Madhusudan

Electrical engineer who understands legacy code and makes them live,Slightly overweight but looks great in pics.Fascinated by evolution of C++.

[phone]

[Madhusudan @sahnnu]

I just published Beheading Headers https://t.co/dY9z7b9cCx

My understanding of Modules.

[Miro Cernetig @MiroC]

Grok estimates the brain can process the equivalent of 43,500 token a day.

We are plateaued!

Grok speaks: “Estimating the daily output of the human brain in terms of "tokens" as used in Large Language Models (LLMs) like myself involves several assumptions and simplifications:

1. **Definition of a Token:** In LLMs, a token roughly corresponds to a word or a punctuation mark, or sometimes a piece of a word for more complex terms. For simplicity, let's equate one token to one word.

2. **Human Speech Output:** An average person speaks around 7,000 to 20,000 words per day. Let's take the midpoint for our calculation, which is 13,500 words.

3. **Internal Thoughts:** The brain's activity isn't just limited to spoken words. Internal thoughts, which could be considered as "processing" or "output" in a computational sense, are harder to quantify but can be significantly higher. Some studies suggest we might "think" up to 50,000 or even 70,000 words per day if we include all forms of internal dialogue and thought.

4. **Non-verbal Communication:** Gestures, facial expressions, and other non-verbal cues also convey information, but translating these into tokens would be speculative. For this calculation, we'll stick to verbal and thought "tokens."

Putting this together:

- **Spoken Words:** Approximately 13,500 tokens per day.
- **Internal Thoughts:** Let's conservatively estimate around 30,000 tokens (considering not all thoughts are fully verbalized).

**Total Daily Output:**
- 13,500 (spoken) + 30,000 (thoughts) = 43,500 tokens per day.

However, this is a very rough estimate because:

- **Complexity of Thought:** Human thoughts are not linear like text input to an LLM. They involve complex, multi-layered processing where one thought might lead to several others simultaneously.
- **Multimodal Nature:** The brain processes and outputs in ways that go beyond simple tokens, including sensory data and emotional states.

In comparison:

- An LLM like myself processes or generates tokens based on prompts or tasks. My daily "output" would depend on usage, but for context, a single session with a user might involve generating thousands of tokens. Over a day, with multiple users, this could easily reach or exceed the human estimate, but it's a very different kind of "output" since it's not continuous thought but rather on-demand generation based on queries.

Keep in mind, this comparison is highly metaphorical and somewhat speculative. The human brain's "output" isn't just in tokens but involves a vast, interconnected network of biological processes far beyond what current AI models can replicate or even fully measure in terms of computational analogy.”

[Akshay Shinde @ConsciousRide]

90% of Microservices interviews in 2026 are just these 12 concepts repeated:

[Madhusudan @sahnnu]

Optimistic locking - >buy some more time . Analogy is Double checked locking singleton

[avrl ☘ @avrldotdev]

System Design Talks 30 - Database Locking
[Walmart System Design Interview]

Problem
On Black Friday sale, 2 users click 'Buy' on the very last item at the exact same millisecond.
If you use standard DB updates, you may sell 2 units for 1 item.
How'd you handle this at scale?

[avrl ☘ @avrldotdev]

System Design Talks 30 - Database Locking
[Walmart System Design Interview]

Problem
On Black Friday sale, 2 users click 'Buy' on the very last item at the exact same millisecond.
If you use standard DB updates, you may sell 2 units for 1 item.
How'd you handle this at scale?

[Dave Farley @davefarley77]

Are we writing fiction, or are we practising engineering? If you know my work, you know I generally prefer data and science over fiction. But @RealGeneKim's fantastic novel, The Unicorn Project, made me realise something...

🧵 1/9

[Madhusudan @sahnnu]

@avrldotdev Optimistic locking - >buy some more time . Analogy is Double checked locking singleton

[Abhishek Singh @0xlelouch_]

93.5% of Senior Backend Engineer interviews are just these 7 concepts repeated again and again.

[Kevin Markham @justmarkham]

In-depth intro to #machinelearning in 15 hours of video by the authors of "Elements of Statistical Learning": http://t.co/Id0ea1UTV0

[Krishna Agrawal @Krishnasagrawal]

Stop wasting hours trying to learn AI. 📘📚

I have already done it for you.

With one list. Zero confusion. And no fluff

📹 Videos:

1. LLM Introduction: https://t.co/Avq3pNZxWY

2. LLMs from Scratch: https://t.co/nGJGCQYi89

3. Agentic AI Overview (Stanford): https://t.co/1JbA2JypnJ

4. Building and Evaluating Agents: https://t.co/02o8b7RAS2

5. Building Effective Agents: https://t.co/Jw0cd23A3K

6. Building Agents with MCP: https://t.co/9uvcpin1j9

7. Building an Agent from Scratch: https://t.co/QiYCLK48tn

8. Philo Agents: https://t.co/TCSX4hnKRf

🗂️ Repos

1. GenAI Agents: https://t.co/VdfLWAA3wW

2. Microsoft's AI Agents for Beginners: https://t.co/3SpFdnMcQa

3. Prompt Engineering Guide: https://t.co/4W7Eh6NKaE

4. Hands-On Large Language Models: https://t.co/LEtKGYBdgU

5. AI Agents for Beginners: https://t.co/3SpFdnMcQa

6. GenAI Agentshttps://lnkd.in/dEt72MEy

7. Made with ML: https://t.co/sL3gbmGUky

8. Hands-On AI Engineering:https://t.co/J79go1Ivlo

9. Awesome Generative AI Guide: https://t.co/xXF08rY8zP

10. Designing Machine Learning Systems: https://t.co/Q39XZLn50b

11. Machine Learning for Beginners from Microsoft: https://t.co/vfkyHL1Gnx

12. LLM Course: https://t.co/wuDkhQ0572

🗺️ Guides

1. Google's Agent Whitepaper: https://t.co/BSjthrG81u

2. Google's Agent Companion: https://t.co/wdzh5zuSvp

3. Building Effective Agents by Anthropic: https://t.co/HmYVgMIA7l.

4. Claude Code Best Agentic Coding practices: https://t.co/H3JpJWlpp3

5. OpenAI's Practical Guide to Building Agents: https://t.co/G1TL1Z2TR3

📚Books:

1. Understanding Deep Learning: https://t.co/EEqkGL7lHe

2. Building an LLM from Scratch: https://t.co/8Ehn91NNxE

3. The LLM Engineering Handbook: https://t.co/yzvXeQAgtV

4. AI Agents: The Definitive Guide - Nicole Koenigstein: https://t.co/FKEzpBewe4

5. Building Applications with AI Agents - Michael Albada: https://t.co/nvBaoR6FxZ

6. AI Agents with MCP - Kyle Stratis: https://t.co/u0M1GfBlsL

7. AI Engineering: https://t.co/A2QxhrpmDc

📜 Papers

1. ReAct: https://t.co/Vbc3vWxQHO

2. Generative Agents: https://t.co/PvNHYnqsOD.

3. Toolformer: https://t.co/HFdds582DI

4. Chain-of-Thought Prompting: https://t.co/GxX5BxXLzV.

🧑🏫 Courses:

1. HuggingFace's Agent Course: https://t.co/zhizqjunzS

2. MCP with Anthropic: https://t.co/LqDZ0oCRfj

3. Building Vector Databases with Pinecone: https://t.co/ncqSVaVHqH

4. Vector Databases from Embeddings to Apps: https://t.co/JOl7SIlguq

5. Agent Memory: https://t.co/5m0UrGwDM9

Repost for your network ♻️
&follow for more stuff on building AI Agents.

[Suraj Kumar Talreja @suritalreja]

India 🇮🇳 gets just ~1 crore foreign tourists a year.

That’s less than Bangkok (3.2 Crore) alone gets.

And this, in a country with:

🕌 40 UNESCO sites
🏔️ The Himalayas
🌴 Goa, Kerala and Andaman beaches
🍛 The best food on Earth
🧘‍♂️ Yoga, Ayurveda, Rich Culture, so many Festivals

We have everything, except tourists.

Now imagine if India touched 10 crore+ foreign visitors every year

Here’s what happens 👇

💰 Billions in forex inflow
🏨 Lakhs of new jobs in hotels, taxis, crafts, food
🏡 More income in small towns & villages
🚆 Better infrastructure across tourist circuits
🌍 And a stronger global image for India

Tourism is soft power + real economy combined.

So what’s stopping us?

❌ Dirty public spaces
❌ Aggressive touting & scams
❌ Patchy infrastructure
❌ Poor marketing abroad
❌ Complicated visitor experience

We don’t need to reinvent the Taj Mahal.

We just need to make India cleaner, simpler, and kinder to our guests.

Solutions are simple:

✅ Keep top 50 tourist zones spotless.
✅ Train taxi drivers, guides & shopkeepers.
✅ Make e-visa truly seamless.
✅ Launch an “India Smiles Back” global campaign.
✅ Turn hospitality into national pride again.

If we fix these, India won’t just have 10 crore foreign visitors,

We’ll have 100 crore stories told around the world. 🌏✨

Because the world already loves India.

We just need to make visiting it a joy, not a challenge. 💛

[Madhusudan @sahnnu]

@OSRTC_Odisha Today on bbsr to paralakhemundi bus, conductor soumya panda sid not issue ticket. SAYS ticket machine not working.

[Madhusudan @sahnnu]

@CYBTRAFFIC @CPHydCity Btw, In the entire stretch of this area. Do you have a single no parking sign ?

F87V+VV6, Masjid Banda Rd, Masjid Banda, Camelot Layout, Gachibowli, Hyderabad, Telangana 500084, India

[Madhusudan @sahnnu]

@CYBTRAFFIC @CPHydCity
Wrongly vehicle picked up for Wrong Parking. Challan no. CYB13CR258374551
When i check portal,
Photo shows only vehicle being towed on the van.

[Madhusudan @sahnnu]

@RAJPUT_NBS

Osrtc toll free number and greviance feature in the app is not functional. Only running volvo buses without proper upkeeping and complaints addressal process will not go long way.

[Madhusudan @sahnnu]

@OSRTC_Odisha @ushapadhee1996 @CMO_Odisha

[Madhusudan @sahnnu]

@suryahse2013 @OSRTC_Odisha @CMO_Odisha @CTOdisha @STAOdisha @ushapadhee1996 @RAJPUT_NBS @BibhutiJena93 @IPR_Odisha @MORTHIndia No one cares

[Madhusudan @sahnnu]

@OSRTC_Odisha @ushapadhee1996 @CMO_Odisha

[Madhusudan @sahnnu]

@OSRTC_Odisha Please respond ASAP

[Madhusudan @sahnnu]

@OSRTC_Odisha Please respond ASAP

[Madhusudan @sahnnu]

@OSRTC_Odisha Us3l3ss service. No timimg

[Madhusudan @sahnnu]

@MartyChargin Why it is selling a profitable unit. And potential buyers?

[Madhusudan @sahnnu]

@OSRTC_Odisha
Very unsafe practices. Doors are not closed. A fellow passenger 500 rupees note fell during ticket purchase and it flew out of the bus. Had the doors been closed it could be saved. This is Bhawanipatna paralakhemundi service