Joydeep Banerjee

Creator

[Jyotiraditya M. Scindia @JM_Scindia]

History has been written at the Home of Cricket. 🇮🇳

Team India has conquered Lord's in the first-ever Women's Test played at the iconic venue, outclassing England on their own turf with a performance where every player rose to the occasion.

More than a victory, this is a defining chapter in Indian cricket, one that will inspire generations of young girls to dream big, wear the India whites and believe that no stage is beyond their reach.

Congratulations, Team India. May you continue to script many more glorious chapters for Indian cricket.

[Ben Lang @benln]

https://t.co/dfEjtC0I62

[Chamath Palihapitiya @chamath]

Too many people are directionless, giving room for many forms of “ the answer” to their problems.

Go to the gym.
Don’t go to the gym.
Only do compound exercises.
Focus on HRV.
Only walk but with a vest.
Must sleep 8 hours.
Must not oversleep.
Must not under sleep.
No meat.
All meat.
All veg.
All farm to table.
Organic is a scam.
No alcohol.
No sugar.
Natural sugar only.
No Rx drugs.
Yes Rx drugs.
Yes peptides.
No peptides.
No HRT.
Yes HRT.

At some point, we will all realize that none of these are the answer. You are looking to fill a hole because your current life has made you an NPC playing someone else’s game by their rules. It’s their game, not yours that is making you unhappy.

Find your game and go play it.

You’ll probably be happy as a result, do a bit of everything from the list above and will end up living to whatever you’re supposed to - very, very happily.

[Joydeep Banerjee @penningjoy]

@ofootball__ Thank you !!❤️❤️❤️❤️❤️

[Joydeep Banerjee @penningjoy]

A nation is judged by the dreams fulfilled and the aspirations met of all who call it their home. It’s judged by how it continues to uphold the values of human dignity, openness, and respect. It’s judged by how it can stand tall, against all odds, and inspire the world to reach for the stars.

America, you ace it!!

Be the beacon of hope and dreams for centuries, and centuries to come!!
The world and the humanity needs you more than ever!!

Happy 250th to you!! Happy Birthday, America! Loads of love and respect!!

Thank you so much!!🙏❤️

#independenceday

[Boris Cherny @bcherny]

As engineering, product, design, DS, etc. melt into a new kind of role, I was reflecting on what roles might look like in the future. For example, when I look at the Claude Code team I see what I think is five archetypes:

1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship
2. Builder: quickly turns a prototype/idea into production-grade product/infra
3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales

Many people span across 2 roles, and sometimes 3 roles. I also notice that these roles are not really tied to job function -- eg. across Anthropic, some designers match category 1, some 2, some 3; same for engineers, PM, DS.

A healthy team needs a mix of these, depending on the product:

- A product that is new and pre-PMF needs people that are strong at 1+2+3
- A product that is growing and has found PMF needs 2+3+4 and some 5
- A product that has strong PMF needs 3+4+5 and some 2

Maybe product roles of the future will look more like this, and less like the domain-specific roles of today?

[Joydeep Banerjee @penningjoy]

@bcherny I have skills for Claude to explain its work using ASCII diagrams after it’s done with a session. I am a highly visual person.. helps a lot. Good to see it baked into the harness now

[Brian Armstrong @brian_armstrong]

Good take

My guess is
- demand for intelligence is near infinite
- but 80% of workloads will be running on 99% cheaper models within 12-18 months
- 20% of workloads will still run on latest gen models where IQ maxing is important (scientific breakthroughs, higher level ochestrator agents?)
- rough analogy might be what % of macbooks or gaming PCs sold have the maxed out specs for CPU/GPU, prices are falling much faster than Moore's law here though
- this leads me to think the limiting factor will be energy and compute, not better models

At Coinbase we're working hard on routing prompts to cheaper models where appropriate, and in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially.

[Joydeep Banerjee @penningjoy]

𝗧𝗵𝗲 𝗕𝗶𝘁𝘁𝗲𝗿 𝗟𝗲𝘀𝘀𝗼𝗻

It’s a famous essay from the AI pioneer Rich Sutton. He argued that research and learning trumps clever when it comes to build AI agents.

Build agents that can learn continuously and improve themselves, instead of teaching them every skill meticulously. Give them an aweful lot of data and computational power, and there you have it- an AI that works. No amount of teaching can beat this level of efficiency.

Link to the essay - https://t.co/vAR94VsL9n

#AI

[The Figen @TheFigen_]

How 3 basic lines manage to mesmerize a man’s mind…

[Joydeep Banerjee @penningjoy]

One of the most influential papers in computer science isn’t about AI.

It’s a 1978 paper called:

“Time, Clocks, and the Ordering of Events in a Distributed System” by Leslie Lamport.

Its core idea pertaining to distributed system still powers modern software.

Imagine three servers:

New York.
London.
Tokyo.

Each performs actions and sends messages.

Then, how do you know which event happened first?

Seems easy. But it isn’t. Most people think the answer is:

“Check the clock.”

Lamport realized that’s the wrong question. Why? Because -

Clocks drift.
Networks delay messages.
Machines disagree.

Perfect time is surprisingly hard.

He proposed that we don’t need perfect time. We need agreement on ordering.

Instead of asking:

“What time was it?”

Ask:

“What happened before what?”

This led to Lamport Clocks.

A logical clock doesn’t measure real time. It tracks causality.

This idea became a foundation for:

• Distributed databases
• Kafka
• Paxos
• Raft
• Kubernetes
• Event sourcing

Takeaway:

In distributed systems, agreement on ordering is often more important than agreement on time.

A simple idea. But a profound shift in thinking.

#systemdesign #distributedsystems

[Joydeep Banerjee @penningjoy]

PPO (Proximal Policy Optimization) is one of the most important ideas in modern AI.

The premise is deceptively simple:

When an AI agent succeeds, don’t let it completely reinvent itself after one lucky win.

PPO trains an AI using two components:

1. Actor — chooses actions
2. Critic — evaluates outcomes

If an action produces a high reward, PPO says:

“Good job. Learn from it.”

But it also says:

“Don’t get carried away.”

It uses a technique called clipping to limit how much the agent can change its behavior after each update.

Think of a great coach training an athlete.

One great game doesn’t mean abandoning everything else and repeating the same play forever.

Instead:

Learn aggressively.
Change conservatively.

That’s the secret.

The result is a learning process that is far more stable, reliable, and less prone to chasing bad shortcuts.

Nearly a decade after its introduction, PPO is still widely used in:

🤖 Robotics
🎮 Gaming
🧠 AI Agents
📈 Decision-making systems

A classic paper that every AI engineer should understand:

https://t.co/DYS6w8vq2w

#AI #MachineLearning #ReinforcementLearning #Agents

[Joydeep Banerjee @penningjoy]

💯

[Uday Kotak @udaykotak]

Google which is cash surplus, just announced an additional capital raise of $80 bn.
Google annual profit is $160 bn, last quarter $62 bn, and market cap $4.5 trillion. That is close to total profits and market cap of all Indian listed companies put together.
It’s a wake up call to all companies to invest into the future, whatever the present maybe.
Now that IPL is done and dusted, time for India to focus on business of business.

[Uday Kotak @udaykotak]

Google which is cash surplus, just announced an additional capital raise of $80 bn.
Google annual profit is $160 bn, last quarter $62 bn, and market cap $4.5 trillion. That is close to total profits and market cap of all Indian listed companies put together.
It’s a wake up call to all companies to invest into the future, whatever the present maybe.
Now that IPL is done and dusted, time for India to focus on business of business.

[Joydeep Banerjee @penningjoy]

Wow!! Anybody in their mid-career, take a few and read this please!! 🙏

[Shreyas Doshi @shreyas]

https://t.co/BZ9YpF6DOo

[Joydeep Banerjee @penningjoy]

Now I genuinely believe the real win for personal AI will be to make local models run on personal devices without internet, being device agnostic can be carried over from one device to another if needed ( or can be synced with the private cloud on demand). The models can be used to run local agentic flows and will fine-tune automatically using something like teacher-discriminator framework.

[Joydeep Banerjee @penningjoy]

SkillsOpt with Hermes is amazing!

I have my Claude code and Codex build features throughout the day and then once they are done, they collate their lessons into a doc. Hermes periodically reads this doc and evaluate if the lessons (with scores) learnt have any material value. Any new lesson worth its salt becomes a skill that can then be progressively picked up agents.

#AI

[Chrome for Developers @ChromiumDev]

AI coding agents can write code, but they can't see if it actually works.

Chrome DevTools for agents 1.0 fixes this. The stable release brings powerful browser debugging, emulation, and automated audits to your AI assistants via our Chrome DevTools MCP server.

👁️ Give your agent eyes on the runtime → https://t.co/jw62MSyKE1

#GoogleIO

[Reads with Ravi @readswithravi]

Take the risk.