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
In 2023, over 45% of Indian white-collar workers said AI tools made parts of their job “irrelevant.”
Meanwhile, plumbers, electricians, chefs, and tailors saw zero impact.
"If your job involves moving atoms, you’re safe. If it involves moving digits, AI is coming for it like lightning."
Let’s simplify the rule:
If your job requires holding a tool, you’re safer.
If it only needs a laptop, start preparing.
AI is a digital intelligence.
It doesn’t cook biryani.
It doesn’t fix wiring.
It doesn’t dig roads or repair air conditioners.
It just eats tasks done by digital workers code, design, writing, finance, legal, even medicine.
But what makes India so vulnerable to this shift?
One word: Outsourcing.
A brutal but honest thread:
And in India, this divide is about to reshape the future of your career.
Meanwhile, plumbers, electricians, chefs, and tailors saw zero impact.
"If your job involves moving atoms, you’re safe. If it involves moving digits, AI is coming for it like lightning."
Let’s simplify the rule:
If your job requires holding a tool, you’re safer.
If it only needs a laptop, start preparing.
AI is a digital intelligence.
It doesn’t cook biryani.
It doesn’t fix wiring.
It doesn’t dig roads or repair air conditioners.
It just eats tasks done by digital workers code, design, writing, finance, legal, even medicine.
But what makes India so vulnerable to this shift?
One word: Outsourcing.
A brutal but honest thread:
And in India, this divide is about to reshape the future of your career.
India’s $250 billion IT-BPO industry is built on solving foreign digital problems with Indian digital talent.
That’s the exact playground AI thrives in.
In the next 5 years, it’s not Western companies that will cut you.
It’ll be Western AI models doing your job faster, cheaper, and without chai breaks.
TCS. Infosys. Wipro. HCL.
These giants built their empires on digital labour.
But OpenAI, Anthropic, and Google just built armies that work 24/7, don’t sleep, don’t strike, and never ask for salary hikes.
In 2024, a U.S. insurance firm reduced 30% of back-office staff in India using GPT-4 powered automation.
By 2026, this becomes normal.
That’s the exact playground AI thrives in.
In the next 5 years, it’s not Western companies that will cut you.
It’ll be Western AI models doing your job faster, cheaper, and without chai breaks.
TCS. Infosys. Wipro. HCL.
These giants built their empires on digital labour.
But OpenAI, Anthropic, and Google just built armies that work 24/7, don’t sleep, don’t strike, and never ask for salary hikes.
In 2024, a U.S. insurance firm reduced 30% of back-office staff in India using GPT-4 powered automation.
By 2026, this becomes normal.
Here’s who else should worry:
Entry-level lawyers (contracts are now AI-checked)
Junior copywriters (GPT-4 writes brand campaigns)
Accountants (AI audits 1000x faster)
Data entry clerks (already extinct in many firms)
Freelancers doing resumes, blogs, or SEO
MBAs running Excel models
All of this is becoming a prompt.
But here’s the plot twist:
AI won’t “destroy all jobs.” It will split them into 3 classes:
Creators of AI – Safe, growing, and rich
Supervisors of AI – Need reskilling, but still needed
Victims of AI – Replaced silently, blamed loudly
In India, 80% of the digital workforce is currently in Category 3.
You’re not lazy. You were just never told this was coming.
Entry-level lawyers (contracts are now AI-checked)
Junior copywriters (GPT-4 writes brand campaigns)
Accountants (AI audits 1000x faster)
Data entry clerks (already extinct in many firms)
Freelancers doing resumes, blogs, or SEO
MBAs running Excel models
All of this is becoming a prompt.
But here’s the plot twist:
AI won’t “destroy all jobs.” It will split them into 3 classes:
Creators of AI – Safe, growing, and rich
Supervisors of AI – Need reskilling, but still needed
Victims of AI – Replaced silently, blamed loudly
In India, 80% of the digital workforce is currently in Category 3.
You’re not lazy. You were just never told this was coming.
@kacaanofsade TPS for Somalia will expire in March 2026.
After all, it’s a very small number of people, so according to transgender logic, TPS expiring is no big deal except for people freaking out about it.
After all, it’s a very small number of people, so according to transgender logic, TPS expiring is no big deal except for people freaking out about it.
@kacaanofsade
@kacaanofsade @threadreaderapp please unroll
@StateDept Why a restrictive immigration policy is so important to Republicans and why the ambassador should take over the monitoring of European refugee policy.
@StateDept If the EU should ensure peace in Europe, we will no longer be able to afford the liberal immigration policy. Mr. Trump is aware of this and will apply pressure.
@StateDept Donald Trump is aware that the financing of the planned NATO goal is on shaky ground as long as immense sums are wasted on the failed migration policy, costing the German taxpayer 50 billion euros annually.
I am losing my marriage.
HOW DIVORCE MISSED ME BY A MINUTE
I have been married for 8 years, but to me, life in marriage has never been easy. From the beginning, it always felt like my husband and I could never find a common ground to operate as a couple. Every time l looked at him, I found myself filled with anger. I had already left him mentally, though physically I was still in the same house with him.
HOW DIVORCE MISSED ME BY A MINUTE
I have been married for 8 years, but to me, life in marriage has never been easy. From the beginning, it always felt like my husband and I could never find a common ground to operate as a couple. Every time l looked at him, I found myself filled with anger. I had already left him mentally, though physically I was still in the same house with him.
On June 12th this year, we had one of those long arguments that lasted the whole night. He spoke less, but I ensured I reminded him of all his vices and flaws. I thought l was fighting for my space, but deep inside I was confirming to myself that I had already checked out emotionally.
Even before then, in April, after one of my usual word fights with him, I left for work. That day I was overwhelmed with emotions and couldn't drive, so I took public transport. I quietly took a corner seat to avoid unnecessary movement. I was lost in my thoughts when a young man, probably in his early 30s, came and sat next to me. Though I wasn't in the mood to talk to anyone, his friendly and appealing approach drew me in.
Even before then, in April, after one of my usual word fights with him, I left for work. That day I was overwhelmed with emotions and couldn't drive, so I took public transport. I quietly took a corner seat to avoid unnecessary movement. I was lost in my thoughts when a young man, probably in his early 30s, came and sat next to me. Though I wasn't in the mood to talk to anyone, his friendly and appealing approach drew me in.
To my surprise, I found myself opening up and explaining the struggles I was going through. He listened. After we alighted near my workplace, he walked with me to a nearby restaurant and bought me a drink to calm me down. I couldn't hold back my tears in his presence, but afterward, I felt composed enough to proceed to work.
Before leaving, he requested a ride for me, and we exchanged numbers.
Before leaving, he requested a ride for me, and we exchanged numbers.
Dear @GemsOfBollywood @SanjeevSanskrit @UdayMahurkar I have written a #Thread on a deeply disturbing trend on TV.
Forget #Urduwood the TV industry is no less destroying in whichever way they can the faith of people.
Here goes 1/n
#Anupamaa #FaithAndFiction #Krishna #Dharma
Forget #Urduwood the TV industry is no less destroying in whichever way they can the faith of people.
Here goes 1/n
#Anupamaa #FaithAndFiction #Krishna #Dharma
2. There is a quiet, dangerous hypocrisy playing out in Indian daily soaps today—especially in shows that constantly invoke Lord Krishna, dharma, and goodness, yet consistently reward evil. This contradiction is deeply disturbing.
#Anupamaa #FaithAndFiction #Krishna #Dharma
#Anupamaa #FaithAndFiction #Krishna #Dharma
3. In #Anupamaa, the narrative repeatedly tells viewers: “Have faith, stay good, trust Krishna.”
But what does the story actually demonstrate?
#Anupamaa #FaithAndFiction #Krishna #Dharma
But what does the story actually demonstrate?
#Anupamaa #FaithAndFiction #Krishna #Dharma
#SupremeCourt will today hear the cases relating to India's Got Latent Controversy and the issue of comedians, including @ReheSamay, making jokes on disability
Matter taken up
Prashant Bhushan : I have filed an intervention on behalf of a disabled professor who has repeatedly asked to be part of the public discussion /discourse etc
intellect-3 is here. this model has been my life the last 2 months and fucked my sleep schedule more than once. a huge team effort that im very proud of. some technical details:
we train on top of glm-4.5-air-base, a 106b param moe, and do the entire post-training pipeline in-house on our open-source stack prime-rl+verifiers+hub. this means two sft stages, a general reasoning and agentic phase, followed by a large-scale rl training run
seeing the stack we have been nurturing the past months in action training a 100b+ moe model across 512 gpus with steady eval gains was beautiful. we also learnt an incredible amount about all the little details you have to get right to get a build a truly good model and believe we are now better equipped than ever to iterate on this at rapid pace
Live now on OpenRouter!
INTELLECT-3 pushes the frontier forward by opening up how high quality models are trained. Weights, code, environments, and a detailed writeup are available to all.
INTELLECT-3 pushes the frontier forward by opening up how high quality models are trained. Weights, code, environments, and a detailed writeup are available to all.
Try it now on OpenRouter: openrouter.ai/prime-intellec…
For API users, note that INTELLECT-3’s reasoning should be preserved between turns. Learn more in our docs:
openrouter.ai/docs/guides/be…
openrouter.ai/docs/guides/be…
For weeks this summer I said Travis Kalanick should run Opendoor.
People laughed.
But they missed the point:
Opendoor needed a builder — not a caretaker.
A founder with aggression, velocity, and vision.
The market treated that as non-consensus.
It was actually obvious.
People laughed.
But they missed the point:
Opendoor needed a builder — not a caretaker.
A founder with aggression, velocity, and vision.
The market treated that as non-consensus.
It was actually obvious.
Travis would have been great.
He understands operational violence.
He understands marketplaces.
He understands refusing to die.
The “Travis era” at Uber was pure asymmetric execution — exactly what OPEN needed.
And for a while, I believed he was the only one who could pull it off.
He understands operational violence.
He understands marketplaces.
He understands refusing to die.
The “Travis era” at Uber was pure asymmetric execution — exactly what OPEN needed.
And for a while, I believed he was the only one who could pull it off.
But then something happened the market didn’t predict:
Kaz Nejatian stepped in — and he carries the same founder aggression,
but with a far deeper understanding of what Opendoor can become.
Kaz sees the full platform.
Not just iBuying.
Not just spreads.
The software company inside the real-estate shell.
And he’s executing like someone who has nothing to lose.
Kaz Nejatian stepped in — and he carries the same founder aggression,
but with a far deeper understanding of what Opendoor can become.
Kaz sees the full platform.
Not just iBuying.
Not just spreads.
The software company inside the real-estate shell.
And he’s executing like someone who has nothing to lose.
We are releasing INTELLECT-3: Scaling async RL to 100B+ MoE on our end to end stack
SOTA for its size in math, code, reasoning, science - outpacing bigger models
Fully open source: weights, data, frameworks, envs, evals
Trained on the same stack we're opening up to everyone
SOTA for its size in math, code, reasoning, science - outpacing bigger models
Fully open source: weights, data, frameworks, envs, evals
Trained on the same stack we're opening up to everyone
Blog with all details
primeintellect.ai/blog/intellect…
primeintellect.ai/blog/intellect…
Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack
Achieving state-of-the-art performance for its size across math, code and reasoning
Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more
Achieving state-of-the-art performance for its size across math, code and reasoning
Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more
INTELLECT-3 is a 106B parameter Mixture-of-Experts model trained with both SFT and RL on top of the GLM 4.5 Air Base model.
Both stages, including multiple ablations, were carried out on a 512-GPU H200 cluster over the course of two months.
Both stages, including multiple ablations, were carried out on a 512-GPU H200 cluster over the course of two months.
Our Training Stack
+ PRIME-RL: Our scalable, asynchronous RL trainer
+ Verifiers: Our unified library used for hundreds of envs and evals on the Environments Hub
+ Sandboxes: Custom container infra optimized for agentic RL
+ Compute: Orchestration & observability for 512 H200s
+ PRIME-RL: Our scalable, asynchronous RL trainer
+ Verifiers: Our unified library used for hundreds of envs and evals on the Environments Hub
+ Sandboxes: Custom container infra optimized for agentic RL
+ Compute: Orchestration & observability for 512 H200s
New blogpost, where I implement a tokenizer in C.
Read here: blog.alpindale.net/posts/simd_tik…
TL;DR the tokenizer is faster than OpenAI's official one by up to 18x. Chinese/Japanese text is bottlenecked by the BPE algorithm, but Code sees a speedup of almost 18x.
