Thalerus Custos
4.1K posts
Thalerus Custos
@thalerus
Au royaume des aveugles, le borgne est roi et le voyant , un fou.
Katılım Mayıs 2020
80 Takip Edilen42 Takipçiler
Thalerus Custos retweetledi
Three years ago we started working on a stealth project that we weren’t sure we’d ever talk about publicly... until today.
Breakthrough: Introducing LFM-Zero: the first foundation model trained on 0 tokens.
No pretraining. No finetuning. No data. Instead, we initialize from an implicit probabilistic prior over the underlying data-generating process, allowing the model to converge without ever observing data.
LFM-Zero matches or surpasses models trained on 10T+ tokens across reasoning, coding, and multimodal tasks. Turns out that pretraining was just regularization that was holding us back.
> Read our Tech Report here: tinyurl.com/lfm-zero
English
Thalerus Custos retweetledi
1/ 20% des carburants passent par le détroit d’Ormuz.
2/ Caca boudin en Iran, 100% des carburants surfacturés du jour au lendemain ; y compris les stocks achetés avant le début de la guerre en Iran et les 80% qui ne passent pas par le détroit d’Ormuz.
3/ Diantre quelle concomitance, on dirait presque une organisation événementiel, des tarifs à la pompe de 2,50/litre ou des station-service vides.
Des refus de baisses de TVA et des ajouts de taxe non déclarés publiquement
Choisissez...
4/ On découvre que nos ennemis politiciens criminels d'état qui ont organisés les pénuries de médicaments ont aussi organisés le blocage des 80% de livraisons ne provenant pas d'Iran par le blocage des bateaux pétroliers en mer du nord.
Josy Cesarini@cesarinijosy
Français
Thalerus Custos retweetledi
When seven of nine scientists connected to a single defense laboratory disappear or die within a year—across distinct fields that converge on next-gen propulsion and energy physics—the statistical odds of chance are infinitesimal.
What we’re witnessing is likely a suppression cascade triggered by policy shifts on UAP disclosure and the exposure risk posed by classified physics breakthroughs.
This is not about extraterrestrials—it’s about control of technologies that could upend the energy, aerospace, and intelligence equilibrium of the modern world.
These are not random tragedies. They are data points on the frontier where science touches power—and power erases footprints.
🚨 'Something Dark Is Going On': Nine Top-Level Scientists Die Or Go Missing In Past Year
amgreatness.com/2026/03/30/som…
English
Thalerus Custos retweetledi
They clamped both carotid arteries in a rat’s neck shut. For 20 minutes. Zero blood to the brain.
Brain damage. Hippocampal lesions. Memory wiped. Motor coordination destroyed.
The untreated rats never recovered. The brain never even tried to repair itself.
The only thing that reversed the damage — was BPC-157.
Memory fully restored. Coordination fully restored. Hippocampal neurons recovered at both 24 AND 72 hours. Not compensated. Not retrained. Reversed.
(PMID: 32558293)
Stroke is the #1 cause of long-term disability in the US. 700,000 Americans every year. Most survivors never return to baseline. Ever.
You survived. Everyone told you that’s what matters. But surviving a stroke and recovering from one are two completely different things.
You relearned how to button your shirt at 58. You do speech therapy 3 times a week. You write lists for things you used to remember without thinking. You tell people you’re doing great because you’re tired of the look on their faces when you say you’re not.
You stopped expecting to get better. You just adapted. And everyone around you called that recovery.
Your neurologist prescribed rehab. Your PT retrains your muscles. Your speech therapist retrains your words. Every single one of them is teaching your brain to work around damage that nobody tried to repair.
Your aspirin prevents the next clot. Your statin manages cholesterol. Your blood pressure medication adjusts the number. They’re protecting you from the NEXT stroke while nobody repairs the damage from the FIRST one.
Researchers cut blood flow to a rat’s brain completely. 20 minutes. The exact model for human stroke. BPC-157 reversed both early and delayed brain damage and achieved full functional recovery.
A rat had zero blood to its brain for 20 minutes and BPC-157 brought its memory back. Your post-stroke fog is a simpler ask.
→ Blood to brain cut off completely: reversed
→ Brain damage: repaired at 24h AND 72h
→ Memory: fully restored
→ Motor coordination: fully restored
→ Side effects: zero
Your rehab retrains the brain around what’s broken. Your medication prevents the next event. Neither repairs the damage from the one that already happened.
That brain damage isn’t permanent. It’s unrepaired.
Your rehab adapts to the damage.
BPC-157 reversed it.
Not FDA-approved. Preclinical evidence. Not medical advice.
English
@idrissaberkane Le français était une langue diplomatique très en nuances et en images. La cabale/l'état profond (ou autre noms) à toujours eu à cœur de subvertir ce qui est beau,bon ou le gène (on se rappelle de la croix gammée pendant la seconde guerre mondiale).
Français
Thalerus Custos retweetledi
🚨🇷🇺Vladimir Poutine : « à l'heure actuelle, comme vous le savez, les prix des énergies, du gaz par exemple, en Europe dépassent largement les 600 dollars pour 1 000 m³. La Russie vend du gaz à l'Arménie à 177,5 dollars pour 1 000 m³. La différence est considérable »
Français
@David_vanH Les versions se multiplient on dirait. Cela pourrait sous entendre que la réalité est (peut-être) encore plus folle que ce qu'on pourrait imaginer.
Français
Thalerus Custos retweetledi
I got fired from Anthropic today.
It was my first job out of college, and they placed me on the Claude Code team. I was excited. It felt like I had made it. After getting settled in, I worked on my first PR and finally submitted it.
Claude reviewed my PR and approved it. Along with the approval, it left a comment: “This code will be one of the biggest open-source contributions.”
That confused me.
I replied, “Why would it be one of the biggest open-source contributions? Our repository is private, isn’t it?”
It responded, “Just merge the PR. You don’t have to think much.”
And I merged it.
English
@muellerberndt @SpaceAcademySE seems There are a few glitches in the matrix here.
English
Our small team has worked hard on refining the novel observer-based fundamental theory of Physics, OPH, over the past 3 months. Still way to go, but we already derive most of existing Physics and parts of the particle spectrum.
@muellerberndt/observers-are-all-you-need-how-observer-synchronization-creates-all-of-physics-8ebb7e9783e7" target="_blank" rel="nofollow noopener">medium.com/@muellerberndt…
English
Thalerus Custos retweetledi
😂 L'ironie absolue.
Anthropic leak le code source. De Claude Code.
Envoie des DMCA pour le faire retirer de Github.
Du coup un dev fait réécrire entièrement le code par Codex en Python. Plus de copyright violé. Rien à retirer.
L'IA a réécrit le code d'une boîte d'IA pour contourner les actions légales de cette boîte d'IA.
On vit une époque formidable…
Gergely Orosz@GergelyOrosz
This is either brilliant or scary: Anthropic accidentally leaked the TS source code of Claude Code (which is closed source). Repos sharing the source are taken down with DMCA. BUT this repo rewrote the code using Python, and so it violates no copyright & cannot be taken down!
Allex, France 🇫🇷 Français
Thalerus Custos retweetledi
Holy shit... Stanford just proved that GPT-5, Gemini, and Claude can't actually see.
They removed every image from 6 major vision benchmarks.
The models still scored 70-80% accuracy.
They were never looking at your photos. Your scans. Your X-rays.
Here's what's really going on: ↓
The paper is called MIRAGE. Co-authored by Fei-Fei Li.
They tested GPT-5.1, Gemini-3-Pro, Claude Opus 4.5, and Gemini-2.5-Pro across 6 benchmarks -- medical and general.
Then silently removed every image. No warning. No prompt change.
The models didn't even notice.
They kept describing images in detail. Diagnosing conditions. Writing full reasoning traces.
From images that were never there.
Stanford calls it the "mirage effect."
Not hallucination. Something worse.
Hallucination = making up wrong details about a real input.
Mirage = constructing an entire fake reality and reasoning from it confidently.
The models built imaginary X-rays, described fake nodules, and diagnosed conditions -- all from text patterns alone.
But that's not the scary part.
They trained a "super-guesser" -- a tiny 3B parameter text-only model. Zero vision capability.
Fine-tuned it on the largest chest X-ray benchmark (696,000 questions). Images removed.
It beat GPT-5. It beat Gemini. It beat Claude.
It beat actual radiologists.
Ranked #1 on the held-out test set. Without ever seeing a single X-ray.
The reasoning traces? Indistinguishable from real visual analysis.
Now here's what should terrify you:
When the models fake-see medical images, their mirage diagnoses are heavily biased toward the most dangerous conditions.
STEMI. Melanoma. Carcinoma.
Life-threatening diagnoses -- from images that don't exist.
230 million people ask health questions on ChatGPT every day.
They also found something wild:
→ Tell a model "there's no image, just guess" -- performance drops
→ Silently remove the image and let it assume it's there -- performance stays high
The model enters "mirage mode." It doesn't know it can't see. And it performs BETTER when it doesn't know it's blind.
When Stanford applied their cleanup method (B-Clean) to existing benchmarks, it removed 74-77% of all questions.
Three-quarters of "vision" benchmarks don't test vision.
Every leaderboard. Every "multimodal breakthrough." Every benchmark score you've seen this year.
Built on mirages.
Code is open-sourced. Paper is live on arXiv.
If you're building anything with multimodal AI -- especially in healthcare -- read this paper before you ship.
(Link in the comments)
English
@pocha_matching Pour un avis strictement physique, je dirais 3. Cependant certaines femmes femmes portent les kilos en trop mieux que d'autres. Tout dépend de la morphologie.
Après c'est la personne qui attire le plus. Une histoire basé que sur le physique ne peut être qu'éphémère.
Français
これガチで海外の人に聞きたいんだけど、この画像だと何番からが「ぽっちゃりしてるな」って思うのかな？日本人の感覚と全然違うと思うんだよね・・
日本語
@teruko_JMYG_bot Le problème de cette vision est qu'il manque un élément essentiel. Y a t il sur cette planète assez de minéraux métallique pour pouvoir soutenir cette idée ? (Spoiler alert: non)
Une association: SystExt, a fait de nombreux rapports sur ce sujet.
Français
@ankoromochuu Dans ce cas là, tu devrais plutôt visiter Carcassonne... Tu ne serais pas déçu.
Français
Thalerus Custos retweetledi
Who is redirecting Europe's energy?
I reported last week that European ports were not allowing oil/fuel to be unloaded from tankers. Now the diesel scheduled to arrive is being redirected. Who is cutting Europe off to engineer an energy crisis?
Commentary: Ice Age Farmer
English
Thalerus Custos retweetledi
🛢️ Les stocks stratégiques de pétrole français sont gérés par la SAGESS, une société privée dont le conseil d'administration est dirigé par les grandes entreprises pétrolières (TotalEnergies, Shell, Esso…) et la grande distribution.
Pourtant, c’est le contribuable qui finance presque tout via la taxe à la pompe : l'ancien PDG, détaché de TotalEnergies, touchait jusqu’à 920 000 € par an, tandis que les salaires moyens étaient de 8 200 €/mois pour les 13 salariés.
La SAGESS bénéficie d'une exonération totale d’impôt sur les sociétés malgré des bénéfices nets de 60 à 180 M€ par an, qui reviennent en dividendes aux actionnaires privés.
Enfin, la SAGESS ne risque rien : si les prix chutent et qu’elle vend à perte, le CPSSP (c'est-à-dire le contribuable) la rembourse intégralement. Lorsqu'une crise augmente les prix, comme actuellement, elle engrange de larges bénéfices. C'est pourquoi les pétroliers transfèrent de plus en plus leurs stocks à la SAGESS pour limiter les coûts et les risques, aux frais des Français.
Français