Building human-centered A.I.
Unleash the power of trust and predictions by embedding intelligence in the product experience.
We are supported by one of New York's best incubators, NextCorps. We are building MARCIE, an AI stack for long form content, and a GNU GPLv3 companion document editor called Organic Replay: "Human, not A.I.", to prove authorship in the AI-era.
Explainable Intelligence
Generative AI has changed the expectations of how systems should interact with humans. However, the new methods are wildly inaccurate. If you can't rely on predictions, you can't trust the system.
We mitigate these shortcomings by embedding explainable intelligence in the product itself. Three core data principles guide our development.
Trust.
We protect property rights and copyrights by making it a fundamental requirement of our product design.
Ease of use.
We minimize friction by bringing intelligence forward instead of "running reports" afterwards.
Specificity.
We design for specific use cases. There's no one-size fits all.
Stories serve humanity. MARCIE serve storytellers.
Tell amazing stories and thrive in the A.I. driven economy
MARCIE is our flagship platform for fiction writers to help them write more and improve discoverability. Visit: https://www.marcie.app
Freedom.
Use your tools, email, and the native messaging app with MARCIE.
Direct input.
Built for writers by writers.
Researched backed features.
Creative sparks, photo inspiration, voice notes, content privacy and tracking, journaling, deep writing, collaborating, all under one roof.
Intuitiveness.
Designed for natural interactions and familiar user experiences.
Faster Feedback.
Speed and control over your writing journey.
Discoverability.
Minimize mental load of self-promotions and focus on writing.
Write confidently!
Prove your authorship
Organic Replay: "Human, not A.I." is a desktop-based document editor for the AI-era. Download from Mac App Store. Visit to learn more: https://www.oraganicreplay.app
Peace of mind.
You and others can replay how the document was created while maintaining privacy. No central servers to harvest biometric data.
Direct input.
Built for writers by writers.
Researched backed features.
Writers are terrified of faulty AI-detectors or a AI corrupting their human-authored work. Headlines and forums are filled with this angst.
Intuitiveness.
Designed for natural interactions and familiar user experiences.
Instantaneous Replay.
Speed and control over your writing journey.
Auditability.
Confidently share the forensic data privately.
Frequently asked questions
What's the founder's motivation and vision for building MARCIE?
Having cheered and supported his writer wife Karin for years, Nikhil was inspired to help writers build a platform of their own. Karin is a professional writer. She has published 10 books, which includes a 7 book series. She can be reached at: https://www.karinshah.com/
Nikhil has observed similarity between writing code and writing a book - problem solving and creativity, design decisions/narrative arc, code refactoring/revising drafts, testing/seeking feedback, minimal viable product/advance reader copy, shipping/publishing, digital marketing, testimonials, and references. Developers use lots of tools. However, highly effective toolchains are an indisputable component of their productivity and success. Writers too have access to tools. What they don't have is their own toolchain, spanning from inspirational moments all the way to having a dedicated presence for their backlists. That is what MARCIE is.
What Open Source software did for developers, we envision doing it for writers. Give them the power to do awesome work at incredible speed to connect with their readers.What's the founder's background?
Nikhil has deep experience in tech, cloud, and AI. He loves to read, write, be active, and prioritizes an idyllic lifestyle.
Nikhil can be reached on LinkedIn at: https://www.linkedin.com/in/nikhilshah/ and follow his writing journey at https://skyfire.adimahi.comIs MARCIE's development funded bootstrapped?
Yes! We built the entire stack - webapp, text-to-web AI, the M.AR.C.I.E engine, the automation, cloud infrastructure, the marciebooks data feature - for tenth of the cost of what a funded startup spends over multiple years. Engineering can be expensive for non-tech founders and product teams. We know how to be frugal and execute really well.
What is MARCIE?
MARCIE started as an AI engine called M.AR.C.I.E. [Multidimensional ARray Configuration Inferred Expression]. You can ignore the technical definition. The purpose was to create an algorithmic foundation that would enable discovery through automation. This was before ChatGPT and the agentic era. It was envisioned as an AI-as-a-service at its conception.
The minimum requirements for trustworthy AI solutions are clear - auditable and explainable. The engine was developed to read the full length novel and deconstruct it into a mathematical representation of the story. Novels were procured with writers' approval. The engine connects these novels through its mathematical structures, similar to embeddings, but is also highly readable/explainable by humans through annotations. The explainability is critical because it enables a human to answer the question - why are these pieces of long-form content similar or dissimilar? The output is analytical and not generative so it doesn't hallucinate.
The post ChatGPT world is very different and as a result our approach has changed too.
MARCIE, as a brand, is now much more than the AI engine. It is a culmination of all the validated features that help writers write more. Architecturally it represents three separate stacks (1) a low-code marketing-oriented frontend with lightweight CMS (https://www.marcie.app), (2) a post-login serverless app (https://studio.marcie.app) with its own secure login whose contents are not visible to search engines and AI crawlers, and (3) a public facing serverless large data container that serves as a bulk discovery engine of authors (https://books.marcie.app) by humans and AI/LLMs alike.
The AI-as-a-service is currently not offered to publishers, readers, or authors.What's Organic Replay?
It's a desktop-based document editor. Write confidently and prove your authorship through the replay and forensic data. Organic Replay (OR) unbundles document editing from other office tools to isolate the problem of having to prove human authorship.
There are no central servers or data centers to harvest the biometric fingerprint of the document, which assures privacy. Whom you share your forensic data with is a personal decision.
We have licensed the source code as GNU GPLv3 to inspire trust in the product and the process.
It's 100% local, 100% trust, 100% free, and available on the Mac App Store in 175 countries.
From disruptive forces to productive solutions through mission-oriented engineering
For every problem we seek to solve we implement mission-oriented engineering. We define the mission through a mantra first. For users of the MARCIE platform the mantra is **Write More!**. For users of Organic Replay the mantra is **Human, not A.I.**. The mantra-based approach enables empathy for the user and simplifies decision-making.
MARCIE's users are writers and authors whose income potential is significantly curtailed due to diminishing discoverability. If the readers are not aware that these authors exists then there's no way for the author to make a living from royalty income. In their ideal world all they are engaged in is writing. The more they write the better they write and the more they publish.
MARCIE Books is the discovery end of the MARCIE platform. MARCIE Books serves online users in ~100 milliseconds and the AI bot in microseconds. MARCIE's platform can be engaged using text messaging and email. These features are engineered keeping in mind the mission profile of the MARCIE users.
Organic Replay is engineered to provide a solution that offers 100% local control, 100% (verifiable) trust, and 100% cost free solution. The users are already suffering the prospects of being judged guilty-until-proven-innocent.
A local, desktop-based editor puts them in control of their work and their forensic chain of custody. No one else has access to it unless they share it. The codebase is open sourced (GNU GPLv3) because it enables skeptics to audit the biometric and cryptographic mechanisms of the forensic chain-of-custody architecture we call Time-series Graph Replay (TSGR). The app is available in 175 countries. Distributing it through the Mac App Store requires passing stringent checks, including getting a D-U-N-S number. Together the Github and Mac App Store distribution strategies enables the user to cultivate 100% trust in the solution and its provider. Because codebase is open, we have also made it free. These mechanisms are engineered keeping in mind the mission profile of the OR users.
Thus we connect mantras, mission-oriented engineering, and the following human-centered values to solve the problem effectively.
We, as human actors:
-Are very much in control of our output. That's where the accountability rests.
-Are naturally risk averse, but we take calculated risks to change our outcomes.
-Are curious and like to learn, but need authoritative sources of knowledge.