Hey, it’s Andreas.
Welcome back to Human in the Loop — your field guide to what just dropped in AI agents, and what’s coming next.

Still on my so-called summer break…but here I am, laptop in tow, toes in the sand. It’s been a standout week for Anthropic — even in the middle of the summer slowdown. McKinsey unveiled a ‘day in the life’ with AI agent Ollie, and Alibaba’s Qwen Code is now giving developers 2,000 FREE daily runs.

And we’re looking closer at the current state of AI in India.

Let’s dive in.

Weekly Field Notes

🧰 Industry Updates
New drops: Tools, frameworks & infra for AI agents

🌀 Claude gets memory recall to retain preferences & past project context
→ Claude now remembers your style, past conversations, and project context. Big step toward persistent agents that adapt to long-term workflows instead of starting from scratch each session.

🌀 Claude Code adds /output-style command for learning & explanatory modes
→ Developers can now switch Claude Code into teaching or explanatory output. A subtle but powerful addition — making agents not just coders, but mentors.

🌀 Anthropic details safeguards for keeping Claude safe & useful
→ Anthropic laid out its latest safety layers — from refusal training to oversight tooling. Shows how the company is positioning Claude as the “safe but capable” enterprise agent.

🌀 Alibaba’s Qwen Code offers 2,000 free daily runs for devs
→ A bold move to lure developers: free daily usage at scale lowers the barrier for coding agents. Smart play in the global LLM competition.

🌀 Docker 4.44 adds tighter MCP integration for faster AI agent builds
→ Docker doubles down on agent-native infra. With built-in MCP hooks, spinning up tool-using agents is now faster and easier for dev teams.

🌀 Pika Labs debuts voice-to-video facial animation model for creators
→ Turns voice recordings into expressive video characters. Creative agents are moving fast from gimmick to production-ready workflows.

🌀 Notion launches hosted MCP server for Claude and other AI integrations
→ Notion quietly became an AI infra provider — hosting MCP servers so teams can plug agents into their workspace without managing backend complexity.

🌀 IBM updates AI Agent Protocols guide for interoperability standards
→ IBM expands its playbook for how agents talk to each other and external systems. The standards race is heating up — and enterprises will care.

🎓 Learning & Upskilling
Sharpen your edge - top free courses this week

📘 LangChain Academy launches free Deep Research agent course
→ Learn how to build multi-agent systems for automated research reports. Practical, hands-on, and completely free — strong pick if you want to get serious about agent workflows.

📘 Microsoft’s Agent Factory blog series shares best practices for agent design
→ Microsoft engineers explain how to scope, design, and iterate on production-grade agents. A tactical series worth bookmarking if you’re building beyond prototypes.

🌱 Mind Fuel
Strategic reads, enterprise POVs and research

🔹 Boston Consulting Group (BCG) on agent-era B2B pricing models
→ Subscription fatigue is real. BCG argues that agents will accelerate outcome-based pricing, shifting enterprise software from “seats” to measurable business impact.

🔹 Deloitte on agentic AI in banking — from fraud detection to treasury
→ A sharp POV: banking workflows are already being re-architected around agents, from real-time fraud checks to treasury automation.

🔹 McKinsey & Company shows a day in the life of fictional AI agent “Ollie”
→ A creative case study to illustrate future workflows. Useful for execs to visualize what agent-driven work looks like in practice.

🔹 KPMG says shadow AI is already here — and can fuel innovation
→ Shadow AI isn’t just a risk — it’s also a source of grassroots innovation. KPMG urges leaders to channel it, not just ban it.

🔹 Palo Alto Networks GenAI report: traffic up 890%, shadow AI risks rising
→ GenAI adoption exploded in enterprise traffic. But Palo Alto warns of compliance and security blind spots as unmonitored AI use surges.

🔹 Infosys report shows only 2% of enterprises have mature Responsible AI practices
→ The maturity gap is stark: most companies talk governance, but only a tiny fraction actually execute at scale. Clear opening for those who can get it right.

♾️ Thought Loop
What I've been thinking, building, circling this week

India’s AI Awakening?

This is a topic I keep coming back to. Over the past 10 years, I’ve worked side by side with Indian colleagues on projects across the globe — and at IBM, one in three employees is based in India. It’s impossible not to notice the depth of skill and scale.

And yet, the paradox remains: India is the world’s largest exporter of tech talent — but one of the smallest inventors of AI. Just look at the global leadership table: Microsoft, Google, YouTube, IBM, Palo Alto, Flex, Micron, Arista, NetApp, Cognizant — all led by Indian-origin CEOs. The diaspora dominates boardrooms.

India is not short on skill or scale. What’s missing is the translation of that capacity into homegrown invention.

For decades, India’s IT giants thrived on services, not invention. R&D spend sits at just 0.65% of GDP (vs. 2.7% in China, 3.5% in the U.S.). Venture funding paints an even starker picture: in 2024, Indian AI startups raised $780M, compared with $97B in the U.S. and $51B in Europe. Add to that the challenge of 22 official languages and 19,500 dialects — many without clean digital datasets — and India looked structurally disadvantaged in the foundation model race.

But recent developments — especially DeepSeek’s sudden rise in China — have changed the equation. Practically overnight, New Delhi shifted gears:

• $1.25B IndiaAI Mission → to fund sovereign models and national AI infrastructure.

• 19,000 GPUs pooled (13,000 H100s included) → through Jio, Yotta, AWS partners, Tata and others.

• Six sovereign models funded → including Sarvam AI’s 70B multilingual LLM with reasoning + voice.

• Startup wave → Soket AI Labs (120B params), Gan.ai (70B), Gnani.ai (14B voice), and CoRover.ai’s BharatGPT (3.2B multimodal, now shipping a “Mini” at 534M).

However, the technical barriers are real:

• Sparse data → Many Indian languages have little digital text or consistent spelling.

• Hard tokenization → Scripts like Hindi, Kannada, Tamil often lack spaces, making word boundaries invisible to global LLM tokenizers.

• Capital gap → India’s startups run on tiny budgets compared to their Western or Chinese peers.

And yet, Indian builders are innovating under constraints. Techniques like “balanced tokenization” and voice-first design are producing efficient multilingual models tuned for India’s reality — and potentially for the Global South at large.

• Talent leverage → Millions of engineers already fluent in global AI. Infrastructure keeps them at home.

• Language moat → Cracking multilingual, voice-first AI creates defensible IP.

• Cost edge → Data centers in India cost ~50% less to build vs. U.S. or Europe.

I’m convinced that if India executes, it won’t just catch up. Whether this shift happens tomorrow or over the next few years is uncertain — but it’s worth keeping a close eye on how Indian startups and policymakers are moving. What they do, and how they do it, could set the standard for sovereign AI done differently: efficient, multilingual, and built for billions outside the Western mainstream.

P.S. I’ll be back in India soon — speaking at a major event in Bangalore. I’m also working on a major project there with many of my talented colleagues, and I’ll be visiting them soon in Calcutta — which keeps me close to the ground on how this ecosystem is really evolving. I’ll keep you posted!

🔧 Tool Spotlight
A tool I'm testing and watching closely this week

MCP is becoming the API layer for the agentic web.

At IBM we’ve been curating a list of all the MCP Servers, Clients, and Developer Tools we’re building at IBM — and it’s growing fast.

→ Servers for cloud, observability, storage, and data products
→ Developer tools to build, test, and connect to MCP servers
→ Clients to integrate directly with your agents and workflows

If you’re building AI agents or experimenting with the Model Context Protocol, this repo is for you and a great place to start with!

Try it now:
→ Explore it here.