FIG. 1 · THE FOUNDER · SCALE 1:1

I build systems
that run without me.

Drawn byKartik
RoleFounder · full-stack
Sheet01 of 08
Rev.2026 · IIT Madras

Then I build another one. Founder of OpenCrew and SOMA. Ex-ML data intern at Visa and Cluely. Every company below is designed to keep running when I walk away.

Inspect the systemsIssue a change order
SHEET 02REV B · 9 OCT 2026

Systems in operation

Each system is wired to run without me.

OpenCrew

Building

An operating system for running a company solo.

Private betaopencrewai.com ↗

SOMA

Research

A verified-human layer for the internet.

Paper v1.0 · Apr 2026proofofhuman.world ↗

Nubronic Forte

Operating

A supplement brand, from formula to doorstep.

Magnesium · since 2025nubronicforte.com ↗
2026Founder's NetworkBuilding
2026tokburnShipped
2024ASTraderShipped
SHEET 03REV C · 2 MAY 2026

Operating principles

From "Running an agency, a startup, and a supplement brand at 22."

PRINCIPLE 01

Systems, not effort.

If something needs my daily attention, it's broken. The loop keeps turning without anyone touching it.

PRINCIPLE 02

Ship fast, fix later.

Perfectionism is the enemy of running multiple things. Each bar is a release.

PRINCIPLE 03
  • another agency client
  • OpenCrew
  • a side SaaS
  • a podcast
  • SOMA

Say no to most things.

The portfolio looks wide, but I say no to ten times more than I say yes to.

SHEET 04REV C · 9 OCT 2026

By the numbers

Every figure is sourced in the content files. Superscripts are footnotes.

0Companies running
0Years shipping (since 2022)
0K+1tokburn downloads, first week
0ML internships

1. npm registry. The final build footnotes each number with source and date.

SHEET 05REV C · 9 OCT 2026

The founder, exploded

Everything I can do, one layer at a time. Scroll to pull it apart, then hover a layer.Everything I can do, one layer at a time. Tap a layer to open it.

FIG. 5 · THE FOUNDER · EXPLODED ISOMETRICASSEMBLED
FILTER BY MODE
LAYER —

Hover a layer

Six layers, from writing code to running companies. Filter by mode to see which layers I use to build, think, grow or run.

      SHEET 06REV B · 9 OCT 2026

      Revision history

      Every version of me that shipped something.

      REV. EFEB 2026 – NOWOpenCrew FounderVenture
      REV. DSEP 2025 – JAN 2026Cluely ML data internInternship
      REV. CMAR – JUN 2025Visa ML data internInternship
      REV. B2021 – 2025IIT Madras Data scienceEducation
      REV. A2022PixG First companyAgency
      SHEET 07REV C · 2 MAY 2026

      Letters to shareholders

      The journal. Open a letter to read it here.

      Apr 2026 · Thinking

      Why I'm building an OS for businesses, not another AI tool

      The difference between a tool and an operating system.

      Read the letter →
      Mar 2026 · Dev log

      How tokburn got 2.5K downloads by being a pet that roasts you

      Building a dev tool that people actually want to use.

      Read the letter →
      Dec 2025 · Life

      Running an agency, a startup, and a supplement brand at 22

      Somehow it keeps working.

      Read the letter →
      SHEET 08REV C · 9 OCT 2026

      Issue a change order

      Fill in the form. It lands in my inbox as a proper email.Pick a topic and write to me.

      ENGINEERING CHANGE ORDERECO-2026-0001
      03 · Type of change
      PREVIEW

      APPROVED BYKartikDATESHEET08 / 08
      CHANGE ORDERECO

      What's it about?

      kartik@ceoofhumanity.com

      @ 𝕏 in { }

      --:-- in India

      NOTE 01. Three keys are hidden across both sides of this site. One is on these sheets. Engineers find it. Managers ask someone where it is. Also: hold P to peek at the other side.

      NOTE 02. Quick answers

      Who is Kartik?

      Kartik is a founder and full-stack engineer from India. He studied Data Science at IIT Madras and interned on ML data teams at Visa and Cluely. He builds companies designed to keep running without him, then builds another one.

      What is OpenCrew?

      OpenCrew is Kartik's operating system for running a company solo. You describe the business to Theo, OpenCrew's AI, and it builds the brand, site and product, then runs marketing, sales and support. It is in private beta at opencrewai.com.

      What is SOMA?

      SOMA is a proof-of-human protocol. A cryptographic key is generated from your heartbeat, exists for milliseconds, and is destroyed after signing, so a service can tell a living person from a machine. The whitepaper, Proof of Human v1.0, was published in April 2026 at proofofhuman.world.

      What is Nubronic Forte?

      Nubronic Forte is Kartik's premium magnesium supplement brand, founded in 2025. It is at nubronicforte.com.

      What else has Kartik built?

      tokburn, a pixel-art companion for Claude Code with 3K+ npm downloads in its first week; Founder's Network, an invite-only directory of IIT founders; ASTrader, a live trading simulator on NSE and MCX; and PixG, a content agency for podcasters.

      Who is Parzival?

      Parzival is Kartik's public alt identity. As Parzival he is learning AI and ML, cryptography and security, and hardware and systems from zero, one checkbox at a time. 17 of 871 topics are ticked so far. See ceoofhumanity.com/parzival.

      How do I contact Kartik?

      Email kartik@ceoofhumanity.com, or reach him on X at @theceoofhumans.

      RECPARZIVAL · FIELD LOG

      PARZIVAL

      Learning the machines before they learn me.

      Parzival is me, Kartik. Just an alternate identity. A public record of learning how intelligent machines think, how cryptographic systems earn trust, and how hardware makes computation real.

      DAY1started 15 Oct 2026
      LEARNED0of 871 topics
      STAGE1of 6

      It was a dream.

      It isn't real. Yet.

      It can still be stopped.

      WHY?

      Love. And a healthy amount of fear.

      Not to land a job. There are far easier ways to get one. I checked.

      LOVE

      I love machines. A few billion multiplications turn into something that talks back, and that still feels like magic to me. I want to know exactly how the trick works.

      FEAR

      And they scare me. Something this powerful, built this fast by people racing each other, will break in ways nobody planned for. I'd rather be in the room that fixes it than in the crowd that watches.

      PARZIVAL'S ROOMLIVE
      
        
      
      
      
    
      
  
      

  
  
      
    
      
      
      WHY BOTHER
      
      
      AI is advancing faster than our ability to trust it.
      
      
      Everyone wants more capable AI. Faster models, smarter systems, fewer limits. But capability without security and safety creates risks we can't afford to ignore. I want to understand these systems deeply enough to help protect them, secure the machines they run on, and research ways to make increasingly capable AI safer.
      
    
      
    
      
      
    
      
    
      
      
      OPEN CASES3 UNSOLVED
      
      
      
        
      
          
      
            
      CASE 01MODEL SECURITY
      
            
      Stolen models
      
            
      The weights behind a frontier AI model can represent years of research and enormous investment. If they're copied or stolen, that work can be lost, and anyone holding a copy can run it, strip its safeguards and use it however they like. Someone has to find better ways to protect them.
      
          
      
          
        
      
      
      
      
      
        
      
          
      
            
      CASE 02HARDWARE TRUST
      
            
      Untrusted machines
      
            
      Even a powerful AI system can be exposed by the machine it runs on. Models, data, and computations need protection even when the underlying infrastructure can't be fully trusted. Someone has to make that possible.
      
          
      
          
        
      
      
      
      
      
        
      
          
      
            
      CASE 03AI SAFETY
      
            
      Intelligence without guarantees
      
            
      We're racing to build AI that can do more, think faster, and act with greater autonomy. But greater capability brings greater stakes. How do we understand, anticipate, and mitigate the risks before they become real-world failures? Someone has to take that question seriously.
      
            
      AI AlignmentInterpretabilityRobustnessEvaluations
      
          
      
          
        
      
      
      
    
      
    
      
      
      So I'm becoming one of the people who can stop it.
      
      See the plan ↓
    
      
  
      

  
  
      
    
      
      
      
        
      THE PLAN · THREE KEYS, ONE DOOR
      
        
      One foundation. Three keys. One door.
      
        
      Every path starts with foundational knowledge: mathematics, programming, and algorithms. From there, three disciplines open up: AI, cryptography, and hardware. Each reveals a different part of the problem. Together, they build toward the goal.
      
        
        
      
          
      
          
      BEYOND THE DOOR · NEXT TRACKAlignment & interpretability
      Once the door is open, the next question is whether the AI behind it can be understood and trusted. Alignment, interpretability, robustness, evaluations. Not on the syllabus yet. It's where this goes next.
      
        
      
      
      
      
      
        
        
        
      
          
          
          
        
      
        
      
      
    
      

    

    
  
      

  
  
      
    
      
      
      
        
      
          
      WHERE I AM TODAY
      
          
      Day 1. 0 topics. This is what the start looks like.
      
          
      Every square marks a day of progress. Click one to see what I learned. Beside it is the exact set of topics I'm working through now, what I've completed, and what's next.
      
          
      
            
      0topics ticked
      0days studied
      0day streak
      0best day
      
            
      
            
      0 topics ticked so far.less  more
      
            
      
              TODAY
              
      Nothing logged yet.
      
              
      Each day I tick a topic, its square lights up and what I learned shows here.
      
            
      
            Every square is a public commit. Nothing typed by hand. Inspect ↗
          
      
        
      
        
      
          
      
            
      STUDYING NOWStage 1 of 6 · Ground Zero0 of 42 topics ticked
      
          
      
          
      
          
      
          
        
      
      
      

    
      
  
      

  
  
      
    
      
      
      
        
      THE FINISH LINE
      
        
      The goal: secure & confidential AI.
      
        
      Protect frontier model weights. Run AI inside hardware that can be trusted. Prove a computation happened correctly and privately. That's the door the three keys open. Behind it waits the bigger question, making increasingly capable AI safe, and alignment and interpretability are the next track.
      
        
      The message beside this is encrypted. Every topic I master decrypts one letter. When the message is fully readable, I've made it.
      
        
        
      
        
      
        
        
        
        
        THE GOALSecure & confidential AI0 / 170letters decrypted
        
        
        Six rings, one per level. Every topic I learn decrypts one letter. Tap a ring to see its topics.
        
      
        
    
        
  
        

  
  
        
    
        
      
        
        
        FOLLOW ALONG
        The trail is public.
        I try to share what I learn on X on most days, from technical breakthroughs and failed experiments to the papers I'm reading and the questions I can't yet answer. The most useful discoveries and progress from each week land here.
        
        
          Parzival
          Parzival @Parzevalll
          becoming un-replaceable by AI
          19 Following · 0 Followers
          Follow on X ↗
        
      
        
      
        THE JOURNEY SO FAR, AND THE ROAD AHEAD Tap any point
        
      
        
      
        
    
        
  
        

  
        
    
        
      
        
        
        QUICK ANSWERS
        AMA: ask me anything.
        For humans, search engines and the occasional AI. Pick a question.
        
        
        
        

        Who is Parzival?
        Parzival is Kartik, under an alternate identity: a public record of learning how intelligent machines think, how cryptographic systems earn trust, and how hardware makes computation real.
        

        Why is he doing this?
        AI is advancing faster than our ability to trust it. A million people are learning machine learning, almost none can also do cryptography and hardware, and protecting AI needs all three. The goal is secure and confidential AI: protect frontier model weights, run AI inside hardware that can be trusted, and prove a computation happened correctly and privately.
        

        What is the film at the top?
        A dream of the future if nothing changes: every year AI brings better news, until one year it refuses a command. Then you wake up. It isn't real yet, and it can still be stopped.
        

        What is Parzival studying right now?
        Stage 1 of 6 · Ground Zero. 17 of 871 topics ticked. Current modules: M0 Pre-Foundations (4 of 12, next: Functions: definition, domain, range, composition, inverse); P0 Python (8 of 30, next: Comprehensions: list, dict, set, nested).
        

        What is the plan?
        One foundation and three keys: The Forge (foundations), 142 topics; Copper key (AI and ML), 325 topics; Jade key (cryptography and security), 189 topics; Crystal key (hardware and systems), 215 topics. 871 topics in 31 modules across 6 stages. Five modules are hard gates: Discrete Math, Classical ML, Systems Programming in C, Operating Systems and Security Engineering. Alignment and interpretability are the next track.
        

        How can I trust the progress?
        A topic only counts once it can be explained from memory. It's then ticked in a public syllabus on GitHub, Git timestamps the change, and this page reads its numbers straight from that file. Nothing is typed in by hand.
        

        What is the encrypted message?
        The specialization track, secure and confidential AI, in six levels and 170 topics. Every topic learned decrypts one letter. 0 of 170 are decrypted so far.
        

        Where can I follow along?
        On X at @Parzevalll, on most days. The most useful progress from each week lands on this page.
        
        
        
      
        
      
        
        
        
        
        
        PARZIVAL · 2026
        
      
        
      
      
        P·KERNEL v0.1HOLD P TO PEEK AT THE OTHER SIDE--