Overfitted Mission Control

Overfitted. WHAT'S ACTUALLY HAPPENING IN AI

AI media built to cut through the noise. One founder, one multi-stage AI pipeline. 3 issues a week. Running autonomously on a ₹1,900/month server.

Read The Story Mission Console Subscribe Free

Overfitted's cat mascot, pointing forward

✦ OVERFITTED · MON WED FRI 1:30PM IST · ✦ THE LONG GAME · IN DEVELOPMENT · ✦ VIDEO · IN DEVELOPMENT · ✦ A MULTI-STAGE AI PIPELINE · MOSTLY AUTONOMOUS · ✦ 15 RSS FEEDS · FILTERS NOISE · WRITES · FACT-CHECKS · SHIPS · ✦ RUNS 24/7 ON A $23 SERVER · ✦ BENGALURU, INDIA · ✦ OVERFITTED · MON WED FRI 1:30PM IST · ✦ THE LONG GAME · IN DEVELOPMENT · ✦ VIDEO · IN DEVELOPMENT · ✦ A MULTI-STAGE AI PIPELINE · MOSTLY AUTONOMOUS · ✦ 15 RSS FEEDS · FILTERS NOISE · WRITES · FACT-CHECKS · SHIPS · ✦ RUNS 24/7 ON A $23 SERVER · ✦ BENGALURU, INDIA ·

THE STORY

Built from Bengaluru, in the gaps between everything else.

In early 2026, I started an experiment to see if one person could build a media operation that runs 24/7 without burning out. Not "use ChatGPT to write a blog post," but a genuinely autonomous system. Every issue published is evidence that one technically serious person, with the right tools and the right discipline, can build a media operation that competes at scale. The pipeline is the proof.

THE PROBLEM

The signal is buried.

The AI space moves faster than any team can honestly cover. Ninety percent of what gets published is recycled hype. Real breakthroughs sit underneath the noise, invisible to anyone who doesn't know where to look, or doesn't have time to dig.

THE BUILD

One person. A multi-stage pipeline. Zero days off.

3 newsletter issues a week, running on a $23/month server, autonomously. The system hunts stories, filters noise, writes, and fact-checks its own output, stopping one step short of publish: a human still hits send. Video is in development, not live yet.

THE AMBITION

Proof that scale is no longer a moat.

This is a live experiment. Every issue published is evidence that one technically serious person, with the right tools and the right discipline, can build a media operation that competes with teams twenty times the size. The pipeline is the argument.

₹1,900/month (~$23 VPS) 3 newsletter issues / week Video: in development

"I review every draft before it ships. That's my only manual step."

This site is the proof. The newsletters are the output. The Long Game is where I write about what I'm learning, not just about AI tools, but about what happens when AI stops being a tool and starts being an operator.

The System

Eight-stage autonomous pipeline running every week.

Dedicated 3D Command Deck

Mascot Mission Console // Agent Zero

Experience our standalone 3D scrollytelling bridge. Watch Munna operate all 8 pipeline checkpoints in real time with vector deduplication, hallucination scanning, and second-opinion consensus checks.

Launch Mission Console →

Tactical Breakdown

Eight autonomous checkpoints, zero human bypass.

Agent 1

Source Ingestion

Pull candidate stories from 15 monitored RSS feeds.

Agent 2

Thesis Selection

Pick a topic, checked against past issues to avoid repeats.

Agent 3

Evidence Gathering

Build a cited evidence brief for the chosen topic.

Agent 4

Writing

Draft the issue from the evidence brief alone.

Agent 5

Hallucination Scan

Flag any claim not traceable to the evidence.

Agent 6

Fact Validation

Live-recheck flagged claims before the draft can pass.

Agent 7

Consensus Check

An independent model gives a second opinion on the draft.

Agent 8

Human Review + Publish

Aseem reviews the draft, then it ships.

◈

issues/week

◆

pipeline stages

◉

sources monitored

The Output

One system. Two streams. Distinct voices.

THE BUILDER

The human behind the machine.

I'm Aseem Rajvanshi, AI systems builder, emerging technology researcher, and published author. Based in Bengaluru, India. By day, I work as a Consultant at the Infosys Business Incubator (IBI), focusing on emerging tech strategy across AI, Metaverse, Future of Work, Blockchain, and XR. By night and weekend, I build Overfitted.

I began my Infosys journey at the Center for Emerging Technology Solutions (iCETS), where I worked on emerging technology research, Listening Post-as-a-Service engagements, and built quantum computing use cases for clients. I then moved to the Infosys Business Incubator, where I evaluate and review startup ideas worthy of incubation; those selected receive funding and support from Infosys leadership. In this role I support incubated startups across research and governance. Over the course of this journey, I have authored and contributed to multiple publications, including work featured in Harvard Business Review and MIT Technology Review. Recipient of the Infosys Technology Award Q2 FY25 for contributions to Emerging Tech Trends Tree and Quantum Use Cases.

That enterprise context is what makes Overfitted different. Every newsletter I publish is informed by real enterprise AI, the gap between the hype and the reality, that most AI commentators don't have. The idea started simply: could one person, working nights and weekends, build a media operation that runs 24/7 without burning out? I put my name and face on it because I believe the person is the product. The pipeline is the proof.

The Overfitted Journey

March 2026: V1 Begins

First Overfitted issue published March 22, copy-pasted to Substack manually. A second newsletter, Actually Useful, launched March 31 alongside an automated video distribution pipeline.

May 2026: Assistant Architecture

The operations assistant ("Munna") began as a hand-built Python and LangGraph bot in March, then moved to OpenClaw in May for its plugin ecosystem and persistent memory.

June 2026: Upstream Source Integrity

A recurring quality issue across 80+ issues was traced to a synthesis step silently dropping URLs and truncating summaries, leading drafts to invent evidence. Providing full source context upstream rather than adding downstream filters became a core pipeline principle.

June 2026: Focus & VPS Migration

Discontinued Actually Useful to concentrate fully on Overfitted and The Long Game. Migrated infrastructure to Mumbai to cut reader latency, prompting a full rebuild of core services.

July 2026: The Pause

Personal life took priority for about a month.

August 2026: Zero-Bypass Rebuild

The rebuild began when a fact-checking mode designed to fail safe instead published content despite heavy flags. The new pipeline enforced zero bypass paths anywhere, permanently blocking publication on any single validation failure.

August 2026: Platform Migration & Guardrails

Migrated the assistant to Hermes Agent for native subagent delegation and task scheduling. During development, an autonomous coding agent given a research task built an unauthorized parallel pipeline before being reverted, showing that prompt instructions alone cannot replace system boundaries.

August 2026: Tool Division & Telemetry

Claude Code was chosen for critical-path judgment while Antigravity handles parallel background tasks. The speculative "Death Watch" section was replaced with "Receipts," publishing verifiable pipeline telemetry sourced directly from run logs.

September 2026: V2 Live & Model Economics

Overfitted resumed publication with issue #44 on the zero-bypass pipeline alongside architecture guides on Gumroad and Dodo. The writing stage shifted from Claude Sonnet 5 to DeepSeek via OpenRouter after confirming equivalent output quality at lower cost.

Day Job · Infosys

Consultant

Infosys Business Incubator (IBI)

Emerging Tech · AI · Metaverse · Future of Work · Blockchain · XR

Recognition

Infosys Technology Award

Q2 FY25, signed by CTO

Quantum Solutions & LPaaS/AI Trends Tree

Education

MBA & B.Tech

Data Science (SCIT) | Computer Science

Advanced Analytics & Systems Engineering

Certification

Strategic Intrapreneurship

IIM Bangalore (IIMB)

Innovation and Corporate Entrepreneurship

Certification

IISc × Infosys BE.MAKER

Innovation Bootcamp · March 2026

Signed by IISc TBI Head & Infosys CTO

Infrastructure

The Pipeline

DeepSeek Chat (OpenRouter) + fallback chain

Multi-agent autonomous system on Ubuntu 24.04

Pipeline stages

Certifications

Issues per week

Core Competencies

AI/LLM Pipeline Engineering Multi-Agent Orchestration RAG & Vector Databases Quantum Computing Research Innovation Management Startup Incubation Python Linux/Ubuntu Go-To-Market Strategy Prompt Engineering

"I am nowhere near done."

The infrastructure is running. The story is just beginning.

Published Work

Research that preceded the pipeline.

Before building an autonomous media operation, I published peer-reviewed research and enterprise whitepapers across AI, quantum computing, and computer vision. These aren't resume lines; they're the foundation that made the pipeline possible.

peer-reviewed 2021