AI

An OpenAI model just disproved an 80 year old math conjecture from Paul Erdos, one of the most prolific mathematicians in history.  We're going to SOLVE everything.

Defined as: erf(z) = (2/√π) ∫₀ᶻ e^(-t²) dt The Error Function (erf) represents the integral of the Gaussian distribution and appears everywhere from probability and statistics to heat transfer, quantum mechanics, and signal processing. Here’s its Maclaurin series expansion: erf(z) = (2/√π) Σ [(-1)^n z^(2n+1) / (n! (2n+1))] from n=0 to ∞.

Brain-inspired hardware that reshapes itself in liquid to match different tasks Reservoir computing is one of the most efficient ways to process time-series data. Instead of training a deep recurrent network with millions of weights, you let a physical system with rich internal dynamics do most of the work, and only train a simple readout. The catch: real-world signals live on very different time scales (a heartbeat lasts ~1 s, a vehicle moves in milliseconds, a pedestrian over tens of seconds), and most physical reservoirs have fixed dynamics baked in at fabrication. Once built, you are stuck with one time window. Dongzi Yang and coauthors propose a clever solution: an in-liquido reservoir based on an array of organic electrochemical transistors (OECTs) sharing a global electrolyte. Two design knobs work together. First, the distance between each transistor channel and the shared gate sets a different ionic relaxation time per node, creating a gradient of dynamics across the 4x4 chip (distributed reservoir states). Second, the ion concentration of the shared liquid globally retunes all nodes at once, shifting the entire spectrum of relaxation times by more than two orders of magnitude, from 0.013 s up to 2.3 s. The results back the idea. With only 966 trainable parameters in the readout (3 to 4 orders of magnitude fewer than competing CNNs), the system reaches 93.2% on Fashion-MNIST, 99.87% on MIT-BIH ECG arrhythmia classification, and an NMSE of 0.0081 on Mackey-Glass chaotic prediction. The most striking demo is task-adaptive trajectory prediction: a fast vehicle and a slow pedestrian require different reservoir time scales. A closed-loop pipeline measures a paired-pulse facilitation index on the fly, and if it falls outside the operating window, a microfluidic pump physically changes the electrolyte until the dynamics match the task. The reservoir literally reconfigures its hardware in real time. This hints at how neuromorphic edge devices for biosignal monitoring, wearables, and continuous physiological sensing could be built. A single low-power chip that switches its temporal regime on demand removes the need to fabricate task-specific accelerators, which is exactly the flexibility that clinical diagnostics, drug response monitoring, and biotech pipelines need when handling heterogeneous time-series data. Paper: Yang et al., ACS Nano (2026) — journal license | doi.org/10.1021/acsnan…

🚨 CRISPR Breakthrough: HIV May Be Cut Out of Human Cells Scientists have used CRISPR gene-editing technology to remove HIV genetic material from infected human cells in lab experiments. Instead of just controlling the virus with medication, this method targets HIV’s hidden DNA and attempts to cut it out completely. Since HIV hides inside human DNA, it has been extremely hard to fully eliminate. This research shows a possible way to directly attack the virus at its source. Experts say it’s still early and human trials and safety testing are needed, but it marks an important step toward a possible future cure for HIV. Reference: National Institutes of Health. (n.d.). CRISPR-based approaches for HIV proviral DNA excision. Nature Biotechnology. (n.d.). Gene editing strategies targeting latent HIV reservoirs.

🚨 SCIENTISTS JUST BUILT A GOLD-COATED OPTICAL FIBER THAT HUNTS MICROSCOPIC TARGETS LIKE A TINY VACUUM CLEANER. Instead of waiting for molecules to slowly drift past, this new fiber actively pulls them in at high speed using a clever combination of light and gold coating. The result? Much faster, far more sensitive detection of tiny particles, cells, or biomarkers. Why this matters: It could supercharge everything from medical diagnostics and environmental monitoring to next-generation sensors and quantum-level experiments. The deeper implication is wild: We’re moving from passively observing the microscopic world… to actively engineering tools that reach out and grab it with precision. What happens when our detectors become this smart and this fast? Follow for more frontier physics and future technology.

🥀🪽Azarnoosh Khorasani was 8 years old. Eight. On January 9, she was shot on Rahnamayi Street in Mashhad during the protests. She was not in a war zone. She was a child in the street. Her body was buried in Bam Chenar village, Kalat-e Naderi, while her family was reportedly pressured into silence. An 8-year-old child. #آذرنوش_خراسانی

Peymaneh Nikparvar, just 18 years old and a science student, was shot to death by Islamic regime terrorists in Bandar Abbas on January 8. Her blood matters too.

پخشان عزیزی هستم و در یک قدمی اجرای حکم اعدام. صدایم باشید!

🚨112 DAYS IN DETENTION Shakila Ghasemi, 26, a Bahá’í woman from Kerman, has now been detained for 112 days. She was arrested on February 2, 2026, after security forces searched her home and confiscated electronic devices belonging to her and her family. According to reports, she has spent weeks under IRGC Intelligence detention, was later transferred to Kerman Prison, and still faces legal uncertainty. No formal charges have been publicly announced. A young girl is being kept in prison without due process, while her family is forced to live inside fear and uncertainty. #شکیلا_قاسمی

🚨: Physicists Mathematically Proved That the Universe Is Ultimately Not a Simulation. This discovery questions the simulation hypothesis and reveals that the foundations of the universe exist beyond any algorithmic system. A new study from UBC Okanagan mathematically demonstrates that the universe cannot be simulated. Using Gödel's incompleteness theorem, scientists have found that reality requires an "algorithmically non-computable understanding," which no computation can replicate.

🚨 Cancer Patient’s Recovery Leaves Doctors Speechless A terminal cancer patient reportedly showed major recovery after receiving an anti-parasitic drug, surprising researchers worldwide. Scientists believe the medicine may weaken cancer cells and help the immune system fight tumors more effectively. The treatment is still experimental, and experts say more studies are needed. But this unexpected case is giving new hope that old medicines could one day become powerful cancer treatments. Source: National Cancer Institute. Drug Repurposing for Cancer Treatment.

For the first time in history, a machine is speaking the brain's own language. Not reading it. Not recording it. Speaking it. Northwestern University engineers printed artificial neurons that send electrical signals indistinguishable from biological ones — directly to living brain cells. The paper was published in April 2026. And what it demonstrates has never been done before in the history of neuroscience. Here is the context that makes this significant. Every brain-computer interface that exists today Neuralink, cochlear implants, deep brain stimulation devices works in one direction. It listens. It reads electrical signals from the brain and translates them into digital commands. The brain speaks. The machine records. What Northwestern University built does something categorically different. Engineers at Northwestern University have taken a striking leap toward merging machines with the human brain by printing artificial neurons that can actually communicate with real ones. These flexible, low-cost devices generate lifelike electrical signals indistinguishable from those produced by biological neurons. Indistinguishable. That word matters more than any other in this paper. The biological cells surrounding the printed neurons could not tell the difference between a signal arriving from a real neuron and a signal arriving from the printed device. They responded the same way. They fired the same way. They communicated the same way. The brain was not tricked into accepting a foreign signal. It accepted the signal because it could not identify it as foreign. Here is the architecture that makes this possible. The devices are flexible and low-cost printed on material that conforms to the soft, curved surfaces of biological tissue rather than pressing rigid electronics against delicate neural structures. This flexibility is what allows them to sit alongside neurons without triggering the immune response that causes scar tissue to form around implanted electronics, the primary reason most brain implants degrade over time. Every existing brain implant fights a losing battle against the body's own immune system. The brain treats rigid foreign objects as threats. It walls them off. The implant's signal quality degrades as scar tissue builds. Eventually it stops working. Printed flexible neurons do not trigger that response. They conform. They integrate. The brain accepts them. Here is the medical direction the researchers are focused on. Spinal cord injuries where the biological pathway between brain and body is severed, printed neurons could bridge the gap, speaking the brain's language to neurons that have lost their natural communication partners. Parkinson's disease, where specific neural circuits degrade, the devices could supplement failing biological signals with synthetic ones the brain accepts as its own. Hearing loss, vision loss, and other sensory deficits where the biological input pathway is damaged but the processing regions remain intact. The researchers describe the primary application as restoring function to damaged neural circuits, providing a communication bridge between biological neurons that can no longer reach each other through natural pathways. That is the therapeutic case. It is extraordinary on its own terms. But here is the implication that the researchers have not yet fully addressed because the technology is too early, and the leap is too large, and the ethical framework does not yet exist to contain the conversation. A device that sends signals to neurons that are indistinguishable from biological neural signals is a device that can, in principle, insert information into a brain. Not extract it. Not read it. Insert it. Every previous brain-computer interface crossed one threshold, the brain communicating to the machine. Northwestern University's printed neurons cross the other threshold. The machine communicating to the brain. In the brain's own language. In a way the brain accepts without resistance. The researchers are focused entirely on repairing what is broken. That is the right focus. The technology is early. The gap between a laboratory demonstration and anything approaching what science fiction has imagined is enormous and full of unsolved problems. But the physical principle has been demonstrated. A machine can now speak to biological neurons in a language the neurons accept as their own. That has never been true before April 2026. Everything that follows medically, ethically, philosophically follows from that single demonstrated fact. Northwestern University printed neurons that talk to real brains. The brains answered. Source: Northwestern University · ScienceDaily · April 18, 2026 · (Link in the descriptions)

@SciTechera But very few commercial quantum computers are being sold and so this is mostly lab research

This is huge. Scientists just improved an AI model using a real quantum computer. Researchers connected Meta's Llama 3.1 8-billion-parameter LLM to IBM Quantum System Two, a 156-qubit quantum computer. Instead of training the full AI on quantum hardware, they added tiny quantum modules called “Cayley Unitary Adapters” into the model. Results were surprising 👀 The quantum-enhanced AI achieved a 1.4% perplexity improvement using only ~6,000 extra trainable parameters compared to the model’s 8 BILLION parameters. Researchers also found cases where: • the quantum AI answered correctly • while both the original model and classical adapter versions failed. Even more impressive: In compressed AI models, the quantum system recovered up to 83% of the performance normally lost during compression. This is important because future AI systems need: • smaller models • lower energy usage • faster inference • cheaper deployment Most “quantum AI” studies only use simulations. But this experiment ran on actual IBM quantum hardware, making it one of the first real demonstrations of a hybrid quantum-enhanced large language model.

Science is amazing

Korean scientists just discovered how to reverse colon cancer. They turned cancer to healthy colon cells without chemo, surgery, or radiation. Here’s how it works (& how to cleanse your colon)

@F1sT @DrNeilStone You misspelt it - TACO

@DrNeilStone @DeepLearn007 TDS is real people lol

Trump missed his own son's wedding so he could surrender to the Islamic Republic

@TheSkinMechanik @NiohBerg Then release and un-redact all the files

@DeepLearn007 @NiohBerg There are no Epstein issues, you fucking idiot.

I think I need to stop reading Trump's endless statements. Every time he posts vague mindgame statements praising the regime, hopeless Iranians inside Iran end up on the verge of suicide from the mental exhaustion. What the hell happened to "Help is on the way", Mr President?

La próxima guerra de la IA no será escribir prompts. Será leer el cerebro de los modelos. Dentro de un LLM, los días de la semana viven como un círculo. Los colores como una superficie. La posición de un coche en un mundo simulado, como una cuerda curva. Las redes neuronales hablan inglés. Pero piensan en formas. El que aprenda a leer esas formas primero va a editar y dirigir modelos como si fueran software. El resto seguirá rezando para que el prompt funcione.

Legacy code across large surfaces with multiple devs working on it = legacy of many vulnerabilities

Trump betrayed the Iranian protesters.