Rajkiran
Panuganti
CEO, Cerebro Labs Inc. · Agentic Harness · LLM Leader
Former Head of AI, KrutrimMicrosoft CopilotFounding team, BingPhD, CS · Ohio StateB.Tech, ECE · IIT Bombayex-Google2 patents
Building Agentic Harness to make web-based agents as powerful and reliable as terminal-based agents. Former Head of AI at Krutrim, where my team trained India's first LLM from scratch. Previously worked on Microsoft Copilot, Microsoft 365 Copilot Chat, and the founding team of Bing.
↗
CEO
Cerebro Labs Inc. · May 2026
↗
Advisor
Codemod · Jan 2024
↗
Advisor
AppAxon · Jan 2024
↗
LinkedIn Newsletter↗Strategic Advisor
Nelumbium Capital · Mar 2023
Writing
The Emerging Communication Stack for Agents
Two communication stacks for AI agents are being built simultaneously: the channel layer (voice, email, SMS, Slack, WhatsApp) where agents reach humans, and the protocol layer (MCP, A2A, AG-UI) where agents reach each other and tools. Here is what is in each, who is building it, and what 2028 looks like.
Read essay ↗Apr 24
Solving Hallucination: Where the Research Stands
11 min
Apr 22
The Agentic Economy Is Already Here
12 min
Apr 21
The Intelligence Layer
7 min
Apr 15
The Falling Price of Intelligence
9 min
Apr 14
Why Looping Is the New Scaling
9 min
Apr 11
The Quiet Skill Revolution
8 min
Apr 7
AI Gets Personal
9 min
Apr 1
When Bigger Models Get Dumber
10 min
Apr 1
The Stack Got Leaked
8 min
Mar 31
The Stack Eats the Model
7 min
Mar 23
The Three Races Happening in AI Right Now
7 min
Mar 22
The Week AI Learned to Do Its Own Research
6 min
Feb 14
Moltbook Is AI Theater, Not AI Progress
3 min
Feb 14
Most AI Agents Aren't Agents
2 min
Feb 13
Voice Models Finally Sound Human. Now What?
3 min
Feb 12
Open Models Are Winning (Just Not How You Think)
2 min
Feb 12
Agentic Workflows: What Actually Ships vs. What Gets Demoed
3 min
Feb 11
Generative UI Is Solving a Problem Developers Don't Have
3 min
Feb 10
MCP Won the Protocol War. Security Lost.
3 min
Feb 10
Enterprise AI Fails on Boring Problems
2 min
Feb 9
AI Coding Tools Hit a Plateau Nobody Wants to Admit
3 min
Feb 8
The Reasoning Models Race Is Getting Weird
2 min
Feb 8
16 Claude Agents Built a C Compiler. Here's What That Actually Means.
3 min
Feb 7
Computer Use vs Operator: Two Visions for AI Automation
3 min
Feb 6
The AI Workflow Tooling Mess
3 min
Feb 5
RAG Is Harder Than It Looks
2 min
Feb 5
Multimodal Agents Sound Great. Deployment Is Brutal.
3 min
Feb 4
Context Windows Keep Growing. It Doesn't Matter as Much as You Think.
3 min
Feb 2
Prompt Engineering Is Mostly Not Engineering
2 min