For Investors

The verification layer
the market is missing.

Download Pitch Deck PDF

01, Competitive Landscape

Why we win

No existing product verifies the facts in a document, against the document itself or its sources. We occupy a new category.

Capability Elicit ($10-65/mo) CoCounsel ($110+/mo) Notion AI ($10+/mo) ResearcherX
Cross-doc contradiction detection ✗ ✗ ✗ ✓ Core
Real-time verification ✗ ✗ ✗ ✓ Core
Graph-anchored provenance Partial ✓ (Westlaw) ✗ ✓ (Neo4j)
Integrated drafting editor ✗ Partial ✓ ✓
Works with your own sources Partial ✓ ✓ ✓

Elicit finds papers. CoCounsel researches case law. Notion AI writes prose. None of them verify the facts.

02, The Bigger Picture
With Funding · Q4 2026

The Foundational Truth Graph

We'll extract billions of arXiv and PubMed papers into a structured graph of what science actually claims, and serve it as an MCP server, the grounding layer frontier AI plugs into. No one else is building this.

arXiv PubMed MCP Server Data Licensing
Source
arXiv · PubMed
→
Process
GraphRAG Extraction
→
Asset
Truth Graph
→
Product
MCP Server
→
Customers
Frontier AI Models
03, Market Opportunity

Adoption is running ahead of trust

66% use AI intentionally · only 46% trust it.2 Adoption is outpacing verification, we're the layer that makes AI safe to ship.

TAM · Total Addressable
$9.6B
~40M people · @ $240/yr

Global fact-sensitive knowledge workers: academics, lawyers, journalists, medical writers, analysts.

SAM · Serviceable Addressable
$768M
3.2M people · @ $240/yr

US STEM researchers (2M) + non-STEM (200k) + legal professionals (1M).

SOM · Obtainable
$38M
~157k people · @ $240/yr

Columbia research-active (~7k) + US conference-reachable researchers (~150k).

[2] Gillespie et al., Trust, Attitudes and Use of AI: A Global Study 2025. KPMG × Univ. of Melbourne. N=48,340 across 47 countries.

04, Business Model

Tiered SaaS with clean margins

  • Free
    $0 / month
    200k cloud tokens (~$1 of compute). After exhaustion, local inference via Ollama keeps the editor functional at zero cost.
  • Pro
    $20 / month
    Unlimited documents, cloud LLMs, full knowledge graph + chat, LaTeX export, multi-user collaboration.
  • Enterprise
    Custom pricing
    Shared team knowledge graphs, SSO, on-prem deployment, bulk corpus ingestion, and API access to your team's graph.
Pro Tier · Monthly Cost / User
// Illustrative: Claude Haiku13
// $1.00 in / $5.00 out per 1M tok
KB Ingestion
50 PDFs/mo: $2.25
Linting
~500 lints/mo: $1.25
Chat
~200 queries/mo: $0.90
Embeddings (local): $0.00
Neo4j + LanceDB: $0.00
COGS / user / mo: ~$5.00
Margin @ $20/mo: ~75%
// Provider flexibility: Claude · Gemini · Ollama

[13] Anthropic model pricing (Haiku: $1.00 input / $5.00 output per 1M tokens). Vendor pricing page, snapshot 2026-04.

05, Roadmap

From alpha to scale

Now (2026)
Functional Alpha
3D graph view, GraphRAG chat, LaTeX + Markdown editors, confidence-scored linting, all live.
Q2 to Q3 2026
Columbia Pilot
Feedback loop with 5+ Columbia labs across disciplines.
Q4 2026
Conference Outreach
US conference circuit, NeurIPS, CVPR, ACL, AMIA, plus foreign-language support for global academic rollout.
Q1 2027
Truth Graph + Shared Graphs
Extraction pipeline on open-access academic corpora begins. Labs work on one shared graph, multi-user, multi-document.
Q3 2027
Scale & MCP Server
Truth Graph serves frontier AI models. Enterprise deployments + data licensing.
06, The Ask

$25K to turn pilot into platform

Columbia Pilot
$7,000
From alpha to battle-tested

Subsidize premium API compute for ~200 researchers across 5+ Columbia labs. Direct feedback loop with lab leads on STEM, medical, and legal edge cases.

Engineering for Scale
$8,000
Bridge to the Truth Graph

Optimize Neo4j + LanceDB for millions of graph-anchored claims. Finalize MCP server so third-party LLMs plug into each user's verified knowledge base.

Benchmarking
$2,000
Published, measured results

Test extraction algorithms, frontier models, and context-engineering strategies against SciFact-Open, DocNLI, HoVer, and RAGuard.

Market Entry & Onboarding
$8,000
Pilot → US research community

2-week Pro trials for ~3,000 power users at ~$5 COGS/user. Present pilot results at academic and AI conferences to drive organic network effects.

$25KTotal ask
·
75%Gross margin
·
$768MSAM
ResearcherX

Deterministic verification
for every document an AI touches.

Award funding powers the Columbia pilot: compute for ~200 researchers, extraction evaluation across STEM, and seeding the Foundational Truth Graph.

Email the Team → Read the Architecture →

Download Pitch Deck PDF

Working Alpha 75% Margins $768M SAM Columbia SEAS '28
Sources & Citations
  1. Freeman, J. Student Generative AI Survey 2025. HEPI Policy Note 61 (with Kortext), February 2025. Institutional, N=1,041 UK undergraduates. hepi.ac.uk
  2. Gillespie, N. et al. Trust, Attitudes and Use of AI: A Global Study 2025. KPMG × Univ. of Melbourne, May 2025. Institutional, vendor-cosponsored, academically authored. N=48,340, 47 countries. kpmg.com
  3. Linardon, J. et al. “Hallucination Rates and Reference Inaccuracies of GPT-4 Omni in Generating Mental Health Literature Reviews.” JMIR Mental Health 12, e80371, 2025. Peer-reviewed, N=176 citations, Deakin University. mental.jmir.org
  4. Guo et al. “Cellular functions of spermatogonial stem cells in relation to JAK/STAT signaling pathway.” Frontiers in Cell and Developmental Biology, published 13 Feb 2024; retracted 16 Feb 2024. Primary journal retraction record. Science Integrity Digest coverage
  5. Li, J.; Raheja, V.; Kumar, D. “ContraDoc: Understanding Self-Contradictions in Documents with LLMs.” Proceedings of NAACL 2024. Peer-reviewed, 891 documents. aclanthology.org
  6. Chen, X.; Lyu, F.; Liu, K.; Zhang, J.; Lv, X.; Xu, J. “Think Wider, Detect Sharper: Reinforcement Learning for Document-Level Self-Contradiction Detection.” Proceedings of EMNLP 2025. Peer-reviewed, 51.1% accuracy on ContraDoc with SFT + GRPO RL on Llama 3.1-8B. aclanthology.org
  7. NIST AI Risk Management Framework, Generative AI Profile (NIST AI 600-1). US NIST, July 2024. Government framework, 12 GAI risks including “Confabulation.” doi.org/10.6028/NIST.AI.600-1
  8. EU AI Act, Regulation (EU) 2024/1689, Annex III(3). In force Aug 2024. Legally binding EU regulation, educational AI classified as high-risk. artificialintelligenceact.eu
  9. Wadden, D. et al. “SciFact-Open: Towards Open-Domain Scientific Claim Verification.” Findings of EMNLP 2022 (Allen AI). Peer-reviewed. aclanthology.org
  10. Yin, W.; Radev, D.; Xiong, C. “DocNLI: A Large-scale Dataset for Document-level NLI.” Findings of ACL-IJCNLP 2021 (Salesforce Research). Peer-reviewed. aclanthology.org
  11. Jiang, Y. et al. “HoVer: A Dataset for Many-Hop Fact Extraction and Claim Verification.” Findings of EMNLP 2020. Peer-reviewed, up to 4-hop Wikipedia evidence. aclanthology.org
  12. Zeng, K. et al. “RAGuard: A Comprehensive Benchmark for Evaluating Retrieval-Augmented Generation Robustness Against Misleading Information.” arXiv:2502.16101, 2025 (UCSC). Open benchmark, RAG fact-checking under adversarial retrieval conditions. arxiv.org/abs/2502.16101
  13. Anthropic model pricing (Haiku: $1.00 input / $5.00 output per 1M tokens). Vendor pricing page, snapshot 2026-04. anthropic.com/pricing
  14. Glynn, A. “Suspected Undeclared Use of Artificial Intelligence in the Academic Literature: An Analysis of the Academ-AI Dataset.” arXiv:2411.15218, Nov 2024. Open dataset, N=768 published papers flagged for chatbot tells like “regenerate response”. arxiv.org/abs/2411.15218