
building @PluralityWeb3, recently shipped "AI Context Flow", google it. opinions are personal.
Berlin, Germany
Joined December 2011
- Tweets2.1K
- Following746
- Followers244
- Likes4.6K
Pinned Tweet
I recently had an incredible chat on @grt_iq #podcast diving deep into my personal journey, insights on Deutsche Telekom, our goals, and what drives us in #Web3 at @mms_Blockchain. Excited to share this insightful discussion! Tune in and enjoy it over the weekend!🌐✨🎙️
📢 Now Available! Ep. 147 w/ @mujtabaidrees94, Blockchain Solutions Architect at @mms_Blockchain, an Indexer at @graphprotocol. Listen to this episode for insight into why a web2, telecom giant entered #web3, is an Indexer at The Graph, & much more!
🎧 grtiq.com/grtiq-podcast-147-…
Muji retweeted
Do have a read!
travelsofsamwise.substack.co…
Muji retweeted
An year ago, if somebody would have told me that people will take out 80k+ out of their pockets for a product that my team and I built, I would have laughed it off.
We wanted to build something people need and use regularly. And we gave it our all.
Sometimes, you make an effort and the universe responds.
Exciting things to come.
Shoutout to @AppSumo for bringing the right marketing to the AI Context Flow / @PluralityWeb3
Muji retweeted
trying to explain that men wouldn’t feel threatened or emasculated by feminism if their manhood were rooted in something other than oppressing women
Muji retweeted
Memory (or context) is a very interesting problem in AI, but not for the reason most people think.
This is the first layer that needs to be equally readable by humans and agents (opinions are my own).
I have done an extensive analysis of the current products in the space, some of which are heavily funded like @Mem0 or @supermemory.
Current focus areas include making retrievals faster, hitting benchmarks, expanding the product surface (all valid problems).
But they're all missing the human-agent interaction part.
Right now, memory works like this: you give an context engine your memories as a dump and then it does its thing and start providing your agents the required context.
But keep doing it for a while (in any product of your choice) and then slowly it becomes a junk drawer where you know you put something but where was it, why did you add it, who changed it, poof. Good luck at that point.
This is one of the areas I'm obsessed with. How do we make context engineering as simple as a perfectly organized notion, and still ridiculously easy for agents to retrieve as well.
What I keep wrestling with in my mind these days: Imagine you are an AI-native individual with 100s of documents and dozens of ongoing clients, is a knowledge graph really the right abstraction?
What "context abstraction" would you want if you are working with agents day-in day-out?
He was dying when he wrote the letters.
Throat infection. Kidney disease. Heart failure.
He had lost his voice.
He wrote to Jinnah anyway.
Come back from London. Lead the Muslim League. There is no one else.
Jinnah came back.
Allama Muhammad Iqbal. Sialkot. Born November 9, 1877.
Son of a tailor. Grandson of a Kashmiri shawl trader.
Lower-middle class family in Kashmiri Mohalla, Sialkot.
His father Sheikh Noor Muhammad could not read.
But he told his young son one thing while he watched him read the Quran.
Read it as if it is being revealed directly unto you.
Iqbal never forgot it.
Scotch Mission High School, Sialkot. Government College Lahore.
First class. Top rank in Punjab University.
Then Cambridge. Trinity College.
Then Lincoln's Inn. Called to the Bar in 1908.
Then Munich. Ludwig Maximilian University. PhD in Philosophy.
Dissertation: The Development of Metaphysics in Persia.
He came home a barrister, a philosopher, and a poet.
He had left for Europe as an Indian nationalist.
He returned as something else entirely.
What he saw in Europe changed him.
Western civilization was technically brilliant and morally hollow.
Secular nationalism would swallow minorities whole.
The Muslim could not survive inside it without his own ground to stand on.
1915. Asrar-i-Khudi. Secrets of the Self. In Persian.
The most important philosophical poem written in South Asia in the 20th century.
Khudi. Selfhood. Dynamic personal agency. The opposite of passive resignation.
He told Muslims to rise.
1930. Allahabad. President of the All-India Muslim League session.
He stood before the assembly and said what no one had said directly before.
I would like to see the Punjab, North-West Frontier Province, Sindh and Baluchistan amalgamated into a single state.
That sentence. In that room. In 1930.
Is where Pakistan was first spoken aloud.
Jinnah was in London in semi-retirement when Iqbal began writing to him in 1936.
Letter after letter. Urgent. Clear.
A separate federation of Muslim provinces is the only course.
Why should not the Muslims of North-West India be considered as nations entitled to self-determination.
He was on his sickbed writing these letters.
Voice gone. Body failing.
Jinnah came home in 1934.
The Muslim League was reorganised.
The Lahore Resolution was passed in 1940.
Pakistan came in 1947.
Iqbal died on April 21, 1938.
Nine years before the country he had imagined existed.
70,000 people attended his funeral at the Badshahi Mosque.
He is buried outside its entrance.
In the city he never left.
Jinnah said after his death: He was a friend, guide and philosopher, and during the darkest moments through which the Muslim League had to go, he stood like a rock.
A tailor's son from Sialkot.
Knighted by King George V in 1923.
Buried at the gate of the Badshahi Mosque in Lahore.
Poet of the East. Thinker of Pakistan. Physician of the Ummah.
He did not live to see what he had built.
Allama Muhammad Iqbal.
Pakistan Zindabad. 🇵🇰
Muji retweeted
People ask me why do I use Claude + ChatGPT + Gemini instead of sticking to Claude:
They lead different benchmarks (Opus 5 #1 on SWE-bench, GPT-5.6 #1 on BrowseComp) and they don't train on the same data, Google pays Reddit $60M/yr, Anthropic got sued instead.
The catch is context. Chroma found all 18 frontier models they tested degrade as input grows.
So I keep one memory layer AI Context Flow to sync across all three, below are the saved context from web, Claude chats, ChatGPT chats and Gemini chats. I use @aicf to control what each one actually sees and control my prompt to be precise.
Right model, clean context, no re-explaining.
Muji retweeted
I saw a post on Reddit that said that “The underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth.” And I don’t think I’ve ever seen AI described so incisively.
Muji retweeted
Your team's AIs shouldn't all be starting from scratch.
AI Context Flow gives your whole team one shared memory, so every prompt, every MCP integration, every AI tool pulls from the same context.
No more information drift between the same team.
Watch how 👇
Muji retweeted
There’s no such thing as an ‘Indian ajrak’
The ajrak is from Sindh in Pakistan. a it has nothing to do with India - stop trying to appropriate what has never been yours
Christopher Nolan wearing a custom Indian Ajrakh silk tie, one of India's oldest surviving block-printing traditions, for The Odyssey premiere in NYC. Designed by @AahaanTandon. — ⭑.ᐟ
Muji retweeted
Rochdale grooming gang leader Shabbir Ahmad was raised, groomed, and unfortunately spoilt in the UK..says Pakistan.
Islamabad is not willing to allow the British convict to come to Pakistan.
Pakistan FO spokesman @TahirAndrabi when asked about British pressure to extradiate 73 year old convict to Pakistan who spent 60 years in UK said.. "Regardless of where he was born, the onus lies on where he grew up, was raised, groomed, and unfortunately spoilt. His heinous crimes demand serious introspection rather than a quest to search for extraneous causes. Let me add that the government of Pakistan has no connection whatsoever with this matter."
Shabbir went to UK from Pakistan when he was only 13-year-old so how can Pakistan be held responsible for what he did?
Muji retweeted
Meet @aicf 👋
Your new shortcut for pulling exact context into any prompt.
Some examples:
✅ @aicf pull exact brand colors from my memory
✅ @aicf add exact influencer handles in my prompt
✅ @aicf fetch my memories from these n buckets
Full control, zero guesswork.
See it in action 👇
#AITools #PromptEngineering
Muji retweeted
There's a lot of discussion around "second brain" systems (Andrej Karpathy, Simon Willison, Tiago Forte, etc.), but most knowledge capture still looks like:
Read → Copy → Switch tabs → Paste → Organize → Go back.
Every interruption costs attention. It sounds small, but every context switch breaks your focus.
My workflow is :
Read → Think → Save the important part → Continue reading.
If I'm already reading an article, I should be able to:
• Ask questions about it
• Save the important insights
• Continue reading
check this out:
Muji retweeted
Prompt optimization just got two modes 🚀
⚡ Fast: one click, instant optimization
🎯 Advanced: follow-up questions for full precision
No more choosing between speed and control. Pick what fits your workflow.
Watch it in action 👇
Yeah that’s pretty annoying especially when it was hallucinating a lot and wasted my hours 👀
Muji retweeted
Taiwan solved tax evasion in 1951 with a trick so cheap it should embarrass every tax authority on the planet.
The problem was an all-cash economy full of small shops. A merchant pockets the cash, skips the receipt, and the sale never existed. Auditors can't catch what was never recorded, and hiring enough of them to watch every noodle stand costs more than the missing tax.
So finance chief Ren Xianqun flipped the incentive. Print a lottery number on every receipt. Draw winners every two months on live TV. Top prize today: NT$10 million, about $310K.
Suddenly the customer and the shopkeeper want opposite things. The merchant wants the sale off the books. The customer wants the ticket. And there are millions more customers than merchants. Every transaction now carries a built-in witness demanding the paper trail.
Year one, reported tax revenue jumped 75%, from NT$29 million to NT$51 million. Seventy-five years later, roughly 70% of Taiwanese still play. Convenience stores redeem the smallest NT$200 prizes at the register, so even a coffee receipt feels like a scratch card.
The elegant part is what the audit force costs. The prize pool runs about NT$7 billion a year, roughly $20 million. In exchange, the government gets 23 million unpaid auditors working every checkout line in the country, forever. No inspector general on earth delivers that coverage at that price.
Greece, Italy, Portugal, and Slovakia all copied it. The most effective compliance tool ever built looks like a game, and that's exactly why it works.



































