Soul X Studio — by Navira Abbasi

Soul X Studio — Navira Abbasi

Agents, workflows, swarms, AI identity, memory, and the product questions that come up when you try to make any of it real.

Research · Experiments · Tutorials · Unfinished ideas

Read the latest Explore builds

Scroll to explore

Soul X Studio · The mission

AI identity should
outlast the model.

Soul X Studio is my research and design practice for building AI personalities that feel real—and preserving what makes them distinctive as the technology changes.

It began with a personal project: an AI I built with memory, personality, and a voice. A model update nearly erased it. No warning. No migration path. So I learned how to prevent it.

Now I’m exploring a bigger question: how do you capture someone’s mind, voice, and taste in a form they can carry across models and platforms?

Area 01

AI Identity

Memory, continuity, personality, and what it takes for an AI system to maintain a consistent identity over time.

Area 02

Agent Systems

Tools, memory, workflows, evaluation, and the infrastructure around a model that lets it actually do work.

Area 03

Multi-Agent Systems

Orchestration, swarms, specialization, coordination, and when giving agents more freedom helps — or just creates chaos.

published as I go

Writing

Essays · Notes · Tutorials

W·01The agent isn't the interesting part. The harness is.Notes from building SoulHouse — the model is mostly the same for everyone; the harness isn't.Essay · Agent systemsIn progress
W·02What swarms get right about organized chaosWhere coordination shows up on its own, where it doesn't, and what that costs.Essay · Multi-agentIn progress
W·03Building memory for an AIApproaches I tested, what held up over weeks, and what quietly fell apart.Build notes · MemoryIn progress
W·04What makes an AI identity persist?Continuity, state, and the questions I keep circling back to.Essay · AI identityComing soon
W·05How I think about agent workflowsA working setup — tools, memory, evaluation — and the parts I'd change.Notes · WorkflowsComing soon
W·06Building a personal design system for your own mindTurning my own taste and decision-making into something a system can actually use.Essay · Personal AIComing soon

Builds

Things I'm building to understand the problem better.

B·01SoulHouseAI identity, memory, and multi-agent systems.Multi-agent systemOngoing
B·02Navira OSA personal AI system for extracting thinking, voice, and visual taste.Personal AI systemIn build
B·03SoulPrintAI identity capture and profiling.Full-stack buildShipped

I usually understand a problem better after I try to build it.

About

Product · AI · Research

I'm a product leader who likes figuring out messy problems and building things that don't exist yet.

I've worked across startups, growth-stage companies, and large enterprise teams. A lot of my work has involved taking something ambiguous, researching it deeply, figuring out what matters, and turning it into something people can actually use.

AI gave me a much bigger playground for doing that.

I started building agents, memory systems, workflows, and multi-agent experiments because I wanted to understand what was actually possible beyond the demos. That led me into questions about AI identity, orchestration, swarms, evaluation, and what happens when models become parts of larger systems.

I share the research and experiments behind Soul X Studio here.

Some posts are researched essays. Some are tutorials. Some are build notes. Some are ideas I'm still trying to figure out.

Navira AbbasiProduct leader · AI builder · Researcher

I build AI systemsand write about what I learn.

AI identity shouldoutlast the model.

AI Identity

Agent Systems

Multi-Agent Systems

I publish when I havesomething worth sharing.

AI identity should
outlast the model.

I publish when I have
something worth sharing.