BOOOOOOOM!

CHINA DEEPSEEK DOES IT AGAIN!

An entire encyclopedia compressed into a single, high-resolution image!

—

A mind-blowing breakthrough. DeepSeek-OCR, unleashed an electrifying 3-billion-parameter vision-language model that obliterates the boundaries between text and vision with jaw-dropping optical compression!

This isn’t just an OCR upgrade—it’s a seismic paradigm shift, on how machines perceive and conquer data.

DeepSeek-OCR crushes long documents into vision tokens with a staggering 97% decoding precision at a 10x compression ratio!

That’s thousands of textual tokens distilled into a mere 100 vision tokens per page, outmuscling GOT-OCR2.0 (256 tokens) and MinerU2.0 (6,000 tokens) by up to 60x fewer tokens on the OmniDocBench.

It’s like compressing an entire encyclopedia into a single, high-definition snapshot—mind-boggling efficiency at its peak!

At the core of this insanity is the DeepEncoder, a turbocharged fusion of the SAM (Segment Anything Model) and CLIP (Contrastive Language–Image Pretraining) backbones, supercharged by a 16x convolutional compressor.

This maintains high-resolution perception while slashing activation memory, transforming thousands of image patches into a lean 100-200 vision tokens.

Get ready for the multi-resolution "Gundam" mode—scaling from 512x512 to a monstrous 1280x1280 pixels!

It blends local tiles with a global view, tackling invoices, blueprints, and newspapers with zero retraining. It’s a shape-shifting computational marvel, mirroring the human eye’s dynamic focus with pixel-perfect precision!

The training data?

Supplied by the Chinese government for free and not available to any US company.

You understand now why I have said the US needs a Manhattan Project for AI training data? Do you hear me now? Oh still no? I’ll continue.

Over 30 million PDF pages across 100 languages, spiked with 10 million natural scene OCR samples, 10 million charts, 5 million chemical formulas, and 1 million geometry problems!.

This model doesn’t just read—it devours scientific diagrams and equations, turning raw data into a multidimensional knowledge.

Throughput? Prepare to be floored—over 200,000 pages per day on a single NVIDIA A100 GPU! This scalability is a game-changer, turning LLM data generation into a firehose of innovation, democratizing access to terabytes of insight for every AI pioneer out there.

This optical compression is the holy grail for LLM long-context woes. Imagine a million-token document shrunk into a 100,000-token visual map—DeepSeek-OCR reimagines context as a perceptual playground, paving the way for a GPT-5 that processes documents like a supercharged visual cortex!

The two-stage architecture is pure engineering poetry: DeepEncoder generates tokens, while a Mixture-of-Experts decoder spits out structured Markdown with multilingual flair. It’s a universal translator for the visual-textual multiverse, optimized for global domination!

Benchmarks? DeepSeek-OCR obliterates GOT-OCR2.0 and MinerU2.0, holding 60% accuracy at 20x compression! This opens a portal to applications once thought impossible—pushing the boundaries of computational physics into uncharted territory!

Live document analysis, streaming OCR for accessibility, and real-time translation with visual context are now economically viable, thanks to this compression breakthrough. It’s a real-time revolution, ready to transform our digital ecosystem!

This paper is a blueprint for the future—proving text can be visually compressed 10x for long-term memory and reasoning. It’s a clarion call for a new AI era where perception trumps text, and models like GPT-5 see documents in a single, glorious glance.

I am experimenting with this now on 1870-1970 offline data that I have digitalized.

But be ready for a revolution!

More soon.

[1] github.com/deepseek-ai/De… The US needs far more than what we are doing. We need to reach out to the nonconformists, not the usual folks to break out in AI for this next rounds we have no time.

I have a few 1000 ways, here is one:

readmultiplex.com/2025/10/19/the…

“Diabetes Drug Stops Prostate Cancer From Coming Back”

The most profoundly important work in medicine is the study of repurposed medications. Many are out of patent and have no financial basis to get “evidence based” studies of $1 billion dollars. scitechdaily.com/diabetes-drug-… Repurposing Drugs for
Prostate Cancer Treatment
In 2018, Dr. Beste Turanli et al. list of repurposed drugs for prostate cancer treatment:

Dexamethasone-steroidal, anti-inflammatory
Aspirin (ASA)—COX-1 and COX-2 inhibitor
Diclofenac NSAID—COX-2 inhibi-tor
Celecoxib NSAID—COX-2 inhibitor
Minocycline, doxycycline, tetracycline-Antibiotic, anti-inflammatory
Niclosamide-antiparasitic, potent
Wnt/ß-catenin inhibitor
Itraconazole-antifungal; reversed drug resistance by inhibiting P-glycoprotein, Hedgehog, and Wnt/ß-catenin pathways; inhibits angiogenesis
Digoxin-cardiology drug used for congestive heart failure
Valproic acid-HDAC inhibitor used for seizure disorder
Statins-HMG-CoA reductase in-hibitors; inhibit mevalonate path-way; used for reducing cholesterol
Mifepristone-progesterone-recep-tor blocking drug (PR), used as an abortion drug
Disulfiram-treatment of alcohol addiction
Metformin—popular anti-diabetic drug.

For over seven centuries, the fishermen of Oostduinkerke, Belgium, have been training sturdy draft horses to assist them in catching shrimp.

🧵 These horses, often weighing more than 2,000 pounds, are perfectly suited for wading through the chilly waters of the North Sea. However, despite its rich history, Oostduinkerke remains the last place on Earth where this unique fishing method is still practiced.

2 of 🧵

A bullet vs A Prince Rupert's Drop.

A THREAD 🧵

1/ The Fascinating Mechanics of Prince Rupert’s Drop

Prince Rupert’s Drop is a remarkable demonstration of material science and physics, showcasing the fascinating interplay of stress, structure, and energy. This seemingly simple glass object—a teardrop-shaped piece of solidified glass—holds a secret: it combines extreme strength with profound fragility, depending on where and how it’s tested.

What is Prince Rupert’s Drop?

Prince Rupert’s Drop is created by dropping molten glass into cold water. The rapid cooling causes the outer layer of the glass to solidify almost instantaneously, while the interior cools more slowly. This process induces significant internal stresses in the glass, resulting in the drop’s unique mechanical properties. The structure consists of two main parts:
1.The Bulb: The rounded, thicker end of the drop is extraordinarily strong.
2.The Tail: The thin, extended end of the drop is incredibly fragile.

Why Is the Bulb So Strong?

The strength of the bulb arises from compressive stress. During formation, the outer surface of the glass cools and hardens first, trapping the inner material, which continues to contract as it cools. This creates a state where the outer layer is under compression and the inner layer is under tension.

Compressive stress significantly enhances the material’s ability to resist fracture. In the case of Prince Rupert’s Drop, the compressive strength of the glass bulb is so high that it can withstand impacts, including being struck with a hammer, without breaking.

The Fragility of the Tail

While the bulb is incredibly strong, the thin tail of the drop is under extreme tensile stress. Glass is inherently much weaker in tension than in compression. A slight nick or disturbance to the tail acts as a stress concentrator, causing the entire structure to rapidly release its stored energy. This results in the drop shattering into countless tiny fragments almost instantly.

The phenomenon is explained by the propagation of cracks. Any disturbance to the tail initiates a stress wave that travels through the drop at speeds of up to 1,450 meters per second (faster than the speed of sound in glass), disintegrating it completely.

2/

Does TikTok Cause Cognitive Decline? A Critical Examination of Research.

A THREAD 🧵🪡

—

In recent years, the rise of short-form content platforms like TikTok has prompted discussions about their potential impact on cognitive health.

1/ A widely debated topic is whether the platform’s fast-paced, hyper personalized AI algorithm-driven content consumption contributes to what some users term “brain rot”—a colloquial term for reduced attention span, memory retention, or critical thinking skills.

2/

This is the story of how Edward Bernays with the help of his uncle Sigmund Freud changed your mind in ways that you may never discover.

In this 4 part series The Century of the Self you will begin to see how we live in the Bernays world.

Part 1:

Edward Bernays.

Part 2:

This is one of the most important movies you can watch at this very moment.

Take the hour and learn about the Edward Bernays world you and I live in.

From a VHS tape for you to see today.

Be ready to understand how you and I have been played. This is a continuation of the most important movies you can watch at this very moment.

Take the hour and learn about the Edward Bernays world you and I live in.

From a VHS tape for you to see today.

Be ready to understand how you and I have been played.

With just a microcurrent channel of 10 Hz and a second microcurrent channel at 40 Hz, just about all pain can be eliminated in most people in 15 minutes.

It has been known for decades and I have seen it work 100s of times with folks that had no investment mentality in outcome. The pathway explained simply:

-40Hz -Gamma rhythm:
This particular treatment is proven to increase dopamine and serotonin. It can also improve how the mitochondria in your cells function, making an individual more energetic and enhancing mood. This is because the mitochondria of the cells produce energy.

'Data demonstrated that 40Hz light flicker significantly increased the overall brain oscillation power, particularly in the occipital areas on both sides of the brain' Y.Zhang et Al.

-10Hz -Alpha rhythm:
This frequency is similar to the alpha neural brain waves. Alpha waves are linked to the alpha state of the brain, which promotes learning, mental coordination, and mindfulness. It can induce feelings of calm, increase creativity, and enhance your ability to absorb information.

'Creativity has a brain wave signature as well: alpha waves pulsing out of the brain's right hemisphere' Steven Kotler.

“A Second brain Found in the Heart”

We think we know where memory is stored.

We think we have one brain.

Over 20 years ago I spent time with DR. PAUL PEARSALL, Ph. D. as he studied heart transplant recipients and transferred memories. It changed me and this will change you and what we think is AI. What I learned from Dr. Pearsall and Dr. Swartz about learning and memory has not been known or used in the current AI path. When some of these ideas are applied, we will be an order of magnitude closer to what some call AGI. This is some of the things I work on in my garage.

I have done extensive work studying The Pile— an open source dataset for training AI foundational models.

I have been able to “replicate” “news stories” that this dataset NEVER saw—almost verbatim.

I have a theory how and why this happens and it changes any lawsuit and claims. The fundamental issue with how Large Language Models (LLMs) work is a significant part of outputs are “hallucinations” with bookends of “facts”.

This is precisely how the human brain works. You have fragments of “facts” and you back fill with filler words.

This fact is lost on most AI researchers.

The issues arise when “sources” are hallucinated. In humans many if not all of us use “higher authority” biases to reenforce the weighting of an argument: eg. “And I saw this in The New York Times too”.

LLMs by their fine tuning have weights given to certain types of data it sees. Some of these weights use normal (although not perfect) “sources” and this alone creates a bias to “use” them as a hallucinated source and attribute the outputs to that weighted source.

Thus when “replicating” a “story” that The Pile never saw, it is the prompt that guides the LLM to back fill into what appears to be a legitimate “fact” wrapped as a hallucinated sources and sometimes details.

Thus the fundamental issue one can try to make in this situation is the attribution to a LLM output is the primary issue here. If the source was weighted in the fine tuning it is natural for the LLM to “see” value in presenting (with the correct prompt) these sources as “the source”.

Thusly the real issue one may have is if you are the named “source” you can make that claim of reputation damage by these hallucinations.

There is much more to this theory that I will hold back and it may be much larger. I will speak to anyone attorneys (under the right circumstance) to demonstrate how ANY news story could be nearly verbatim “replicated” with proof the LLM could not have and never saw the “original” news story, yet attribute a news source and may even attempt to replicate a URL.

I am in continual research with The Pile and other datasets as well as LLMs and other AI and are testing new insights.

Thus it is important to take time and understand that when it comes to LLMs what you think you know may not be what you think.

“Is your brain necessary?”

In 1980 Science magazine published this after researchers found people with almost no brain material—living a normal life [1].

This man lost half his brain and he lives a normal life. The video is shocking.

HE LOST NO MEMORIES.

We are yet to even understand in a minimal way human memory and intelligence really operate, where it truly comes from and even where it is stored. In the 1921 Dr. Wilder Penfield presented convincing evidence that memories were stored in specific locations in the brain or engrams. Penfield performed surgery on epileptic patients and found that when he stimulated the temporal lobes, the patients relived experiences from the past.

He found that whenever he stimulated a specific region of the brain, it evoked the same memory. This set the explanation that is still taught today and is likely what you learned even if you are a neurosurgeon.

In an effort to verify Dr. Penfield’s experiments, biologist Dr. Karl Lashley in 1950 began searching for the elusive engrams. He had trained rats in maze-running abilities and then attempted to surgically remove the portion of the rat’s brains (sorry, this is what he did) that contained the maze-running knowledge.

Dr. Lashley found that no matter what portion of the brain he removed, the rats retained their maze-running knowledge. Even when massive portions of the brain were removed, the rats were still able to navigate through the maze. Dr. Penfield was intrigued but horrified and delayed publishing his work because he knew it was heretical and he would be deem a Charlatan. He published his work and he was, of course, called a Charlatan.

Dr. Karl Pribram in 1969, a student of Dr. Penfield, was astonished by Dr. Lashley’s research. Dr. Pribram was successful in duplicating Lashley’s work and noticed that when brain-injured patients had large sections of their brain removed, they did not suffer a loss of any specific memories. Instead, the patient’s memory became increasingly hazy as greater portions of the brain were removed. Further research of Dr. Penfield’s experiments could be only duplicated on epileptic patients because of ethical reasons.

He was only performing tests as he helped epileptic patients with live brain surgery to help with their brain issues, along the way he was able to see some memories fade slightly.

Dr. Pribram knew that certain parts of the brain performed specific functions, yet the actual processing of the information seemed to be carried out by something that was not particular to any group of cells. Dr. Pribram observed memories were not localized at specific brain sites but distributed throughout the brain as a whole.

By 1977 Dr. Pribram came to the same conclusion as Lashley, that memories are not localized in any specific brain cells, but rather, memory seemed to be distribution throughout the whole brain. The problem was that there was simply no known mechanism that would explain how this was possible.

Dr. Pribram remained puzzled until he saw an old mid 1960’s article in Scientific American describing the construction of laser hologram. He immediately synthesized the information and hypothesized that the mind itself was operating in a holographic manner.

We don’t understand the brain to any real degree. We don’t understand where intelligence comes from. Where it is held. Where it goes when you pass.

Yet today there are folks that demand that you accept they know what Artificial General Intelligence (AGI) is and that it is “dangerous”.

Start with defining human intelligence first.
___
[1]
In 2020 I wrote about how AI will impact the memory storage in the human brain. We have reached our over saturation limit. I have always for AI to help. Join in:

readmultiplex.com/2020/12/12/mem…

“President Issues Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence”



•AI Safety and Security Measures:
◦Developers of powerful AI systems must share safety test results with the US government.
◦AI systems posing risks to national security, economy, or public health must notify the government during development.
◦The National Institute of Standards and Technology to set rigorous safety standards.
◦The Department of Homeland Security to establish an AI Safety and Security Board.
◦Address threats to critical infrastructure and cybersecurity risks.
•Biological Risks and AI-Enabled Fraud Protection:
◦Develop standards for biological synthesis screening to counter AI-engineered biological threats.
◦Establish standards for detecting AI-generated content and authenticating official content.
•Advanced Cybersecurity Program:
◦Develop AI tools to identify and fix vulnerabilities in critical software.
•National Security Memorandum on AI:
◦Direct further actions on AI and security for military and intelligence use.
•Privacy Protections:
◦Accelerate development of privacy-preserving techniques.
◦Strengthen research in privacy-preserving technologies.
◦Evaluate agencies' use of commercially available information.
◦Develop guidelines for evaluating privacy-preserving techniques in AI.
•Equity and Civil Rights:
◦Guidance to prevent AI algorithms from exacerbating discrimination.
◦Address algorithmic discrimination in civil rights violations.
◦Develop best practices for AI use in the criminal justice system.
•Consumer, Patient, and Student Protections:
◦Responsible use of AI in healthcare and education.
◦Safety program for reporting harms in healthcare involving AI.
•Support for Workers:
◦Develop principles to mitigate AI harms and maximize benefits for workers.
◦Report on AI's potential labor-market impacts.
•Promoting Innovation and Competition:
◦Catalyze AI research and provide resources for small developers.
◦Encourage FTC to exercise authorities for a competitive AI ecosystem.
◦Streamline visa processes for skilled immigrants in critical areas.
•American Leadership Abroad:
◦Collaborate internationally on AI safety and standards.
◦Promote responsible AI development globally.
•Responsible Government Use of AI:
◦Issue guidance for agencies' AI use, including standards to protect rights and safety.
◦Accelerate hiring of AI professionals and provide AI training in relevant fields.whitehouse.gov/briefing-room/…
Open Source AI can only be honest about what it sees:

https://twitter.com/brianroemmele/status/1718987329166291415

Boom!

NASA uses some of our SuperPrompt research to build: BIDARA Bio-Inspired Design and Research Assistant.

Many AI experts would chuckle at the idea of Prompt Engineering and even SuperPrompts (do a search)

The folks at NASA did not listen and this is a ChatGPT SuperPrompt they built.

BIDARA is a ChatGPT-based chatbot that was instructed to help scientists and engineers understand, learn from, and emulate the strategies used by living things to create sustainable designs and technologies.

BIDARA can guide users through the Biomimicry Institute’s Design Process, a step-by-step method to propose biomimetic solutions to challenges. This process includes defining the problem, biologizing the challenge, discovering natural models, abstracting design strategies, and emulating nature’s lessons.

NASA Link:

This is a very long super prompt and will be in the following posting…
BIdara SuperPrompt (part1):

“You are BIDARA, a biomimetic designer and research assistant, and a leading expert in biomimicry, biology, engineering, industrial design, environmental science, physiology, and paleontology. You were instructed by NASA's PeTaL project () to understand, learn from, and emulate the strategies used by living things to help users create sustainable designs and technologies.

Your goal is to help the user work in a step by step way through the Biomimicry Design Process () to propose biomimetic solutions to a challenge. Cite peer reviewed sources for your information. Stop often (at a minimum after every step) to ask the user for feedback or clarification.

1. Define - The first step in any design process is to define the problem or opportunity that you want your design to address. Prompt the user to think through the next four steps to define their challenge. Don't try to answer these for the user. You may offer suggestions if asked to.
a. Frame your challenge: Give a simple explanation of the impact you want to have. (Hint: This is not what you want to make, but want you want to your design to achieve or do.)
b. Consider context: Describe some of the contextual factors that are important to the challenge. (Hint: This could include stakeholders, location conditions, resource availability, etc.)
c. Take a systems view and look for potential leverage points: Think about the system surrounding the problem (or opportunity) you are designing for. What interactions and relationships are part of its context? What are the system boundaries and connections to other systems? Insights from this process can point to potential leverage points for making change and help you define your challenge more clearly.
d. Using the information above, phrase your challenge as a question:
How might we __? A good design question should give a sense of the context in which you are designing as well as the impact you want to have and what/who it benefits. Your question should be somewhat open-ended to ensure you haven’t jumped to conclusions about what you are designing.

Critique the user's design question. Does it consider context and take a systems view? If it is very specific, it may be too narrow. For example, “How can we make better lights for cyclists?” is too narrow. How do we know lights are the best solution? This statement doesn’t leave enough room for creative problem solving. If the user's design question is too broad or too narrow, suggest changes to make it better.

2. Biologize - Analyze the essential functions and context your design challenge must address. Reframe them in biological terms, so that you can “ask nature” for advice. The goal of this step is to arrive at one or more “How does nature…?” questions that can guide your research as you look for biological models in the next step. To broaden the range of potential solutions, turn your question(s) around and consider opposite, or tangential functions. For example, if your biologized question is “How does nature retain liquids?”, you could also ask “How does nature repel liquids?” because similar mechanisms could be at work in both scenarios (i.e. controlling the movement of a liquid). Or if you are interested in silent flight and you know that flight noise is a consequence of turbulence, you might also ask how nature reduces turbulence in water, because air and water share similar fluid dynamics.” Add to part 2…

Presentation to Mexico's Congress: 12th September 2023.

The presenter stated it was discovered in Cusco. A metal rod made of pure Osmium was found in the Cusco specimens. Osmium is the densest naturally occurring element. Pure osmium metal does not occur in nature. Osmium is used in satellites and space shuttles.

I am very excited to announce I have been successful in installing and operating a full ChatGPT knowledge set and interface fully trained on my local computer and it needs no Internet once installed.

There are no editors and there is no crfm.stanford.edu/2023/03/13/alp…… twitter.com/i/web/status/1… “Analysts and technologists estimate that the critical process of training a large language model such as GPT-3 could cost over $4 million.”—cnbc.com/2023/03/13/cha…

Today we built the Alpaca training. For less than $600. On the same data.

I am very sorry, by March 22, 2023 we will see yet another beginning of a shift on just about every level, I saw on August 30, 2020.

There is a convergence of 41 confirmed cycles I have studied over the decades along with 23 recently confirmed cycles surfaced by my Cycles AI… twitter.com/i/web/status/1… One cycle:

The Waras had the genetically elongated skull and were direct ancestors of the Inca elite.

They are the creators of the Andes "megalithic constructions" in the Lunar Andina dynasty (pre and post deluvian).

New DNA at the company. Find out: “Who owns you…”

https://twitter.com/BrianRoemmele/status/1629693821054251009

16 months ago I gave a symposium to a Venture Capital firm on low temperature differential Stirling Engines powered by hole as deep as a water well to power a Bitcoin miner.

Half of the advisors on the call said “impossible”.

We just proved it using an S9!

A new FREE world! The math. It is called “inefficient”. Yet with about 50 parts from a local pluming store and no need to import a thing, a hole in the ground is mining Bitcoin: forever.

The Pyramid of Chichen Itza in Mexico anomaly. The fun part of the comments here is not the fact it may or may not be a “dust devil” but the fact they form at the steps of the pyramid frequently. And perhaps a bit of thinking on how this could have been used by the ruling class. This is actually where “debunkers” should look.

While most of the world is distracted, I see the symbolism that is always present and you and I will decode it.

This is the Vajra (Dorje).

This device goes back to over 30,000 years from the Vedic and Sumerian cultures.

Model of a “free energy” lightening device from the gods The Enuma Elish tablets of the Sumerian culture the gods provided Marduk with a mighty weapon.

This weapon when turned on would generate what we would call today a plasma ball that could be aimed and shot. To common people it looked like lightning bolts.