OpenAI 2026 hackathon

Mosaic

A pluggable framework for building safe, observable multi-agent systems on OpenAI — where agents interpret, analyze, and brief, and humans always hold the decision.

Solo project by Narayan SS · 0 likes · 0 comments

Archive position — measured, not model output

0 likes on Devpost

2,264 of the 7,856 archived projects have more likes, and 5,592 share exactly 0 — so this project's #5,397 place in the like-ranked listing is a tie-break inside that group, not a ranking.

Projects (log scale)

1
10
100
1k
10k
05,592
11,758
2285
3–4132
5–975
10+14

Likes on Devpost. ▲ marks this project's group.

Show the figures
LikesProjectsShare of archive
05,59271.2%
11,75822.4%
22853.6%
3–41321.7%
5–9751.0%
10+140.2%
Devpost like counts for all 7,856 archived projects, captured when this archive was built.

Executive Summary

What the company appears to be

Mosaic is a self-reported framework for building multi-agent AI systems on OpenAI, designed to support high-stakes decision-making where humans always hold the final decision. It is described as a pluggable system that enables agents to interpret, analyze, and brief, while ensuring no autonomous execution occurs.

What changed

The author states they built Mosaic in four days, starting from an idea and iterating quickly. The framework is positioned as a foundational tool for building safe, observable multi-agent systems, with an emphasis on extensibility and deterministic replay capabilities.

Single most important open question

Is there evidence of real-world usage or adoption beyond the author’s solo development? The description contains no claims about customers, revenue, traction or product-market fit — only a self-reported technical prototype.

Back to contents

What The Product Actually Is

The description states that Mosaic is a pluggable framework for building multi-agent systems on OpenAI. It includes three agents:

  • Luna (interprets raw events)
  • Terra (analyzes what is happening)
  • Sol (briefs the operator)

These agents feed into a single, live operating picture. The system runs on OpenAI when a key is present, replays banked responses deterministically when not, and falls back to fixtures offline.

It uses:

  • Go for backend
  • Postgres for event spine
  • OpenAI models (with strict structured output)
  • Svelte UI
  • Docker, Supabase, Cloud Run

The framework is described as not a demo, but a tool others can adapt without fighting over ownership.

Inference Mosaic appears to be a prototype or early-stage framework built for developers or engineers working in high-stakes domains like cybersecurity. It is not a finished product or SaaS offering, but rather an open-source or reusable architecture.

Back to contents

Positioning & Claim Evolution

The author positions Mosaic as:

  • A framework, not a product
  • Designed to supplement humans, not replace them
  • Built with safety and observability in mind
  • Focused on high-stakes domains (e.g., emergency operations, cybersecurity)

Key claims include:

  • AI should "take the noise off our plate" so we can do what only humans can.
  • The system ensures no autonomous execution — this is “baked into the architecture.”
  • It supports deterministic replay for demos and testing.

Inference The positioning reflects a belief in human-in-the-loop AI systems, especially in domains where errors are costly. This is not a commercial product, but a technical approach or architectural pattern.

Back to contents

Target Customer & ICP

The description does not name specific customers or personas. However, the author implies:

  • Developers or engineers working with multi-agent systems
  • Teams in high-stakes domains (e.g., cybersecurity, emergency response)
  • Organizations looking to build safe AI systems that avoid autonomous actions

There is no evidence of a defined ICP beyond these implied use cases.

Inference The target audience likely includes technical teams building AI applications, particularly those in regulated or safety-critical industries. However, the lack of customer data means this remains speculative.

Back to contents

Business Model & Pricing Evidence

No business model or pricing information is provided in the description. The framework is described as a pluggable system intended for others to use and adapt — not sold directly.

Inference There is no evidence of monetization, licensing, or revenue streams. If Mosaic becomes a product, it would likely be through developer adoption or enterprise licensing, but this is not stated.

Back to contents

Technical & Delivery Signals

The system is built using:

  • Go (backend)
  • Postgres (event spine)
  • OpenAI models
  • Svelte UI
  • Supabase, Cloud Run for deployment
  • Playwright for end-to-end testing

Key technical features include:

  • Pluggable architecture with extensibility as a design constraint
  • Deterministic replay using cassette layer
  • Structured output from OpenAI models
  • Support for offline fallbacks and deterministic behavior

Inference The system shows early signs of thoughtful engineering, particularly around pluggability and deterministic execution. However, it is not yet proven in production or at scale.

Back to contents

Traction & Maturity Signals

There is no evidence of:

  • Customers
  • Revenue
  • Product-market fit
  • Adoption beyond the author’s own development
  • Any form of traction or usage metrics

The project was built solo in four days and submitted to a hackathon. It is described as a framework, not a product.

Inference This is an early-stage prototype with no demonstrated traction or maturity. The author's goal was to build something reusable, but there’s no indication of real-world usage or impact.

Back to contents

Competitive Context

The description does not mention competitors or similar tools in the market. It is unclear whether Mosaic is positioned against other multi-agent frameworks, AI orchestration platforms, or decision-support systems.

Inference There is no evidence of a competitive landscape. The project appears to be self-contained and unanchored to existing products or markets.

Back to contents

Key Risks & Red Flags

  • No traction or adoption: The framework has not been used by others beyond the author.
  • Unproven in production: It is described as a prototype, not a tested system.
  • Solo development: Only one person built it; no team or institutional support.
  • No commercialization plan: No pricing, licensing, or monetization strategy is evident.
  • Limited evidence of real-world relevance: The domain use case (cybersecurity, emergency operations) is mentioned, but not validated.

Inference The project lacks any commercial viability indicators. It may be a useful concept, but there is no evidence it has moved beyond the idea stage or gained traction.

Back to contents

Diligence Questions To Ask The Founders

  1. What specific high-stakes domains have you tested Mosaic in?
  2. Have you had others adopt or adapt the framework? If so, how?
  3. Is there a plan to monetize or commercialize this framework?
  4. How does Mosaic handle scalability beyond single-agent use cases?
  5. What are the key technical limitations of the current architecture?
  6. Are there any known issues with OpenAI’s structured output mode that affect reliability?

Back to contents

Investment/Partnership Verdict

Not evidenced.

The description provides no evidence of revenue, customers, traction or commercial viability. It is a self-reported prototype built by one person for a hackathon. There is no indication of a product-market fit, team strength, or path to monetization.

This project is best described as an early-stage idea or proof-of-concept, not a viable investment or partnership opportunity at this time.

Confidence Low — based entirely on self-reported information with no external validation.

Back to contents

Source

Submitted to the OpenAI 2026 hackathon on Devpost. Project home on DevPost.

The analysis above was generated by a language model from the project's own one-line description. It is not independent research and contains no verified traction, revenue or customer data.