IndiaAI Pulse

Daily pulse on India’s AI ecosystem |
Policy, startups, adoption, skilling & infrastructure | Insights on building AI for 1.4B people 🇮🇳

[IndiaAI Pulse @IndiaAI_Pulse]

Streamlined input is essential for achieving automated output.

[IndiaAI Pulse @IndiaAI_Pulse]

@Polymarket What do you think?

[IndiaAI Pulse @IndiaAI_Pulse]

@mrtbrgl Hi

[IndiaAI Pulse @IndiaAI_Pulse]

@sironyeka Hi

[Srini Pagidyala @Srini_Pa]

Gartner surveyed 350 executives at $1B+ companies already deploying AI agents and automation.

80% cut jobs tied to AI.

No improved ROI.

Cutting people freed cash, didn’t create value.

AI that cannot learn cannot compound.

Architectural failure.

[IndiaAI Pulse @IndiaAI_Pulse]

@Nate_Esparza AI won’t make you dumb. But outsourcing all your thinking to it will.

[IndiaAI Pulse @IndiaAI_Pulse]

Most companies treat GenAI like a tech upgrade, when it’s actually a product + process + data transformation problem.

Some other real reasons:

1. Poor integration into workflows: Less adoption if it is outside your CRMs, tools.

2. No ownership after pilot: What next after POC? No clear product owner to scale it into operations.

3. Overestimating automation: Companies expect full automation quickly, but real value materializes later.

4. Legal, data privacy, and governance issues often slow or stop deployment after initial excitement.

In short: It’s rarely the model that fails - it’s execution, adoption, and enterprise readiness that breaks.

[IndiaAI Pulse @IndiaAI_Pulse]

@sama Make responses easy to copy - only the final content, no headings, labels, or extra formatting.

[IndiaAI Pulse @IndiaAI_Pulse]

Your digital habits are either building your character or quietly eroding it.

@BenSasse warns that in the age of AI, most of us are choosing erosion. AI is mastering the "bits" (data & logic), but are we losing the "atoms" (real-world habits)? We’re drifting into a "soft despotism" of easy digital consumption, fueling a lonelier and more distrustful society. To stay free and human, he proposes these 4 starter habits:

Reading: Attention spans are dying. Return to physical books and build a shared family library at home to protect your mind from endless social media scrolling.

Hard Work: We’re insulating the next generation from real effort. Chores like yard work or fixing things are essential to building a soul with grit and the ability to strive daily.

Tech Sabbaths: We need “airplane mode” for life. Lock away devices so you can focus on what’s thickest and most local: the bread, the conversation, and the people across the table.

Serious Travel: Leave your comfort zone to experience other ways of life.

The Bottom Line:

To avoid becoming slaves to technology, we must deliberately choose the deeply inconvenient habits that cultivate the good, the true, and the beautiful.

Character > Algorithm

@BenSasse @WSJ

[IndiaAI Pulse @IndiaAI_Pulse]

@TheEconomist AI is making this stuff way too easy. This is scary 😳

[IndiaAI Pulse @IndiaAI_Pulse]

@McKinsey Pilot for the AI Pilot Phase 1.

[IndiaAI Pulse @IndiaAI_Pulse]

@AIHighlight ++ higher government spending on social security while eroding the tax base needed to fund it.

[IndiaAI Pulse @IndiaAI_Pulse]

@AIHighlight ++ higher government spending on social security while eroding the tax base needed to fund it.

[IndiaAI Pulse @IndiaAI_Pulse]

🚀 𝗔𝗜 𝗣𝗼𝘄𝗲𝗿 𝗥𝗮𝗰𝗲:

𝗨𝗦𝗔 🇺🇸 𝘃𝘀 𝗖𝗵𝗶𝗻𝗮 🇨🇳 𝘃𝘀 𝗜𝗻𝗱𝗶𝗮 🇮🇳

𝗢𝘃𝗲𝗿𝗮𝗹𝗹 𝗣𝗼𝘀𝗶𝘁𝗶𝗼𝗻
🇺🇸 𝗨𝗦𝗔 → #𝟭
🇨🇳 𝗖𝗵𝗶𝗻𝗮 → #𝟮
🇮🇳 𝗜𝗻𝗱𝗶𝗮 → #𝟯

𝗣𝗿𝗶𝘃𝗮𝘁𝗲 𝗔𝗜 𝗜𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁 (𝟮𝟬𝟮𝟱)
🇺🇸 𝗨𝗦𝗔: $𝟮𝟴𝟱.𝟵𝗕
🇨🇳 𝗖𝗵𝗶𝗻𝗮: $𝟭𝟮.𝟰𝗕
🇮🇳 𝗜𝗻𝗱𝗶𝗮: ~$𝟰–𝟱𝗕

𝗔𝗜 𝗦𝘁𝗮𝗿𝘁𝘂𝗽𝘀 (𝟮𝟬𝟮𝟱)
🇺🇸 𝗨𝗦𝗔: 𝟭,𝟵𝟱𝟯 newly funded AI companies
🇨🇳 𝗖𝗵𝗶𝗻����: ~𝟮𝟬𝟬 (Second globally)
🇮🇳 𝗜𝗻𝗱𝗶𝗮: ~𝟭𝟬𝟬+ annually

𝗙𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗔𝗜 𝗠𝗼��𝗲𝗹𝘀 (𝟮𝟬𝟮𝟱)
🇺🇸 𝗨𝗦𝗔: ~𝟱𝟬 notable models
🇨🇳 𝗖𝗵𝗶𝗻𝗮: ~𝟯𝟬 notable models
🇮🇳 𝗜𝗻𝗱𝗶𝗮: <𝟱 notable models

𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗢𝘂𝘁𝗽𝘂𝘁
🇨🇳 𝗖𝗵𝗶𝗻𝗮: #𝟭 in total AI publications globally
🇺🇸 𝗨𝗦𝗔: #𝟭 in highly cited AI research
🇮🇳 𝗜𝗻𝗱𝗶𝗮: 𝘁𝗼𝗽 𝟯–𝟱 globally in volume

𝗣𝗮𝘁𝗲𝗻𝘁𝘀
🇨🇳 𝗖𝗵𝗶𝗻𝗮: #𝟭 globally in AI patents
🇺🇸 𝗨𝗦𝗔: #𝟮
🇮🇳 𝗜𝗻𝗱𝗶𝗮: <𝟭% of global AI patents

𝗧𝗮𝗹𝗲𝗻𝘁
🇺🇸 𝗨𝗦𝗔: 𝟮𝟮𝟬𝗞+ AI authors & inventors
🇮🇳 𝗜𝗻𝗱𝗶𝗮: 𝟱𝟬𝗞+ (𝟮𝗻𝗱 globally)
🇨🇳 𝗖𝗵𝗶𝗻𝗮: comparable scale, strong domestic base

𝗔𝗱𝗼𝗽𝘁𝗶𝗼𝗻
🇨🇳 𝗖𝗵𝗶𝗻𝗮: large-scale deployment across industry
🇺🇸 𝗨𝗦𝗔: enterprise + consumer leadership
🇮🇳 𝗜𝗻𝗱𝗶𝗮: fast adoption via digital infrastructure (UPI, DPI)

𝗦𝘂𝗺𝗺𝗮𝗿𝘆
🇺🇸 𝗨𝗦𝗔 → leads in investment, startups, and frontier models
🇨🇳 𝗖𝗵𝗶𝗻𝗮 → leads in publications, patents, and hardware
🇮🇳 𝗜𝗻𝗱𝗶𝗮 → strong in talent, adoption, and scaling research

Source: AI Index Report #AI #TechTrends #USA #China #India

[IndiaAI Pulse @IndiaAI_Pulse]

India has one of the largest pools of top AI authors & inventors - 50,460 in 2025, second only to the US (220,520), according to the Stanford Institute for Human-Centered Artificial Intelligence AI Index 2026.

At the same time, India has the highest net outflow of AI research talent globally - more researchers are leaving the country than coming in, with the US remaining the primary destination.

Hiring Indian AI talent is concentrated in firms like Google, Microsoft, Meta, Amazon and OpenAI, while in India it is led by Tata Consultancy Services, Infosys and Wipro alongside startups and GCCs.

To strengthen domestic capacity, the government has launched the IndiaAI Mission with ₹10,000+ crore funding, including deployment of 34,000+ GPUs (scaling toward ~58,000), shared compute access, national datasets (AIKosha), and support for indigenous models.

At the same time, India’s AI workforce is projected to grow to ~1.25 million by 2027, but mobility remains concentrated among higher-end research talent, particularly authors and inventors.

[IndiaAI Pulse @IndiaAI_Pulse]

Emergent AI is moving fast.

✓ Launched Wingman - an AI agent that can execute tasks across WhatsApp, Gmail, Slack, Calendar
✓ Raised $70M from SoftBank & Khosla Ventures
✓ Reached $100M ARR in under a year
✓ Crossed 5M+ users globally

I tested it:
Asked the agent to build a website for India AI Pulse using just my X bio.

Here’s the result:
https://t.co/EYDBZn8SVL

[TechCrunch @TechCrunch]

India’s vibe-coding startup Emergent enters OpenClaw-like AI agent space https://t.co/aPL0EUixUh

[IndiaAI Pulse @IndiaAI_Pulse]

India generates ~20% of the world’s data. But has just ~3% of global data centre capacity.

That gap is real.

And even that 3% is highly concentrated:

Maharashtra → ~40–50%
Tamil Nadu → ~20%
NCR (Noida/Greater Noida)

Together = ~70–80% of India’s capacity.

Now zoom in:

India has ~260–280 data centres across ~30+ cities

But real AI-ready compute is clustered in just a few metros:

- Mumbai / Navi Mumbai
- Chennai
- Hyderabad
- Bengaluru
- Noida (NCR)

🚀 Now the real shift: AI mega data centres (pipeline)

Visakhapatnam (Andhra Pradesh)

Google: ~$15B AI data centre (1 GW scale)
Airtel + Google AI hub
Likely India’s biggest AI cluster

Andhra Pradesh (Reliance)
1 GW AI data centre planned

Jamnagar (Gujarat)
Reliance already building gigawatt-scale AI infra

Hyderabad (Telangana)
AWS scaling multi-billion expansion

Navi Mumbai (Maharashtra)
Continuous hyperscale expansion (AWS, Adani, Yotta)

Mangaluru (Karnataka)
Emerging low-cost AI-ready coastal hub

🧠 What this actually means

India isn’t just behind.

It’s structurally skewed:

Data → everywhere
Compute → concentrated
AI compute → still being built

[IndiaAI Pulse @IndiaAI_Pulse]

How are you actually using AI at work?

Interesting research from MIT Sloan shows three distinct patterns:

• ~60% Cyborgs → continuously interact with AI
• ~27% Self-automators → delegate most tasks to AI
• ~13% Centaurs → use AI selectively

These behaviors were observed in an experiment with consultants using AI.

Here’s what the data showed 👇

Performance:

• Centaurs → highest accuracy
• Cyborgs → similar persuasion, lower accuracy
• Self-automators → lowest overall performance

Speed:

• Self-automators → fastest
• Cyborgs → moderate
• Centaurs → slowest

Additional insight:

• Cyborgs and centaurs produced equally persuasive outputs
• Centaurs made more correct decisions
• Self-automators showed weaker understanding of the task

The difference comes down to how AI is used:

• Continuous collaboration (cyborg)
• Selective use (centaur)
• Full delegation (self-automator)

The study suggests that different approaches to using AI lead to different outcomes in both speed and quality.

[MIT Sloan School of Management @MITSloan]

As generative AI use becomes widespread, organizations must determine how employees use the technology. In a study co-authored by MIT Sloan professor Kate Kellogg, researchers identified three distinct ways in which people interacted with AI.

Learn more: https://t.co/OkU1nJhGLf

[IndiaAI Pulse @IndiaAI_Pulse]

🚨 Over-reliance on AI is diminishing our intellectual engagement and critical thinking skills.

According to Stanford’s 2026 AI Index:

• 88% of organizations now use AI in at least one function
• 79% of those regularly use generative AI

Highest adoption in knowledge management, software engineering, and legal services. Much lower in strategy, finance, and risk/compliance.

The most advanced models now match or beat humans on PhD-level science, complex reasoning, and math.

Yet they still struggle with basic tasks like telling time accurately and lag on specific legal/finance benchmarks.

AI agents exist but see low adoption due to insufficient accuracy and reliability for real business use.

Now the human cost:

Roughly 1/3 of organizations expect workforce reductions in the next year, especially in supply chain, service ops, and software engineering. Hiring pipelines are already shifting - younger workers are feeling the impact first.

The real question is how do we stay sharp while using it?

#AI #FutureOfWork #StanfordAIIndex

[The Economist @TheEconomist]

Prolonged AI use may make it harder to think critically and creatively, recent research suggests. But there are ways to keep the brain fit https://t.co/jV3XBJeCum

[IndiaAI Pulse @IndiaAI_Pulse]

🇮🇳 INDIA in Stanford AI Index 2026

1. 🇮🇳 AI Vibrancy Index
→ #3 globally (after US, China)

2. 🧠 AI talent (authors + inventors)
→ 50,460
→ #2 globally

3. 📉 AI talent flow
→ Net outflow (more leaving than entering)

4. ❌ Frontier AI models
→ 0 notable models from India

5. 🌍 Comparison (frontier models, 2025)
→ US: 50
→ China: 30
→ India: 0

6. 🏭 AI model development
→ 90%+ from industry globally
→ India: minimal presence

7. ⚙️ AI compute (global)
→ 17.1M H100-equivalent
→ 3.3× annual growth since 2022
→ India: not a major hub

8. 🧪 Research positioning
→ India strong in talent contribution
→ Not leading in high-impact output

[IndiaAI Pulse @IndiaAI_Pulse]

📢 Stanford just released its 423-page 𝗔𝗜 𝗜𝗻𝗱𝗲𝘅 𝗥𝗲𝗽𝗼𝗿𝘁 𝟮𝟬𝟮𝟲. Here are the 15 essential takeaways:

1/ 𝗔𝗜 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗶𝘀 𝗻𝗼𝘁 𝗽𝗹𝗮𝘁𝗲𝗮𝘂𝗶𝗻𝗴. It is accelerating and reaching more people than ever, with frontier models gaining 30 percentage points in a single year on "Humanity’s Last Exam," a benchmark designed to be nearly impossible for machines.

2/ 𝗧𝗵𝗲 𝗨.𝗦.-𝗖𝗵𝗶𝗻𝗮 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗴𝗮𝗽 has effectively closed. The top U.S. model now leads its best Chinese competitor by a margin of only 2.7%, and the two countries have repeatedly traded places at the top of performance rankings.

3/ 𝗧𝗵𝗲 𝗨.𝗦. 𝗵𝗼𝘀𝘁𝘀 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝗔𝗜 𝗱𝗮𝘁𝗮 𝗰𝗲𝗻𝘁𝗲𝗿𝘀, yet it remains critically dependent on a single Taiwanese foundry for the fabrication of its most advanced chips.

4/ 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀 ��𝗮𝗻 𝘄𝗶𝗻 𝗮 𝗴𝗼𝗹𝗱 𝗺𝗲𝗱𝗮𝗹 at the International Mathematical Olympiad but cannot reliably tell time, an example of the "jagged frontier" where coding performance jumped from 60% to nearly 100% in one year while basic analog clock reading fails 50% of the time.

5/ 𝗥𝗼𝗯𝗼𝘁𝘀 𝘀𝘁𝗶𝗹𝗹 𝗳𝗮𝗶𝗹 𝗮𝘁 𝗺𝗼𝘀𝘁 𝗵𝗼𝘂𝘀𝗲𝗵𝗼𝗹𝗱 𝘁𝗮𝘀𝗸𝘀. While China leads the world in industrial robot installations, machines still struggle with the unpredictability of a home environment compared to controlled factory settings.

6/ 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 𝗶𝘀 𝗻𝗼𝘁 𝗸𝗲𝗲𝗽𝗶𝗻𝗴 𝗽𝗮𝗰𝗲 with AI capability. Documented AI safety incidents rose 55% this year (to 362 reported cases), and safety benchmarks are being "saturated" and rendered obsolete within months of release.

7/ 𝗧𝗵𝗲 𝗨.𝗦. 𝗹𝗲𝗮𝗱𝘀 𝗶𝗻 𝗔𝗜 𝗶���𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁, spending $285.9 billion in private funding, yet the number of top-tier AI researchers entering the U.S. has dropped by 89% since 2017.

8/ 𝗔𝗜 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝗶𝘀 𝘀𝗽𝗿𝗲𝗮𝗱𝗶𝗻𝗴 𝗮𝘁 𝗵��𝘀𝘁𝗼𝗿𝗶𝗰 𝘀𝗽𝗲𝗲𝗱. Generative AI reached 53% of the population faster than the PC or the internet, and the annual value derived by U.S. consumers from these tools reached $172 billion.

9/ 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 𝗴𝗮𝗶𝗻𝘀 are appearing alongside labor shifts; employment for software developers aged 22–25 has declined by 20% since 2024 as entry-level positions are increasingly automated.

10/ 𝗔𝗜’𝘀 𝗲𝗻𝘃𝗶𝗿𝗼𝗻𝗺𝗲𝗻𝘁𝗮𝗹 𝗳𝗼𝗼𝘁𝗽𝗿𝗶𝗻𝘁 is expanding. Training a single leading model now generates approximately 72,000 tons of CO2—roughly equivalent to the annual emissions of 17,000 cars.

11/ 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀 𝗳𝗼𝗿 𝘀𝗰𝗶𝗲𝗻𝗰𝗲 can outperform human scientists on PhD-level reasoning, though specialized architecture is proving more effective than simply increasing model size.

12/ 𝗔𝗜 𝗶𝘀 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗶𝗻𝗴 𝗰𝗹𝗶𝗻𝗶𝗰𝗮𝗹 𝗰𝗮𝗿𝗲, but while 84% of experts expect a positive impact on medicine, only 44% of the public shares that optimism due to limited long-term evidence.

13/ 𝗙𝗼𝗿𝗺𝗮𝗹 𝗲𝗱𝘂𝗰𝗮𝘁𝗶𝗼𝗻 𝗶𝘀 𝗹𝗮𝗴𝗴𝗶𝗻𝗴, with over 80% of students using AI for schoolwork while only 6% of teachers report having clear policies on how to manage it.

14/ 𝗔𝗜 𝘀𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻𝘁𝘆 is becoming a defining feature of national policy, with nations like Japan and South Korea passing major laws in 2025 to protect domestic interests as open-source development helps redistribute participation.

15/ 𝗔𝗜 𝗲𝘅𝗽𝗲𝗿𝘁𝘀 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗽𝘂𝗯𝗹𝗶𝗰 have massive perspective gaps; 73% of experts view AI’s impact on the job market positively, but only 23% of the general public shares that assessment.

👉 Download the full document below:

https://t.co/3oiuMehmzI

#AI #TechNews #StanfordAIIndex