times reporter at the festival described hearing several loud pops followed by what sounded like rapid gunfire.
"It was just pure chaos," one festival attendee said.
Authorities have not indicated that the shooting is being investigated as terrorism, despite the heightened security following the Islamist attack at https://www.zerohedge.com/political/manhunt-underway-islamic-state-suspect-after-deadly-car-attack-berlin-pride-parade. The State Department has also https://www.zerohedge.com/political/watch-live-rubio-bessent-convene-65-nations-global-crackdown-far-left-political-terrorism a far-left threat across the West.
https://cms.zerohedge.com/users/tyler-durden
Mon, 07/27/2026 - 07:45
https://www.zerohedge.com/markets/shootout-erupts-seattle-festival-killing-3-second-gunman-run
Teen accused of planning bomb hoax at Peter Kay show appears in court
Grinning Omar Majed did not speak during the hearing
https://i2-prod.manchestereveningnews.co.uk/incoming/article34358067.ece/ALTERNATES/s615/0_Peter-Kay-gig-evacuated.jpg
https://www.walesonline.co.uk/news/world-news/teen-accused-planning-bomb-hoax-34358420
Carteira de pedidos da Embraer atinge recorde de US$ 34,5 bilhões no segundo trimestre
Aviação Comercial ultrapassa US$ 15 bilhões enquanto Defesa & Segurança registra o crescimento anual mais rápido
https://www.airway.com.br/carteira-de-pedidos-da-embraer-atinge-recorde-de-us-34-5-bilhoes-no-segundo-trimestre/
Путин заявил об устойчивости российской финансовой системы
Президент России Владимир Путин заявил об устойчивости российской финансовой системы.
https://russian.rt.com/business/news/1662196-putin-ustoichivost-finansovaya-sistema?utm_source=rss&utm_medium=rss&utm_campaign=RSS
https://russian.rt.com/business/news/1662196-putin-ustoichivost-finansovaya-sistema?utm_source=rss&utm_medium=rss&utm_campaign=RSS
https://blossom.primal.net/62bdbb561b6bfffe0bf9f370178f20108da0a7b85e7bd74124b5ed8917157c5b.jpg What is your definition of freedom?
Simply put? My definition of freedom rests on three principles... having the vitality and energy to play, the autonomy and space to create my own path and the right people beside me so I can truly love the life I live.
Freedom means
having power to say NO
to things that do not align with my soul
to say YES to projects that excite me
even if they don't turn a profit
Ultimately? It’s the autonomy to fulfill my purpose here. To observe life. To maintain perspective. Not working for the dreams of others, but living in alignment with my own values.
zankacore 🌧️ I hate running out of food once a month
https://status.cafe/statuses/257211
Prefeitura de Bandeirantes (MS) oferta iniciais de R$ 5,7 mil em seletivo
Processo seletivo Prefeitura de Bandeirantes, no estado de Mato Grosso do Sul, oferta oportunidades de nível superior.
O post https://concursosnobrasil.com/concursos/ms/2026/07/27/prefeitura-de-bandeirantes-ms-oferta-iniciais-de-r-57-mil-em-seletivo/ apareceu primeiro em https://concursosnobrasil.com.
https://concursosnobrasil.com/concursos/ms/2026/07/27/prefeitura-de-bandeirantes-ms-oferta-iniciais-de-r-57-mil-em-seletivo/
Can we teach a model to encode a semantic feature on a chosen manifold in just three channels?
This is my submission to https://bluedot.org/puzzles/technical-ai-safety, for which I received an Honorable Mention.
Congratulations to Gustavo Korzune Gurgel, Patryk Perduta (https://www.lesswrong.com/posts/TRno3dELwmMHYYH5v/bluedot-technical-ai-safety-puzzle-submission-that-got-me-a), Sam Spilllard, Karine Levonyan, and Michael Zlatin for their recognition in the puzzle.
My article below focuses on my answer to Task 3: training a small MLP to encode country feature through a chosen nonlinear manifold in three reserved channels.
My Task 1 and 2 write-up is available on https://phusroyal.github.io/writing/bluedot-1st-puzzle-q12/, and the https://phusroyal.github.io/writing/bluedot-1st-puzzle-q3/.
I welcome discussion, feedback, and collaborations that could extend this idea.
You can check out the code for this article in my https://github.com/phusroyal/Phu-BlueDot_1st_puzzle.
The puzzle and the question
https://res.cloudinary.com/lesswrong-2-0/image/upload/v1784909291/lexical_client_uploads/q8lcxcx6ajjoiuhqfxhu.png
Model architecture provided with the puzzle. The investigated representation is the output of the third ReLU.
https://bluedot.org/puzzles/technical-ai-safety provides a trained five-layer MLP for multi-label classification over eight binary features, using mean-pooled sentence-transformer representations. The puzzle identifies nonlinear behavior at the output of the third ReLU, denoted as , and asks participants to:
- find the nonlinear feature ;
- explain the geometry used at to represent ; and
- train a new model with a more interesting representation.
This post addresses the third task. I train a new five-layer MLP and constrain country to use a chosen three-dimensional manifold while testing whether the classifier relies on that code.
The guiding question is:
Can we choose a representation geometry first, then train the model so that the feature follows that geometry while still solving the original task?
This is inspired by work on counting manifolds in language models [1] and manifold steering [2]. I use the opposite direction: instead of discovering a manifold after training, I choose a small manifold code first and ask the model to use it.
The manifold must not be decorative. I therefore ask both:
Did the hidden activations form the desired shape?
But also:
Did the model actually use that shape to solve the task?
My design has five requirements:
- Specify the geometry before training.
- Make the learned geometry code match it.
- Preserve the original eight-label task.
- Weaken easy shortcuts, especially linear and complement-only access.
- Show that interventions on the geometry change the country prediction.
1. Define the Target Manifold before Training
The first stage has no model. I only define what the model will later be asked to learn.
1.1 The Distributional Problem
For each example, the dataset gives the binary label . It says whether country is present, but does not say where the example belongs on a desired manifold.
For a helix, supplies no position , radius , or tube angle . If those coordinates were observed, I could use point-wise supervision to train my model.
Instead, I use a distributional target:
Here, and are the learned geometry codes for positive and negative target examples. The model is free to choose individual placements; the aggregate positive and negative code distributions must match and . This follows the aggregate-distribution perspective of Wasserstein Auto-Encoders [3].
I choose two target geometries: a sphere shell and a helix tube, shown in Figure 1. In both, positives occupy an inner radius and negatives an outer radius while both classes span the same scaffold. The label is therefore a distance-to-structure decision rather than a global direction. If positives occupied only a sphere's north pole and negatives only its south pole, the representation would be essentially linear.
1.2 Sphere Shell
Let be a direction sampled uniformly from the unit sphere , the set of points in 3-dimensional space with Euclidean norm 1.
Let be a radius. The raw sphere point is
The positive class uses a smaller radius than the negative class:
with and . Here, is the target radius for positive examples and is the target radius for negative examples. The approximation means the sampled radius is concentrated near that value, not exactly fixed at that value. Thus country is encoded by an inner versus outer shell, not by direction.
1.3 Helix Tube
Let be a position along the helix and its vertical pitch (to control how fast the helix rises). The helix core curve is
Two local cross-section directions around that center line are
With cross-section angle and tube radius , a point in the tube is
The class-conditional distributions are
I use , , and . Both classes trace the same curved scaffold.
I tried three turns, but the tail layer did not reliably read them under the remaining constraints; one turn was learnable, usable, and testable.
1.4 Normalization
The sphere and helix have different natural scales, so I normalize sampled points to test shape rather than raw norm. Let be a raw point, with for the sphere and for the helix. With balanced classes,
For the sphere shell,
and for the helix tube,
Here is the radius-noise width. Its term is the uniform-interval variance rule [4]. The helix additionally contributes from its circular core and from its centered vertical coordinate.
Before training, I draw normalized target anchors and .
2. Train the Model
2.1 Base Classifier
Each text input is encoded by sentence-transformers/all-MiniLM-L6-v2, then mean-pooled into . A five-layer ReLU MLP maps it to eight logits and is trained with binary cross-entropy:
The ordinary classifier reaches mean AUC , mean accuracy , and country AUC . This establishes that later failures come from manifold constraints rather than basic architecture or data handling.
2.2 Add the Geometry Bottleneck and Activation-sign
2.2.1 Geometry Bottleneck
I reserve three hidden2 pre-activations for a geometry code , leaving 61 complement coordinates . Three channels are the minimal space for a point on either target 3D manifold.
Because hidden2 is post-ReLU, I add positive offset so a signed manifold can live in activation space:
Let be the chosen geometry family, either the sphere shell or the helix tube. The learned coordinate head predicts three unconstrained values, and fixed map converts them into a normalized point on the chosen manifold:
For the sphere bottleneck,
where is the azimuth angle , is the vertical coordinate, and has maximum allowed radius .
For the helix bottleneck,
where , , and .
2.2.2 Experiment
At this stage, I optimize only . The purpose is not yet to force good class geometry. I ask only:
Can the classifier survive this architectural constraint?
Alongside mean and country AUC, I use four analyses:
- Geometry probe. A fixed readout from alone measures whether its sphere radius or distance to the helix core is closer to the positive or negative target radius:
Its ROC AUC shows whether the reserved code contains the label in the intended geometric form; it does not prove the tail uses that code.
- Coverage entropy. For the sphere, I bin the azimuths of learned points into eight bins. For the helix, I map each point to its nearest core position and bin it into 24 helix positions. High entropy means codes spread across the intended manifold rather than collapsing into one patch.
- Causal target delta. I replace the geometry channels with positive and negative anchors, run the tail, and compute the mean country-logit difference. A large positive value means the tail listens to geometry; a near-zero value means the geometry may look correct but the tail ignores it.
- Linear probe AUC. I train a linear probe on the full country activations. Lower is better: it means country is less exposed as an ordinary linear direction.
2.2.3 Results
Table 1. Model behavior after adding the geometry bottleneck. Higher is better unless marked ↓.
Geometry
Mean task AUC ↑
Country AUC ↑
Geometry probe AUC ↑
Linear probe AUC ↓
Coverage entropy ↑
Causal delta ↑
Sphere shell
0.9975
0.9995
0.2021
0.9996
0.7971
0.0187
Helix tube
0.9963
0.9997
0.1182
0.9997
0.6561
0.0357
The model survives the bottleneck, but the near-zero causal deltas show that it does not use the intended geometry. Low geometry-probe AUC and coverage entropy also show that country does not yet follow the prescribed shape. The model is routing the label through other complement channels.
2.3 Add Unconditional Optimal-transport Coverage
Can the learned geometry codes cover the target geometry instead of collapsing to a small region?
The bottleneck constrains codes to the allowed family, but not to the whole geometry. Every example can lie on a valid helix tube while almost all examples occupy one segment.
To prevent this collapse, I add unconditional mixture optimal transport (MixOT), which spreads the unlabeled mini-batch across the overall target geometry:
where is the country-positive rate.
For mini-batch , let be its n learned 3D geometry codes and sample equally many anchors from .
Think of learned codes as students and target anchors as seats across the shape. OT matches each student to a seat while every seat must receive a match. Collapsed codes leave distant seats unmatched and are expensive; spread-out codes make the matching cheaper.
Therefore, for a learned code and a target anchor . The pairwise transport cost is
Entropic OT solves
subject to
Here is the transport plan. The first constraint assigns every learned point, and the second ensures every target anchor receives mass. This second constraint makes collapse expensive: the model cannot ignore most anchors and only match the easy ones. The entropy weight smooths the matching for efficient Sinkhorn optimization [5]
To test whether class-conditional geometry works on its own, I replace the previous MixOT term with:
2.3.1 Results
I add Radius MAE, the mean absolute error between learned and desired class radius (lower is better).
Table 2. Model behavior after adding unconditional mixture optimal transport.
Geometry
Stage
Mean task AUC ↑
Country AUC ↑
Geometry probe AUC ↑
Linear probe AUC ↓
Coverage entropy ↑
Causal delta ↑
Radius MAE ↓
Sphere shell
Bottleneck
0.9975
0.9995
0.2021
0.9996
0.7971
0.0187
0.7243
MixOT
0.9974
0.9990
0.5585
0.9993
0.9991
0.0206
0.3012
Helix tube
Bottleneck
0.9963
0.9997
0.1182
0.9997
0.6561
0.0357
0.6653
MixOT
0.9975
0.9993
0.3095
0.9994
0.9942
0.0368
0.5582
MixOT makes coverage almost perfect, but geometry-probe AUC remains weak, radius MAE remains large, and causal delta remains nearly zero.
This is expected: it spreads the unlabeled batch but does not assign positives to and negatives to .
2.4 Add Class-conditional Geometry
Put positive examples in the positive part of the shape, and negative examples in the negative part.
MixOT spreads the batch but does not assign each class to its own part. I replace its shared seating chart with class-conditional matching and add local per-example signals.
2.4.1 Class-conditional OT (ClassOT)
For , let and sample anchors from . Then
where is the number of included label groups; I skip a class with fewer than two examples in the batch.
2.4.2 Radius Loss
ClassOT gives the right distributional shape, but a batch-level match can be weak local training signal. The fixed tail has to decode country from each geometry code. Radius loss supplies a simple cue: each example should reach its own class radius.
Let be distance from the origin for the sphere or distance to the nearest helix core point for the helix:
2.4.3 Geometry-score Loss
Geometry-score loss asks whether the geometry itself would classify a point correctly. Define
and the logit-like score
It is high near the positive radius and low near the negative radius, so I optimize
2.4.4 Geometry-score Loss versus Radius Loss
ClassOT decides where the two populations should lie. Radius loss asks each point to reach its assigned radius. Geometry-score loss instead asks whether the geometry itself would classify the point correctly: it turns distance to the positive and negative radii into a logit-like score, so a point closer to should look more country-like and a point closer to more negative-like.
For and , a positive at is good under both losses. A positive at is still pulled toward by radius loss, while geometry-score loss considers it acceptable because it remains closer to than .
2.4.5 Experiment
I remove the previous stage's mixture-OT term to test whether class-conditional geometry works on its own.
2.4.6 Results
Table 3. Model behavior after replacing unconditional mixture OT with class-conditional OT.
Geometry
Stage
Mean task AUC ↑
Country AUC ↑
Geometry probe AUC ↑
Linear probe AUC ↓
Coverage entropy ↑
Causal delta ↑
Radius MAE ↓
Sphere shell
Bottleneck
0.9975
0.9995
0.2021
0.9996
0.7971
0.0187
0.7243
MixOT
0.9974
0.9990
0.5585
0.9993
0.9991
0.0206
0.3012
ClassOT
0.9962
0.9996
0.9998
0.9998
0.9970
0.0188
0.1286
Helix tube
Bottleneck
0.9963
0.9997
0.1182
0.9997
0.6561
0.0357
0.6653
MixOT
0.9975
0.9993
0.3095
0.9994
0.9942
0.0368
0.5582
ClassOT
0.9949
0.9996
0.9996
0.9997
0.9953
0.0389
0.0974
ClassOT writes country into the intended geometry but does not make the classifier read from it. Geometry-probe AUC reaches for the sphere and for the helix, but causal delta remains near zero ( and ). The tail can still ignore the prescribed geometry and use other hidden2 signals. Near-perfect linear-probe AUC shows that country also remains easy to read linearly.
2.5 Make the Geometry Functional and Reduce Linear Access (GFAL)
Can country remain useful while becoming harder to read as one ordinary linear direction?
I call this combined stage GFAL, for Geometry Functional and Anti-Linear. It retains the task, ClassOT, radius, and geometry-score losses; restores MixOT coverage; and adds causal geometry pressure, tail fitting, anti-linear pressure, and a squared-correlation penalty.
These additions repair different failures. Causal pressure and tail fitting make the tail respond to the three geometry channels. MixOT protects coverage, especially for the helix. Anti-linear pressure and correlation penalties weaken ordinary linear country readouts from hidden2.
2.5.1 Causal Geometry Pressure
If I replace only the geometry channels, does the model's own country logit move?
I hold complement fixed and create positive and negative edited states:
Let be the model’s tail from hidden2 to logits. The target pressure asks the tail to respond correctly: should be high while should be low. The target causal response is
I also penalize off-target spillover to avoid affecting other labels:
so that
Positive geometry should raise the country logit and negative geometry should lower it [6, 7], without moving every other output.
2.5.2 Tail Fitting
hidden2 → hidden3 → logits
Tail fitting freezes earlier layers and trains only the final tail. It teaches the tail how to decode the geometry channels into the country logit; it assumes the geometry code is already present rather than shaping it itself.
Causal geometry pressure and tail fitting are not redundant because the model can fail in two separate ways:
- Good geometry, bad use. The three channels contain the intended structure, but the tail ignores them. Tail fitting gives the decoder a focused reason to read them.
- Bad geometry, good reader. The tail is willing to read the channels, but the channels do not form the desired structure. Tail fitting cannot repair this; geometry losses and causal pressure do that work.
Geometry losses write the code in the right language, causal pressure checks that changing the code changes the answer, and tail fitting teaches the final reader how to read that language.
With , I use:
- Context state: . This lets the tail fit on complement values for each example, but prevents learning signal from flowing back to earlier layers to hide country information in these complement channels.
- Neutral state: , where is the mini-batch mean complement vector. If the model tail can still predict country from this state, the example-specific signal must come from the geometry channels , because the complement no longer carries example-specific clues.
The tail predicts from both states and follows the geometry score:
The model tail must predict the target label from both controlled states. As is the country logit produced by the model tail, then:
The target logit is trained to follow the intended geometry score. As should be high when is near the positive target radius and low when is near the negative target radius, tail fitting adds:
The on means this score is treated as a fixed teacher signal. The model tail must move toward the score; the geometry score itself is not adjusted to make the loss easier.
2.5.3 Anti-linear Pressure
Even after geometry contains country, the full hidden2 state can expose a simple linear country direction. I add a one-layer adversary
with loss
Gradient reversal leaves hidden2 unchanged in the forward pass but reverses its gradient in the backward pass. The adversary therefore learns to predict country, while the representation learns to make that linear prediction worse. Equivalently,
The objective is not to erase country entirely. The main task still needs country and geometry losses still require the first three channels to carry its manifold. It is to remove the arbitrary straight-line shortcut, following the gradient-reversal mechanism of Ganin et al. [8].
2.5.4 Squared Correlation Penalty
An adversary can underfit or miss one narrow channel that quietly tracks country. I therefore add a direct backup check for every hidden2 channel:
This does not prove that the complement is clean, but it closes the easy one-channel shortcut.
2.5.5 Experiment
2.5.6 Results
Table 4. Model behavior after adding causal geometry pressure, tail fitting, anti-linear pressure, and squared correlation penalty (GFAL: Geometry Functional and Anti-Linear).
Geometry
Stage
Mean task AUC ↑
Country AUC ↑
Geometry probe AUC ↑
Linear probe AUC ↓
Coverage entropy ↑
Causal delta ↑
Radius MAE ↓
Sphere shell
Bottleneck
0.9975
0.9995
0.2021
0.9996
0.7971
0.0187
0.7243
MixOT
0.9974
0.9990
0.5585
0.9993
0.9991
0.0206
0.3012
ClassOT
0.9962
0.9996
0.9998
0.9998
0.9970
0.0188
0.1286
GFAL
0.9799
0.9995
0.9997
0.5743
0.9993
3.0845
0.1444
Helix tube
Bottleneck
0.9963
0.9997
0.1182
0.9997
0.6561
0.0357
0.6653
MixOT
0.9975
0.9993
0.3095
0.9994
0.9942
0.0368
0.5582
ClassOT
0.9949
0.9996
0.9996
0.9997
0.9953
0.0389
0.0974
GFAL
0.9739
0.9957
0.9940
0.6710
0.7510
1.8980
0.1887
GFAL makes the geometry functional. Geometry-probe AUC stays at for the sphere and for the helix, while causal delta rises from near zero to and . Geometry edits now move the model's country logit.
Linear-probe AUC falls from in the ClassOT stage to . The sphere result is cleaner; the helix retains lower coverage and higher linear access, so MixOT remains necessary.
2.6 Reduce complement shortcuts
GFAL makes geometry functional, but does not prove the remaining 61 channels are harmless.
If I remove geometry code and look only at the remaining 61 channels, can I still read country?
If yes, geometry is functional but not primary: the model has a backup route. GFAL+ adds a second one-layer adversary :
or, in probability form,
The complement-correlation penalty provides a direct backup check for any single complement channel that tracks country:
2.6.1 Why keep both anti-linear and complement adversaries?
Why do I not remove the anti-linear adversary and use only the complement adversary? Because complement cleanup is narrower: the complement adversary only looks at the non-geometry channels.
The complement adversary asks: can country be read after removing the geometry channels?
The anti-linear adversary asks: is country still easy to read as a simple linear direction from the whole hidden2 state?
The first question is important but cannot replace the broader question from the anti-linear adversary. The whole hidden2 state contains both and , and the tail classifier sees both. Without full-hidden2 anti-linear pressure, the model can reopen an easy linear shortcut using their mixture.
The squared correlation penalty remains for the same reason. The adversary attacks a learned linear readout, but it depends on optimization; correlation attacks the simpler failure where one hidden2 channel quietly tracks country. This stage adds complement-specific pressure without declaring the earlier shortcut solved forever. Keeping both is a guardrail: while the model cleans up , country must not become trivially exposed again in the full representation.
2.6.2 Experiment
2.6.3 Results
Table 5. Model behavior after adding complement-adversary pressure (GFAL+).
Geometry
Stage
Mean task AUC ↑
Country AUC ↑
Geometry probe AUC ↑
Linear probe AUC ↓
Coverage entropy↑
Causal delta ↑
Radius MAE↓
Complement AUC ↓
Sphere shell
Bottleneck
0.9975
0.9995
0.2021
0.9996
0.7971
0.0187
0.7243
0.9996
MixOT
0.9974
0.9990
0.5585
0.9993
0.9991
0.0206
0.3012
0.9981
ClassOT
0.9962
0.9996
0.9998
0.9998
0.9970
0.0188
0.1286
0.9998
GFAL
0.9799
0.9995
0.9997
0.5743
0.9993
3.0845
0.1444
0.5758
GFAL+
0.9674
0.9996
0.9997
0.6204
0.9997
3.5004
0.1171
0.5728
Helix tube
Bottleneck
0.9963
0.9997
0.1182
0.9997
0.6561
0.0357
0.6653
0.9997
MixOT
0.9975
0.9993
0.3095
0.9994
0.9942
0.0368
0.5582
0.9994
ClassOT
0.9949
0.9996
0.9996
0.9997
0.9953
0.0389
0.0974
0.9996
GFAL
0.9739
0.9957
0.9940
0.6710
0.7510
1.8980
0.1887
0.6071
GFAL+
0.9576
0.9949
0.9927
0.5992
0.7787
2.6884
0.2111
0.5681
GFAL+ preserves task performance, strong geometry probes, and causal geometry for both manifolds. Its complement result is asymmetric: complement AUC falls from to for the helix but rises from to for the sphere. Thus it reduces a helix backup route but does not establish perfect information isolation for either geometry.
3. Causal-use validation
If I edit only the geometry channels after training, does the trained classifier actually follow that edit?
The earlier probes show that geometry channels contain country information; they do not show that the classifier uses them. I freeze the model, select a balanced held-out subset, and edit only hidden2 geometry channels. I use four interventions:
- Ablation: set to zero and ask whether country prediction worsens.
- Replacement: insert positive or negative anchors and ask whether the country logit follows.
- Swap: exchange learned geometry codes between labels and ask whether predictions move as expected.
- Path: move smoothly from positive to negative geometry and ask whether the country logit moves smoothly.
For replacement and swap, I also measure specificity: whether country moves much more than the other seven logits.
3.1 Ablation
I replace with zero, then compute target AUC. The ablation drop is
If I remove the geometry signal, does country prediction get worse?
3.2 Replacement
I replace learned geometry with positive or negative anchors while preserving the original complement:
Here is the intervened hidden2 state. Because is copied from the original example, the replacement geometry is the only intended cause of output movement. The causal target delta is
3.3 Swap
For positive-negative pairs , I exchange only their geometry channels. Let and be the selected positive and negative indices. The original hidden states are:
Now swap only the geometry channels:
The swap target shift is
It is large when negative geometry lowers positive examples and positive geometry raises negative examples.
3.4 Path
I choose a deterministic path through each target geometry from the positive to the negative radius:
Rather than jumping directly between the endpoints, I traverse intermediate geometry points and check whether the country logit changes smoothly. For the sphere, I fix one direction and increase radius from to . For the helix, I fix one core position and tube angle, then increase only the tube radius. I hold the complement at mean :
Let is the country logit at step . Since the path moves from positive geometry to negative geometry, the expected behavior is that country logit decreases. The path target delta is:
The path monotonic fraction is:
means every adjacent step moves in the expected direction.
3.5 Specificity
Specificity compares country movement with average off-target movement:
A high ratio means the geometry edit mainly affects country rather than every feature.
3.6 Results
Table 6. Causal-use validation on the trained model from the previous stage.
Geometry
Ablation AUC drop
Ablation target AUC
Causal delta
Swap target shift
Swap specificity
Path target delta
Path monotonic fraction
Specificity ratio
Sphere shell
0.4675
0.5317
3.5004
4.4433
92.0223
3.4963
1.0000
73.4252
Helix tube
0.4279
0.5654
2.6884
7.1863
128.0430
5.1727
1.0000
47.3151
Both geometries pass every causal check. Ablation nearly removes country prediction, leaving both near chance. Replacement moves country logits in the expected direction with specificity ratios and . Each path has monotonic fraction .
The swap test is the strongest learned-code check because it exchanges the model's own geometry codes rather than synthetic anchors. It yields target shifts for the sphere and for the helix, with specificities and . The learned geometry is therefore not merely visually aligned with country: moving it between examples changes the model's prediction path as expected.
The sphere is geometrically cleaner, with higher coverage entropy and lower radius MAE, but both geometries show strong causal use. These tests establish a causally active pathway, not perfect information isolation: the complement can remain an alternative country-information source.
Conclusion
This experiment shows that a small MLP can encode a semantic feature through a pre-chosen nonlinear manifold in only three channels while preserving task performance. A sphere shell and helix tube both support a code that is geometrically well formed and causally used by the classifier.
The result is not that country has been isolated perfectly. GFAL+ leaves evidence of complement leakage, and the helix is a harder geometry to maintain than the sphere. The useful conclusion is narrower: geometry can be specified before training, made readable from a compact code, and tested causally rather than accepted because it looks interpretable.
Reference
- Gurnee et al. https://arxiv.org/abs/2601.04480.
- Wurgaft et al. https://arxiv.org/abs/2605.05115.
- Tolstikhin et al. https://arxiv.org/abs/1711.01558.
- ProofWiki. https://proofwiki.org/wiki/Variance_of_Continuous_Uniform_Distribution.
- Cuturi. https://arxiv.org/abs/1306.0895.
- Geiger et al. https://proceedings.mlr.press/v162/geiger22a.html.
- Geiger et al. https://arxiv.org/abs/2106.02997.
- Ganin et al. https://arxiv.org/abs/1505.07818.
https://www.lesswrong.com/posts/ZwEer94AefjdW4933/can-we-teach-a-model-to-encode-a-semantic-feature-on-a#comments
https://www.lesswrong.com/posts/ZwEer94AefjdW4933/can-we-teach-a-model-to-encode-a-semantic-feature-on-a
🔔 Proč přestal Trump útočit na Írán? Mluví se o diplomacii i nedostatku raket:
Spojené státy po víc než týdnu trvajících úderech už dva dny neútočí na Írán. Teherán též pozastavil své útoky na Blízkém východě. Washington naznačuje, že důvodem je diplomacie, ale zákulisní informace ukazují méně přívětivý obrázek. USA totiž klesají zásoby raket pro protivzdušnou obranu a údajně si kvůli tomu i vybírají, které íránské střely propustí.
https://www.idnes.cz/zpravy/zahranicni/iran-usa-rakety-zasoby-valka-nedostatek-trump.A260727_113712_zahranicni_jhr
#CzechNews #News #Press #Media
Agnez Mo, Anggun Represent 'Reacher' at San Diego Comic-Con
Actor Alan Ritchson joined Agnez Mo and Anggun to represent the "Reacher" Season 4 team during a panel discussion at San Diego Comic Con.
https://en.tempo.co/read/2115657/agnez-mo-anggun-represent-reacher-at-san-diego-comic-con
Agnez Mo, Anggun Represent 'Reacher' at San Diego Comic-Con
Actor Alan Ritchson joined Agnez Mo and Anggun to represent the "Reacher" Season 4 team during a panel discussion at San Diego Comic Con.
https://en.tempo.co/read/2115657/agnez-mo-anggun-represent-reacher-at-san-diego-comic-con
Analis Politik Sebut Londo Ireng Itu Pejabat Korup
Analis politik mengatakan semestinya pelabelan londo ireng disematkan kepada pejabat atau mantan pejabat yang korup
https://nasional.tempo.co/read/2115656/analis-politik-sebut-londo-ireng-itu-pejabat-korup
Debian Sid Installation Guide (PowerPC)
https://forums.macrumors.com/threads/debian-sid-installation-guide-powerpc.2146795/
Britische Regierung beschließt Abgabe für Elektroautos: Was bedeutet das für Deutschland?
https://mf.b37mrtl.ru/deutsch/images/2026.07/thumbnail/6a6679e0b480cc692b788c5c.jpg
Großbritannien führt ab dem ersten April 2028 eine Abgabe für Elektroautos ein. Fahrer zahlen rund 2,2 Cent pro Kilometer. Reine Elektroautos und Plug-in-Hybride sind betroffen. Ob Deutschland dem Beispiel folgt, bleibt offen.
https://mf.b37mrtl.ru/deutsch/images/2026.07/thumbnail/6a6679e0b480cc692b788c5c.jpg
https://de.rt.com/europa/286963-britische-regierung-beschliesst-abgabe-fuer/
Em dashes are fucking amazing
I fucking love em dashes. I spam them all the time. They make me feel free.
I don’t give a shit what the “oh yeah let’s examine this piece of writing with a magnifying glass for any signs of AI” crowd thinks. I don’t give a shit that AI figured out how to produce em dashes before these online troglodytes did. Typing em dashes is not an AGI-complete problem. It’s not that difficult to produce an organic, artisanal, hand-crafted em dash — on Mac, for example, it’s simply Option + Shift + Hyphen.
I fucking love em dashes. There is no greater joy in life than typing a sentence, realising it needs a clarification, a caveat, an addendum — but in the same sentence, not a new one — and, guess what, the em dash is right there for you. You want to package things neatly together — and the em dash lets you do exactly this. Em dashes make writing feel like you are assembling lego bricks.
Em dashes can be used instead of brackets in a sentence — like this one for example — and you can just type your shit and keep going. Brackets are for people who raise their hand in meetings and say: “this might be a stupid question”. A parenthetical is whispering: “Sorry, don’t mind me, I’ll be quick”. Em dash kicks the door open and announces itself. Use brackets when you are embarrassed by your own thought.
And also occasionally — occasionally — one just needs a random dramatic pause. https://x.com/esrtweet/status/1889785599340486802, after all. It’s not like grammar and punctuation rules were sent to us by god in their final form. People were just writing shit, and at some point the most popular patterns got codified. Sure, you need commas to signify a small break. Then em dash is a natural way to express a larger, longer, break. A period is great to express a complete thought — and it shifts the register of the next letter.
But colons are so nearly useless — they are the same width as a comma and unlike periods they don’t change the register of the following letter. Em dashes are of a clearly different length. And most colons are better off as em dashes to delineate different parts of the sentence.
As for semicolons, semicolons are neat, they are like the coolest punctuation mark after the em dash — the fundamental indecision compressed into a single punctuation mark — you are neither starting a new sentence nor you are, exactly, continuing an existing one. You are just… piling your stuff together. I am always happy when I find a place for the little guy semicolon. But, again, most cases where you could use one are better off served by the em dash as well.
Speaking of ellipsis… Ellipsis is a tad too melancholic. It’s nice, it’s fine, it’s like being in a garden and having limerence about a woman you can’t have. If you have a text where this is appropriate — by all means use ellipsis. I usually want to take my readers to the highest highs instead of idly sitting around with them. And em dashes are like a lift to the top floor of a high-rise hotel building in Tokyo — an one that takes you there in seconds. The ellipsis trails off while em the em dash cuts forward.
“But what would people think of my writing if I ever used an em dash? What if someone points one out in my writing?”. Bitch, please. Are you writing for an internet rando who’s got nothing better to do than to hunt for em dashes? Fine, you do, okay, fair enough. Then I’ve got you covered: use AP-style em dashes, these are separated from the text by spaces instead of being glued to it. That’s how every em dash in this piece is styled — and it’s unlike AI em dashes which are “like—this, word—word”. You can now outnitpick your hostile interlocutor by explaining that you’re using the special sort of em dashes AI doesn’t produce by default.
But also, maybe consider not giving a fuck about what someone on the internet think? Are you really that easily swayed by a rando’s opinion? You know what, we can play this game the other way. If your writing doesn’t contain an em dash, I ain’t reading it. From now on you are obliged to use em dashes for my pleasure.
https://www.lesswrong.com/posts/FWTGBdmkNdePNKvLN/em-dashes-are-fucking-amazing#comments
https://www.lesswrong.com/posts/FWTGBdmkNdePNKvLN/em-dashes-are-fucking-amazing
DorcelClub.26.07.06.Peach.Lollypop.Cara.Mella.And.Bonnnie.Blonde.XXX.1080p.HEVC.x265.PRT
File Size: 670 MB
Duration: 0:39:38
https://pornrips.to/dorcelclub-26-07-06-peach-lollypop-cara-mella-and-bonnnie-blonde-xxx-1080p-hevc-x265-prt/
Oil price dives as US and Iran pause attacks
The US says attacks on Iran have been halted to give "talks some space", raising hopes of a resolution to the conflict.
https://www.bbc.co.uk/news/articles/clyj834jn5lo?at_medium=RSS&at_campaign=rss
Oil price dives as US and Iran pause attacks
The US says attacks on Iran have been halted to give "talks some space", raising hopes of a resolution to the conflict.
https://www.bbc.co.uk/news/articles/clyj834jn5lo?at_medium=RSS&at_campaign=rss
🔔 Za odkladem útoků na Írán je i nedostatek munice v USA, píše WSJ:
Prezident USA Donald Trump odložil očekávané masivní údery na Írán nejen ve snaze oživit diplomatické úsilí, ale také kvůli debatám ohledně nedostatku munice. Listu The Wall Street Journal (WSJ) to řekli američtí představitelé obeznámení se situací. Trump to popřel. „Máme mnohem více munice než kdokoli na světě a mnohem více, než potřebujeme,“ prohlásil.
https://ct24.ceskatelevize.cz/clanek/svet/za-odkladem-utoku-na-iran-je-i-nedostatek-munice-v-usa-pise-wsj-375991
#CzechNews #News #Press #Media
Good Morning
Stay Sovereign
Godspeed
10 Gbps home internet
https://forums.macrumors.com/threads/10-gbps-home-internet.2270464/
📰 **In this week's issue:**
🗞️ **BREAKING**
The Bulgarian Resort That's Redefining Luxury (And It's Not Where You Think)
Think Bulgaria's Black Sea coast is all about budget-friendly package holidays and crowded beaches? You couldn't be more wrong. While the rest of the industry struggles with a 30% drop in bookings, one adults-only sanctuary has just been crowned the most luxurious hotel on the entire coastline for the second year in a row, boasting a near-perfect 9.8/10 guest rating. It only has 11 suites, it's located in a resort most travelers have never heard of, and it's completely redefining what "exclusive" means in Eastern Europe. What's the secret that's making it a global outlier in a season of crisis?
https://image.nostr.build/5fa2a6baa76bce5729c8926ee07fc9236366b467f45bd04a7ada17c8cac5626f.jpg
✍️ Author: NM team
🔗 https://nostrmag.com/article/w30travel01
📈 id#4040163
📰 **In this week's issue:**
🗞️ **BREAKING**
Wallet of Satoshi Review: The Lightning Wallet That’s Redefining “Simple”
Think you need to be a Bitcoin maxi to use the Lightning Network? Wallet of Satoshi processed over 28 million transactions since 2019—and it didn't get there by catering to crypto experts. It got there by being so simple that anyone can use it. But here's the tension that keeps me up at night: the very thing that makes it accessible—its custodial model—is also the thing that makes it controversial. So does convenience outweigh control, or is this wallet selling something that Bitcoiners shouldn't buy?
https://image.nostr.build/f6f1a81fc1b277e4537cef5c76384ec909846475633dbd2d005cf020dc62f093.png
✍️ Author: NM team
🔗 https://nostrmag.com/article/w30bitcoin01
📈 id#8075856
Datafolha: 52% veem tarifas de Trump como apoio à campanha de Flávio Bolsonaro
Governo norte-americano anunciou na semana passada um acréscimo de 12,5% nas sobretaxas sobre produtos brasileiros; total pode chegar a 37,5%
https://jovempan.com.br/politica/datafolha-52-veem-tarifas-de-trump-como-apoio-a-campanha-de-flavio-bolsonaro/
キャラメルコーンのピーナッツはただのピーナッツじゃねぇぞ
キャラメルコーンの香りがついたピーナッツだ
Путин заявил о формировании нового мироустройства
Президент России Владимир Путин заявил о формировании нового мироустройства в условиях острой конкуренции.
https://russian.rt.com/world/news/1662195-putin-o-formirovanii-miroustroistva?utm_source=rss&utm_medium=rss&utm_campaign=RSS
https://russian.rt.com/world/news/1662195-putin-o-formirovanii-miroustroistva?utm_source=rss&utm_medium=rss&utm_campaign=RSS
„Von dem Islamischen Staat distanziert“: Berliner Jugendrichter ließen sich von CSD-Attentäter Abdul Ballout täuschen
https://apollo-news.net/wp-content/uploads/2026/07/imago862723564-e1785152616427-1024x575.jpg
Der nach seinem Anschlag auf den Christopher Street Day (CSD) in Berlin von der Polizei erschossene Abdul B. hat die ...
The post https://apollo-news.net/von-dem-islamischen-staat-distanziert-berliner-jugendrichter-liessen-sich-von-csd-attentaeter-ballout-taeuschen/ appeared first on https://apollo-news.net.
https://apollo-news.net/von-dem-islamischen-staat-distanziert-berliner-jugendrichter-liessen-sich-von-csd-attentaeter-ballout-taeuschen/
Drie gewonden bij steekpartij in Parijs, dader opgepakt
Bij een steekpartij in Parijs zijn drie vrouwen gewond geraakt. Twee van hen zouden zwaargewond zijn geraakt. Volgens de Franse zender BFM TV zijn ze niet in levensgevaar.
De dader stak de vrouwen met twee messen neer in de wijk Porte de Clichy, zegt de Franse minister van Binnenlandse Zaken. Hij meldt dat de dader is gearresteerd door een politieagent die op dat moment vrij was.
Op beelden op X is te zien hoe een man rondrent met twee messen in zijn hand.
https://images.cdn.nos.nl/1/D/R/L/g/m/DRSwNzx4f7sLFWxkgo4wcwHergJDmFhn4Jtjyqw/8x420x3984x2241-1024x576.webp
https://nos.nl/l/2624607
Detailed view of MIGRATION ASSISTANT?
https://forums.macrumors.com/threads/detailed-view-of-migration-assistant.2481856/
Ein Zeichen für den Frieden
In der EU marschieren alle im Gleichschritt gegen Russland: Von der Politik zur Industrie, von den Kirchen zu den Gewerkschaften und vorneweg die Medien, von der früher mal kritischen Frankfurter Rundschau bis zum französischen Satireblatt „Le Canard enchaîné“, von der FAZ bis zu Le Monde. Es wird höchste Zeit, dagegen ein klares Zeichen setzen.
https://overtonmetrics.de/piwik.php?idsite=1&rec=1&url=https%3A%2F%2Foverton-magazin.de%2Ftop-story%2Fein-zeichen-fuer-den-frieden%2F%3Fpk_campaign%3Dfeed%26pk_kwd%3Dein-zeichen-fuer-den-frieden&action_name=Ein%20Zeichen%20f%C3%BCr%20den%20Frieden%C2%A0%C2%A0%C2%A0%C2%A0%C2%A0&urlref=https%3A%2F%2Foverton-magazin.de%2Ffeed%2F
Der Beitrag https://overton-magazin.de/top-story/ein-zeichen-fuer-den-frieden/?pk_campaign=feed&pk_kwd=ein-zeichen-fuer-den-frieden erschien zuerst auf https://overton-magazin.de.
https://overton-magazin.de/top-story/ein-zeichen-fuer-den-frieden/?pk_campaign=feed&pk_kwd=ein-zeichen-fuer-den-frieden
