я Заки Ахмадов

CISSP OSCP ISO27K1,32. Humanitarian, with passion for photography & for beauty, RT ≠ Endorsement

[Jeffrey Emanuel @doodlestein]

DeepSeek just released a pretty shocking new paper. They really buried the lede here by referring to it simply as DeepSeek OCR.

While it’s a very strong OCR model, the purpose of it and the implications of their approach go far beyond what you’d expect of “yet another OCR model.”

Traditionally, vision LLM tokens almost seemed like an afterthought or “bolt on” to the LLM paradigm. And 10k words of English would take up far more space in a multimodal LLM when expressed as intelligible pixels than when expressed as tokens.

So those 10k words may have turned into 15k tokens, or 30k to 60k “visual tokens.” So vision tokens were way less efficient and really only made sense to use for data that couldn’t be effectively conveyed with words.

But that gets inverted now from the ideas in this paper. DeepSeek figured out how to get 10x better compression using vision tokens than with text tokens! So you could theoretically store those 10k words in just 1,500 of their special compressed visual tokens.

This might not be as unexpected as it sounds if you think of how your own mind works. After all, I know that when I’m looking for a part of a book that I’ve already read, I imagine it visually and always remember which side of the book it was on and approximately where on the page it was, which suggests some kind of visual memory representation at work.

Now, it’s not clear how exactly this interacts with the other downstream cognitive functioning of an LLM; can the model reason as intelligently over those compressed visual tokens as it can using regular text tokens? Does it make the model less articulate by forcing it into a more vision-oriented modality?

But you can imagine that, depending on the exact tradeoffs, it could be a very exciting new axis to greatly expand effective context sizes. Especially when combined with DeepSeek’s other recent paper from a couple weeks ago about sparse attention.

For all we know, Google could have already figured out something like this, which could explain why Gemini has such a huge context size and is so good and fast at OCR tasks. If they did, they probably wouldn’t say because it would be viewed as an important trade secret.

But the nice thing about DeepSeek is that they’ve made the entire thing open source and open weights and explained how they did it, so now everyone can try it out and explore.

Even if these tricks make attention more lossy, the potential of getting a frontier LLM with a 10 or 20 million token context window is pretty exciting.

You could basically cram all of a company’s key internal documents into a prompt preamble and cache this with OpenAI and then just add your specific query or prompt on top of that and not have to deal with search tools and still have it be fast and cost-effective.

Or put an entire code base into the context and cache it, and then just keep appending the equivalent of the git diffs as you make changes to the code.

If you’ve ever read stories about the great physicist Hans Bethe, he was known for having vast amounts of random physical facts memorized (like the entire periodic table; boiling points of various substances, etc.) so that he could seamlessly think and compute without ever having to interrupt his flow to look something up in a reference table.

Having vast amounts of task-specific knowledge in your working memory is extremely useful. This seems like a very clever and additive approach to potentially expanding that memory bank by 10x or more.

[я Заки Ахмадов @ahmadtaha]

Scientists claim 'vast city' is hiding below Egypt's Giza Pyramids https://t.co/KxrlrRybGk

[Jay Anderson @TheProjectUnity]

🚨This is the official PRESS RELEASE of the Massive Underground Structure Findings at the Giza Plateau from the Khafre Pyramid Project

Speaker: Nicole Ciccole

[Leonard Murphy Personal Account @lennyism]

The Khafre mystery gets deeper and deeper; here are more updates and resources to explore. https://t.co/AUEvQy2lZs

[Leonard Murphy Personal Account @lennyism]

Perplexity added some validation and additional details:

SAR Scanning of the Khafre Pyramid: Sources and Research Documentation

The team of Corrado Malanga, Armando Mei, Filippo Biondi, and Nicole Ciccolo has indeed released information about Synthetic Aperture Radar (SAR) scanning of the Giza plateau, with a specific focus on the underground structures beneath the Khafre pyramid. This research represents a significant application of satellite technology to archaeological investigation. Below is a comprehensive analysis of the available sources documenting this project.

Primary Source Documentation

Official Press Releases and Announcements

Multiple sources confirm that the team has made official announcements about their findings. On March 15, 2025, they released information summarizing key findings regarding underground structures detected beneath the Khafre Pyramid1. This followed earlier announcements, including a press release from February 2025 where Nicole Ciccolo, serving as head of communications, presented findings about "a vast underground city" beneath the Giza pyramids2.

A subsequent official statement released on February 14, 2025, explicitly identified the research as the "Khafre Research Project SAR Technology" and detailed the collaboration among Professor Corrado Malanga, Filippo Biondi, Armando Mei, and Nicole Ciccolo6. This statement specified that the team used Capella Space satellite data to conduct their analysis, confirming the satellite technology component mentioned in the summary6.

Technical and Scientific Background

The research builds upon earlier work by team members, particularly Filippo Biondi, who has been exploring the application of Synthetic Aperture Radar to pyramid structures. In July 2024, Biondi shared information about using SAR technology to study the "harmonic resonance" of Khafre's Pyramid, providing insight into the methodological approach that likely informed the current findings3.

The technical foundation for this research appears in a 2022 paper titled "Synthetic Aperture Radar Doppler Tomography Reveals Details of Undiscovered High-Resolution Internal Structure of the Great Pyramid of Giza," co-authored by Filippo Biondi7. This paper describes how SAR data can be processed to reveal internal structures within pyramids by analyzing micro-movements generated by background seismic waves, transforming radar signals into phononic information to detect millimetric vibrations17.

Reported Findings

Underground Structures

According to the March 18, 2025 report, the team's analysis revealed extensive underground structures beneath the Khafre Pyramid, including:

Five identical structures near the base of the pyramid, connected by geometric pathways and containing five horizontal levels with sloping roofs1
Eight cylindrical structures resembling vertical wells, surrounded by descending spiral pathways1
Two large cubic structures measuring approximately 80 meters per side at a depth of 648 meters, where the eight cylindrical structures merge1
The entire structure extending approximately two kilometers beneath the surface and continuing beneath all three pyramids of the Giza complex1

Research Methodology

The team has described their methodology as involving:

Processing of SAR data from multiple angles to create 3D reconstructions of internal pyramid structures1
Use of "Doppler tomography," a patented technique developed by Biondi that enables detection of underground structures6
Converting photonic radar information into phonic signals to capture vibrational data within the acoustic band2
Analysis of data from Capella Space satellites6

Conclusion

The sources collectively confirm the core elements of the summary regarding the team's composition, their use of SAR technology, their focus on the Khafre pyramid and the broader Giza plateau, and their application of satellite technology. The most recent findings were announced in March 2025, with earlier press releases in February 2025 establishing the foundation of the project.

The team has scheduled detailed presentations of their findings, with both press conferences and technical conferences planned to share additional data with the scientific community and the public6.

References1 "SAR Scan of Khafre Pyramid Shows Huge Underground Structures." Greg Reese Substack, March 18, 2025.2 "Official press release. Giza Plateau: discovery of a huge city under..." YouTube, February 7, 2025.3 "Filippo Biondi, Ph.D. on LinkedIn: harmonicsar." LinkedIn, July 27, 2024.6 "discovery of a huge city under the Pyramids." YouTube, February 14, 2025.7 "Synthetic Aperture Radar Doppler Tomography Reveals Details of..." arXiv, August 1, 2022.

Citations:
https://t.co/pHbDLIhUps
https://t.co/TA2V4oJcWm
https://t.co/pMFcrE2ixH
https://t.co/1D6P39MFgO
https://t.co/gcqatl7TKP
https://t.co/V6jSWDw8Es
https://t.co/ROP8qGEnvw
https://t.co/10IZjDQJgR
https://t.co/fXSAgQcVqG
https://t.co/8f6ttuV8TZ
https://t.co/No8i8ImOUm
https://t.co/ZLqCEzIwgI
https://t.co/Lc1tKR2bZy
https://t.co/xnJdemRLXH

Answer from Perplexity: https://t.co/XfWI4hOW7Y"

[Global𝕏 @GlobaltrekX]

Pyramid Structure Discovery:

Here's an image created by @CoreyWilken showing just how big these structures are in comparison that have recently been discovered beneath the Pyramid of Giza.

Click on image for Full scale. Absolutely Mind blowing 🤯

[я Заки Ахмадов @ahmadtaha]

Do you remember when you joined X? I do! #MyXAnniversary

[я Заки Ахмадов @ahmadtaha]

BTC = $71,300
🚀 🚀 🔝ATH

[я Заки Ахмадов @ahmadtaha]

Looks like a pursuit from Need For Speed 😁

[Big Nate Andrews..... ♋️ @50shadesofblac4]

He had no chance🫠🫠

[NetBlocks @netblocks]

⚠️ Confirmed: Metrics show a significant disruption to internet connectivity in #Egypt attributed to an incident on the FEA and SEA-ME-WE 4 subsea cable networks, with national connectivity currently at 64% of ordinary levels 📉

[я Заки Ахмадов @ahmadtaha]

Live from Amsterdam, a Palestine support demonstration

[George Galloway @georgegalloway]

The foul allegations of rape have been dropped by the Israeli government. The forty beheaded babies has been downscaled to one dead baby, not beheaded, and killed by persons unknown. Two thirds of Israelis killed on October 7 were military personnel. The killers of the remaining one third are definitively revealed to have been in part the Israeli armed forces themselves. Those with influence who spread the propaganda to the contrary stand exposed as War Criminals and now much blood stains their character for ever. It is a spot which will not out. #Gaza_War #GazaCeasefire #GazaHolocaust #GazaHospital

[Viral Content @_viral_content]

Best act of kindness seen for today 🥰

[Tom London @TomLondon6]

I am Jewish

I don't claim to be any expert on the Jewish religion

But in my 62 years I have picked up a bit

Seems to me that the correct Jewish position right now is to oppose the ongoing GENOCIDE by all means possible

I will go on the march TOMORROW along with MANY Jews

[Black-Jewish Alliance ✊🏿✡ @BlackJewishA]

Over a thousand Jews on the Jewish bloc today, proudly in our kippot and kefiyeh. Some have brought olive branches. This is the biggest it's ever been.

On the streets of London today, Jews stand with Palestine. ✊🏿✡️

[Jeremy McLellan @JeremyMcLellan]

10 Ways To Support Israel 🇮🇱

[~هدوووء~ @DoOodi554]

@jacksonhinkle