Thread Reader helps you read and share Twitter threads easily!
I'm @ThreadReaderApp a Twitter bot here to help you read threads more easily. To trigger me, you just have to reply to (or quote) any tweet of the thread you want to unroll and mention me with the "unroll" keyword and I'll send you a link back on Twitter 😀
— Thread Reader App (@threadreaderapp) November 25, 2017
X thread is series of posts by the same author connected with a line!
From any post in the thread, mention us with a keyword "unroll"
@threadreaderapp unroll
Follow @ThreadReaderApp to mention us easily!
Practice here first or read more on our help page!
Recent
1/ Iran and Russia have a partnership treaty, but that doesn't mean Russia should help China in any way, says a Russian commentator. He argues that Iran is merely a "Chinese gas station" and a "situational partner", and Russian support should be limited to "likes and reposts". ⬇️ 
2/ Opinion on the US/Israel/Iran war among Russian commentators and warbloggers has been mixed in recent days, with some arguing for Russia to support Iran, some saying it should condemn Iran for 'going crazy', and others saying Russia should stay out of it.
3/ Russian journalist and politician Andrei Medvedev recently caused controversy when he argued that Iran was only an opportunistic ally of Russia, and has a long-term history of contributing to Russian losses and defeats such as in Afghanistan.
We thought "Tree Search" was the endgame for AI reasoning.
We were wrong.
A new paper (March 2, 2026) just flipped the script.
It proposes a radical new paradigm: "Reasoning as Gradient."
And it just hit SOTA on MLE-Bench with a single GPU.
This changes how we build agents.
We were wrong.
A new paper (March 2, 2026) just flipped the script.
It proposes a radical new paradigm: "Reasoning as Gradient."
And it just hit SOTA on MLE-Bench with a single GPU.
This changes how we build agents.
1. The "Tree Search" Trap
Until now, the logic was:
"To make an agent smarter, let it generate 100 options (Tree of Thoughts) and pick the best one."
It works, but it's inefficient.
It's like trying to find a needle in a haystack by burning the haystack.
Until now, the logic was:
"To make an agent smarter, let it generate 100 options (Tree of Thoughts) and pick the best one."
It works, but it's inefficient.
It's like trying to find a needle in a haystack by burning the haystack.
2. The New Approach: Gome
The paper introduces "Gome" (Gradient-based Optimization for MLE).
Instead of randomly searching for a better answer, it treats reasoning errors like a "gradient."
It looks at why it failed, calculates the "direction" of the fix, and updates the prompt/memory instantly.
The paper introduces "Gome" (Gradient-based Optimization for MLE).
Instead of randomly searching for a better answer, it treats reasoning errors like a "gradient."
It looks at why it failed, calculates the "direction" of the fix, and updates the prompt/memory instantly.
If they will do it to Jon McGreevy, they will do it to you or me, and they will do it to any public official who does not play along. A thread. /1
2/ It starts with Deep Involvement. The target embeds in a covert operation, gains insider knowledge of systemic abuse, and defects to go public.
This is followed immediately by Legal Entrapment—fabricated charges designed to justify initial punishment.
This is followed immediately by Legal Entrapment—fabricated charges designed to justify initial punishment.
3/ Post-release, the whistleblower rises in influence. But mainstream media leverages those prior "legal charges" to poison the well.
While they gain a following, their credibility is systematically chipped away from the outside.
While they gain a following, their credibility is systematically chipped away from the outside.
@cescontos Artes conceituais de Zack Snyder JUSTICE LEAGUE (#1):
instagram.com/p/DLkcT9OMR9K/…
instagram.com/p/DLkcT9OMR9K/…
@cescontos Artes conceituais de Zack Snyder JUSTICE LEAGUE (#2):
instagram.com/reel/DLh3yQ0Oo…
instagram.com/reel/DLh3yQ0Oo…
@cescontos Artes conceituais de Zack Snyder JUSTICE LEAGUE (#3):
instagram.com/p/C3Q2IFtPAXl/…
instagram.com/p/C3Q2IFtPAXl/…
During a job interview, if they ask: “Why should we hire you over other candidates?”
THE GOLDEN RESPONSE: 
THE GOLDEN RESPONSE: 
The “Generic Answer” Trap
In 2026, saying things like “I’m hardworking, passionate, and a quick learner” instantly weakens your position.
Why?
Because every candidate says the same thing.
Hiring managers aren’t looking for effort.
They’re looking for impact.
In 2026, saying things like “I’m hardworking, passionate, and a quick learner” instantly weakens your position.
Why?
Because every candidate says the same thing.
Hiring managers aren’t looking for effort.
They’re looking for impact.
The Psychology of the Answer
The best candidates don’t compete by listing traits.
They compete by showing value.
Instead of sounding like someone asking for a job, they sound like someone who solves problems.
The goal isn’t to prove you’re better.
The goal is to prove you’re useful.
The best candidates don’t compete by listing traits.
They compete by showing value.
Instead of sounding like someone asking for a job, they sound like someone who solves problems.
The goal isn’t to prove you’re better.
The goal is to prove you’re useful.
It's become popular to say that anyone with ADHD or level 1 autism is annoyingly making it up for a "trendy identity." But have you considered that being annoying is part of it? Link in replies.
My ADHD makes me socially inept and then my OCD tells me I did worse, more socially inept things that I don't remember doing.
cartoonshateher.com/p/am-i-neurodi…
cartoonshateher.com/p/am-i-neurodi…
Yann LeCun's (@ylecun ) new paper along with other top researchers proposes a brilliant idea. 🎯
Says that chasing general AI is a mistake and we must build superhuman adaptable specialists instead.
The whole AI industry is obsessed with building machines that can do absolutely everything humans can do.
But this goal is fundamentally flawed because humans are actually highly specialized creatures optimized only for physical survival.
Instead of trying to force one giant model to master every possible task from folding laundry to predicting protein structures, they suggest building expert systems that learn generic knowledge through self-supervised methods.
By using internal world models to understand how things work, these specialized systems can quickly adapt to solve complex problems that human brains simply cannot handle.
This shift means we can stop wasting computing power on human traits and focus on building diverse tools that actually solve hard real-world problems.
So overall the researchers here propose a new target called Superhuman Adaptable Intelligence which focuses strictly on how fast a system learns new skills.
The paper explicitly argues that evolution shaped human intelligence strictly as a specialized tool for physical survival.
The researchers state that nature optimized our brains specifically for tasks necessary to stay alive in the physical world.
They explain that abilities like walking or seeing seem incredibly general to us only because they are absolutely critical for our existence.
The authors point out that humans are actually terrible at cognitive tasks outside this evolutionary comfort zone, like calculating massive mathematical probabilities.
The study highlights how a chess grandmaster only looks intelligent compared to other humans, while modern computers easily crush those human limits.
This proves their central point that humanity suffers from an illusion of generality simply because we cannot perceive our own biological blind spots.
They conclude that building machines to mimic this narrow human survival toolkit is a deeply flawed way to create advanced technology.
Says that chasing general AI is a mistake and we must build superhuman adaptable specialists instead.
The whole AI industry is obsessed with building machines that can do absolutely everything humans can do.
But this goal is fundamentally flawed because humans are actually highly specialized creatures optimized only for physical survival.
Instead of trying to force one giant model to master every possible task from folding laundry to predicting protein structures, they suggest building expert systems that learn generic knowledge through self-supervised methods.
By using internal world models to understand how things work, these specialized systems can quickly adapt to solve complex problems that human brains simply cannot handle.
This shift means we can stop wasting computing power on human traits and focus on building diverse tools that actually solve hard real-world problems.
So overall the researchers here propose a new target called Superhuman Adaptable Intelligence which focuses strictly on how fast a system learns new skills.
The paper explicitly argues that evolution shaped human intelligence strictly as a specialized tool for physical survival.
The researchers state that nature optimized our brains specifically for tasks necessary to stay alive in the physical world.
They explain that abilities like walking or seeing seem incredibly general to us only because they are absolutely critical for our existence.
The authors point out that humans are actually terrible at cognitive tasks outside this evolutionary comfort zone, like calculating massive mathematical probabilities.
The study highlights how a chess grandmaster only looks intelligent compared to other humans, while modern computers easily crush those human limits.
This proves their central point that humanity suffers from an illusion of generality simply because we cannot perceive our own biological blind spots.
They conclude that building machines to mimic this narrow human survival toolkit is a deeply flawed way to create advanced technology.
This visual maps different AI goals to show how adaptable intelligence completely beats older performance ideas.
Traditional targets only focus on copying human jobs.
The new framework prioritizes fast learning across important tasks.
It targets high adaptability over static performance.
Specialized experts easily beat systems mimicking rigid human behavior.
----
Paper arxiv.org/abs/2602.23643…
"AI Must Embrace Specialization via Superhuman Adaptable Intelligence"
Traditional targets only focus on copying human jobs.
The new framework prioritizes fast learning across important tasks.
It targets high adaptability over static performance.
Specialized experts easily beat systems mimicking rigid human behavior.
----
Paper arxiv.org/abs/2602.23643…
"AI Must Embrace Specialization via Superhuman Adaptable Intelligence"
This table shows why the most famous definitions for AGI are totally flawed.
It shows they fail because they wrongly assume human minds are perfectly general, demand impossible computing resources, or simply cannot be tested.
i.e. the industry needs a smarter goal.
It shows they fail because they wrongly assume human minds are perfectly general, demand impossible computing resources, or simply cannot be tested.
i.e. the industry needs a smarter goal.
Please continue supporting my orphaned children. They need you. Your donations are a lifeline for them. Today is the 16th of Ramadan. May God bless you. Share with your friends.✊️🙋♂️🙏🤲🚨📢🚀👩❤️👨 chuffed.org/project/147976…
AlinaCobirlau @BRATZULA @cyynnnnnm @doctor_rahmeh @HKX37 @karim2025215780 @Lin11W @LogicalImaan @marcharles8789 @mynameprada @Normanjam67 @piersmorgan @sahouraxo @piersmorgan @Vossvik7 @owenjonesjourno @AndosMac @Andromeda11711 @the_andrey_x @MahdiBaladi @Jackiew80333500 @Lovetor
@gemmamarierose @GapingAnuss @gemmamarierose @basedgizmo @xenomorphporn @Lilkimfan555 @Stayupstaydown @ApVeng @lostboyshaim @mhdkira @jh_yeo40187
A lot of academics still think AI apps generate fake references to papers that don't exist.
They are living in 2023.
You can easily integrate a database of 280M research papers with Claude and ChatGPT to get answers with references to published papers.
Here's how to do it:
They are living in 2023.
You can easily integrate a database of 280M research papers with Claude and ChatGPT to get answers with references to published papers.
Here's how to do it:
1. Go to scite[.]ai/mcp and select Claude ai
Copy the URL given in the second line. Then click on the blue "Add Scite" button.
This will open Claude Connectors. Type in the name Scite and paste the URL in the "Remote MCP" field.
Then click on "Add."
Copy the URL given in the second line. Then click on the blue "Add Scite" button.
This will open Claude Connectors. Type in the name Scite and paste the URL in the "Remote MCP" field.
Then click on "Add."
2. This will add a Scite Connector to your Claude.
Scroll down a little and you will see a Scite Connector. Click on "Connect" and log in to your Scite account.
Give Claude permission to access Scite. And that's it.
You will need a paid subscription for Scite to do this.
Scroll down a little and you will see a Scite Connector. Click on "Connect" and log in to your Scite account.
Give Claude permission to access Scite. And that's it.
You will need a paid subscription for Scite to do this.
NEW MAP UPDATE - IRAN - MIDDLE EAST 🇮🇷🇮🇱🇺🇸 DAY 6
After 6 days of war, military escalation is increasing. Iran hit 2 new countries, Azerbaijan 🇦🇿 and Turkiye 🇹🇷while massive airstrikes continue.
Europeans 🇫🇷🇬🇧🇮🇹🇪🇸🇩🇪🇳🇱🇬🇷 are sending reinforcements to Cyprus.
🧵THREAD🧵1/20 ⬇️
After 6 days of war, military escalation is increasing. Iran hit 2 new countries, Azerbaijan 🇦🇿 and Turkiye 🇹🇷while massive airstrikes continue.
Europeans 🇫🇷🇬🇧🇮🇹🇪🇸🇩🇪🇳🇱🇬🇷 are sending reinforcements to Cyprus.
🧵THREAD🧵1/20 ⬇️
This morning, yet another country sustained at least 2 drone strikes. This time, Iran attacked Azerbaijani 🇦🇿 enclave of Nakhitchevan.
At least two drones impacted the city and an airport. Yesterday, missiles were destroyed 2 times while en route for the Incirlik base in Turkiye.
At least two drones impacted the city and an airport. Yesterday, missiles were destroyed 2 times while en route for the Incirlik base in Turkiye.
Since the start of the Israelo-American operation to topple the regime in Tehran, thousands of air and missile strikes happened in Iran.
After gaining total control of the air, US and Israeli fighter jets are constantly flying in from the Syrian and Irakian territory.
After gaining total control of the air, US and Israeli fighter jets are constantly flying in from the Syrian and Irakian territory.
🧵 This ancient Indian carving explains KARMA better than modern philosophy ever could.
Look closely.
Every knot is connected.
Nothing stands alone 👇
Look closely.
Every knot is connected.
Nothing stands alone 👇
1️⃣ In Sanatan Dharma, karma is not punishment.
It is connection.
Every action ties a knot.
Every intention strengthens it.
You don’t escape karma.
You move within it.
It is connection.
Every action ties a knot.
Every intention strengthens it.
You don’t escape karma.
You move within it.
2️⃣ See how every loop touches another?
That’s life.
Your words affect someone.
Their reaction affects another.
And the cycle continues.
Nothing is isolated.
That’s life.
Your words affect someone.
Their reaction affects another.
And the cycle continues.
Nothing is isolated.
I torched 37lbs of fat in 19 weeks by changing one thing.
Nutrition.
This is exactly what I ate every single day:
= Thread =
Nutrition.
This is exactly what I ate every single day:
= Thread =
Struggling to drop fat?
Learn this:
Nutrition is the key to fat loss.
→ Stop doing hours of cardio.
→ Start dialing in your nutrition.
It's how I lost 37lbs of fat in 19 weeks.
After 17 years simplicity always scales.
Below is exactly what I ate every day:
Learn this:
Nutrition is the key to fat loss.
→ Stop doing hours of cardio.
→ Start dialing in your nutrition.
It's how I lost 37lbs of fat in 19 weeks.
After 17 years simplicity always scales.
Below is exactly what I ate every day:
1. Breakfast:
Low Fat Greek Yogurt + Whey Protein:
⚬ 450g Low Fat Greek Yogurt
⚬ 1 Scoop Chocolate Whey Protein
⚬ 100g Frozen Blueberries
+One protein shake (whey, creatine, water)
Protein: 92g
Carbohydrate: 27g
Fat: 5g
*Elite breakfast, takes 30 seconds
Low Fat Greek Yogurt + Whey Protein:
⚬ 450g Low Fat Greek Yogurt
⚬ 1 Scoop Chocolate Whey Protein
⚬ 100g Frozen Blueberries
+One protein shake (whey, creatine, water)
Protein: 92g
Carbohydrate: 27g
Fat: 5g
*Elite breakfast, takes 30 seconds
