a23one is a 23-product, one-network system built to compound. With initial grants from Google Cloud and Cloudflare, work began in '23 that eventually led to a more ambitious vision; building product that builds product. The guiding belief is that demand for product will scale with the ability to produce said product.
Across five acts from Genesis to Terminus, each act lifts the next: product that powers product, product that builds product, product that owns product. Two hallmark consumer applications have emerged along the way — brainful, which expands the capacity to learn and do, and tidalsight, which produces high-performance financial portfolios generating outsized returns. 500K+ lines of code power the network, processing 4 million monthly requests with thousands of weekly active users.
The origins of the network.
The introductory product: brainful, a next generation thought processor for learning and research that amplifies R&D for new products.
win '22An integrated thinking environment — where writing, research, and discussion live together. 75+ orchestrated AI models, 114 languages, millions of monthly requests.
product that powers product
The digital plumbing powering the next generation of AI-first product.
win '25With growing concerns of exposing sensitive data to third-party services, llmshield is a context-preserving open-source Python library for preventing PII leakage in AI applications.
Spun out of brainful's AI layer to meet growing demand for robust and fault-tolerant AI infrastructure. A single OpenAI-compatible API endpoint routing to dozens of LLM providers and endpoints.
Regular embeddings have limitations in security, privacy, and retrieval accuracy. polyembed aims to tackle all three through local inference, homomorphic embeddings, and systematic context injection for higher-accuracy retrieval.
Securities analytics, insights, and ratings platform. Multi-agent AI continuously analyses fundamentals, social sentiment, and macro signals across equities, crypto, commodities, and currencies — delivering conviction-rated trading intelligence with integrated algorithmic execution.
ML has been a traditionally gatekept field. mlpotion aims to democratise ML models by providing no-code end-to-end ML pipelines as a service. This allows you to distil AI or proprietary data into lightweight prediction models that are exposed as APIs or downloadable models.
As agentic development accelerates the volume of code shipped across a23one, the burden of release notes, documentation, and roadmap planning grows with it. releasetape automates all three — one-click setup per repository, continuous output.
Cross-product link intelligence and attribution. Every link across an organisation's network becomes a first-party data point. From tracking user journeys to understanding feature heatmaps, oburai helps teams and orgs make data-driven decisions for product.
product that builds product
The recursive engine where products author products.
spr '26Prompt to production. In response to a growing necessity for idea and feasibility validation, spawnpod builds product using the agent-first a23stack tailored for high-velocity and production-grade prototyping — from GitHub repo to live deployment previews.
Grafted AI for complex software. Config as code in a single Python file — the chat loop, memory, streaming, and observability run on agentgraft, while tools and actions execute inside the host application via signed webhooks. Drop in a script tag, expose your functions, ship intelligent agents in hours.
product that owns product
Where product holds its own capital.
the end of an era
The 23rd and terminatory product.