People have been uploading this video over and over again after an alleged deletion.
Many believe that this particular clip is a proof that indeed Egypt was ‘robbed’.
Do yourself a favor
Stop what you're doing.
This is important.
Even if you don't have a GPU.
Go download one of the latest local models and just keep it in storage.
There may come a time when you can no longer access intelligence freely
12-27B is enough.
Two days ago the US banned Claude Fable 5.
Yesterday China dropped GLM 5.2.
Today GLM 5.2 is #1 on @bridgebench BS at 100.0, and #1 on Reasoning at 42.8, beating Fable 5.
At 1/10th the cost and 300 tokens per second.
You cannot export control your way out of an open source race.
The ban didn't slow China down.
Unban Fable 5.
⚠️Sensitive Content ⚠️
🚨A horrific scene of a child burning in the fire caused by an Israeli airstrike, as an entire family was killed overnight in Gaza.
HORRIFIC: Israeli settler terrorists are attacking the village of Shouqba right now — setting fields and vehicles on fire, storming a soccer field, and beating children.
Residents are pleading for help from neighboring villages.
🇱🇧🇮🇱 Jabal Amel Hospital in Tyre has been severely damaged in an Israeli airstrike.
This is the 5th time this hospital has been hit.Not the 5th hospital in Lebanon. The 5th strike on this one building.
Source: WarFront Witness on Telegram
This video of Israel’s latest war crime is being systematically suppressed by Twitter.
I screen recorded the moment the platform removed 19 retweets from the post — evidence of Western war crimes being targeted by Elon’s algorithm.
Minutes later, it removed 21 more retweets👇
BREAKING: Israel just carried out a massacre in Tyre, completely demolishing three buildings — where many remain trapped inside — and wrecking the Jabal Amel hospital.
Exact casualty figures unknown but thought to be very serious.
More footage below
This #CVPR2026 paper from our research team is trending #1 on @HuggingFace 🤗
Meet LocateAnything: a vision-language detection model that rethinks bounding box prediction. For AI agents and robots, “seeing” is only useful if a model can pinpoint where something is fast enough to act.
Trained on 138M high-quality samples, LocateAnything decodes bounding boxes in parallel instead of one coordinate at a time, improving localization accuracy while dramatically increasing throughput for visual grounding and detection.
Project page: https://t.co/O7JMe8tzFM
Researchers proved that every single elementary function, sin, exp, log, sqrt, comes from one single binary operator.
It is like finding the “God Particle" for calculus.
In computer science, every complex program breaks down to a single logical operator: the NAND gate. It is the fundamental building block of all digital reality.
But for continuous math, physics, engineering, machine learning, we thought we needed a massive toolbox.
Addition. Subtraction. Trigonometry. Logarithms.
Every scientific calculator and neural network has to juggle all of them.
Until today.
But this paper proved that every single mathematical function can be generated by a single, bizarre binary operator.
eml(x,y) = exp(x) - ln(y).
Combine that with the number 1, and you can build everything.
Pi. The square root. Sine and Cosine. Arithmetic.
It is all just the exact same operator, repeating over and over again in a binary tree.
Nobody anticipated this existed. It was found by systematic exhaustive search.
But the implications for AI are massive.
Instead of an AI struggling to combine different mathematical rules to discover a new scientific law, it can just use a single, uniform architecture.
One trainable circuit. One repeatable node.
We thought the language of the universe was complex.
It turns out, it's just one equation repeating in the dark.
Ever wondered who invented robotics?
Ismail al-Jazari, a 12th-century Muslim inventor and engineer, is often called the “father of robotics” for his pioneering work on automata, self-operating mechanical devices powered by water and intricate mechanisms.
His name was Ismail al-Jazari.
A thread on the forgotten father of robotics 🧵
MICROSOFT RESEARCH JUST PUT A FREE DATA ANALYSIS TOOL ONLINE THAT REPLACES $70/MONTH TABLEAU SEATS.
You describe the chart you want. It builds it. No SQL or formulas. No degree required.
30 CHART TYPES. Works on screenshots, CSVs, live databases, and plain text. Zero dollars.
Here is what is going on.
Tableau charges $70 per user per month. Power BI Pro runs $10 to $20 per seat. Excel with Copilot is another subscription on top of that. Data analysis has always been expensive because the tools that make it easy cost serious money.
Microsoft Research just put the alternative online for free.
It is called Data Formulator. You load your data, describe what you want to see in plain language, and AI builds the chart, transforms the data, and writes the code behind it automatically. No dragging pivot tables. No writing SQL. No figuring out why your VLOOKUP broke.
And here is where it gets interesting.
↳ paste a screenshot of a table and it extracts the data automatically
↳ connect it directly to MySQL, PostgreSQL, Azure, S3, or any URL with live refresh
↳ describe a chart in plain English and it figures out what data transformations are needed to make it
↳ agent mode lets it plan and explore your data across multiple turns on its own
↳ build full shareable reports directly inside the tool
↳ runs on OpenAI, Claude, Gemini, or fully local with Ollama
The new version has a unified AI agent that handles everything, a persistent workspace so your data stays organized across sessions, and sandboxed code execution so nothing runs on your machine without permission.
Microsoft sells Power BI to enterprises for real money every month. Their own research team then built a free open source tool that does a large chunk of the same job and put it on GitHub under an MIT license.
Someone at the Power BI team is having a very interesting week.
🚨 : Was Mossad Behind 9/11 Attacks?
A stunning new revelation from Lebanon has revived one of the most controversial questions of the 21st century: Did Israeli intelligence have a role in the 9/11 attacks on the United States?
According to reports, the Lebanese army recently smashed an Israeli spy ring operating in the Bekaa Valley region, near the Syrian border. The operation uncovered a connection that has sent shockwaves through intelligence circles worldwide.
The Jarrah Brothers’ Confession
The spy ring’s operatives have been identified as two brothers – Ali and Youssef al-Jarrah – from the Jlela region. Their nephew is none other than Ziad al-Jarrah, one of the four hijackers who reportedly piloted United Airlines Flight 93, which crashed in Pennsylvania on September 11, 2001.
Under interrogation by Lebanese authorities, the Jarrah brothers allegedly admitted to their links with Mossad, Israel’s national intelligence agency. While details remain murky, the implication is that a close relative of a 9/11 hijacker was actively working for Israeli intelligence – raising inevitable questions about prior knowledge or even coordination.
The “Dancing Israelis” Incident
For years, skeptics of the official 9/11 narrative have pointed to a bizarre episode that the FBI itself documented. On the morning of September 11, 2001, five Israeli citizens (commonly reported as four) were arrested in New Jersey after they were seen taking pictures of the burning Twin Towers from a van, while reportedly dancing and celebrating in apparent jubilation.
The men, later identified as Urban Moving Systems employees, were detained by the FBI but released after brief questioning – reportedly under instructions from the U.S. Justice Department. They promptly left the country. Their behavior has long fueled theories that they may have been Mossad agents documenting the attack.
Goodbye Claude Code subscription fees.
Someone just built a proxy that runs Claude Code completely free... and it's wild.
You literally plug in a free NVIDIA API key and point Claude Code at localhost.
That's it.
It handles everything:
- Converts Anthropic API calls to NVIDIA NIM format
- Unlocks 40 requests/min for free
- Supports Kimi K2, GLM 4.7, MiniMax M2, Devstral and more
- Streams thinking tokens and tool calls live
- Even includes a Telegram bot so you can run Claude Code from your phone
No API bill. No rate limit panic. No vendor lock-in.
Honestly, this goes beyond router tools like OpenRouter.
It doesn't just swap the model... it turns Claude Code into a free agent you can control remotely.
The project is open-source on GitHub.
It's called free-claude-code.
I'm deleting every codebase documentation tool because of this.
Google launched CodeWiki and it turns any GitHub repo into documentation a normal human can actually understand.
You paste a repository and it automatically maps the entire project, explains the architecture, builds diagrams, creates tutorials, and gives you a chatbot that understands the codebase.
The difference from every other AI code explainer is the structure. Most tools summarize files. This turns the whole repo into an interactive wiki you can actually navigate.
→ Generates architecture diagrams automatically
→ Explains what each part of the codebase does
→ Detects dependencies and how files connect
→ Creates step-by-step tutorials from the repo
→ Turns complex systems into readable documentation
→ Lets you ask questions through a repo-aware chatbot
→ Makes onboarding to any codebase feel 10x faster
Basically:
you paste a repo you don't understand.
CodeWiki turns it into something you can read, explore, and ask questions about in minutes.
This is what documentation should have been all along.
Link below 👇
🔥 LLMs just got 8.5× faster — and it breaks the old speed limit 🤯
AI models were always slow because they generate text one token at a time.
Then came speculative decoding: A small model drafts multiple tokens → the big model verifies them together.
Faster, but still limited because drafting was still sequential.
Now comes DFlash ⚡
It removes that bottleneck completely.
Instead of predicting tokens one-by-one, it uses a block diffusion model that generates multiple tokens in parallel in a single step.
No sequence delay. No waiting.
Even better:
Drafting cost stays constant
Uses hidden signals from the main model
Better prediction accuracy
No quality loss
Result: 📊 48.5 tokens/sec → 415 tokens/sec
Already rolling into vLLM, SGLang, Transformers with models like Llama 3.1, Qwen, Kimi, GPT-OSS.
This isn’t just faster AI…
This is a new inference architecture.
Microsoft open-sourced a no-code data analysis tool!
You drag and drop, type what you want, and AI builds the chart, transforms the data, and writes the SQL for you.
→ Works on screenshots, CSVs, and live databases
→ Connect OpenAI, Claude, or run Ollama locally
→ No $70/month seat. No PhD in DAX.
100% open source.
THE ENTIRE AI INDUSTRY JUST GOT HUMILIATED
a tiny model trained in just a few hours on a single graphics card is planning 48x faster than billion-dollar supercomputers.
It actually understands physics instead of just memorizing patterns.
yann lecun was right the whole time
for three years every major lab told you the same story. scale is all you need. just throw more GPUs at it. just train on more tokens. eventually the model will "wake up" and understand the world.
it was a lie. or at minimum, a very expensive bet that just lost.
LeCun kept saying generative AI is a dead end. predicting the next pixel or the next token is fundamentally wasteful, the model burns trillions of parameters memorizing surface details instead of learning how reality actually works.
he proposed JEPA instead. predict abstract concepts in a compressed thought space. don't paint the world pixel by pixel, understand it.
the problem was JEPA kept collapsing. left to its own devices the model would cheat, mapping a dog, a car, and a human to the same point in latent space. technically minimizes the loss. learns absolutely nothing.
every fix was ugly. seven loss terms. frozen encoders. EMA tricks. stop-gradients. the kind of duct-tape engineering that should have been a red flag.
then LeCun's team dropped LeWorldModel.
they replaced all the hacks with one regularizer that forces the latent space into a gaussian distribution. the model can no longer cheat. to make accurate predictions it has to actually encode physics.
15 million parameters. single GPU. trains in hours.
plans 48x faster than foundation world models.
detects physically impossible events on its own.
meanwhile OpenAI is raising another $40B to train GPT-6 on a data center the size of manhattan.
the entire scaling thesis just got embarrassed by a model that fits on a gaming PC.