Private beta · AI research for institutional investors

Your research desk,
with perfect memory.

10kgraph reads every filing, debates every thesis, and sizes every position — grounded in a knowledge graph that never forgets what your fund has already figured out.

Live thesis — NVDA
live
Verdict
NVDALONG
Size3.2% NAV
Conviction0.81
Horizon6–9 months
Reviewedjust now
PM copilot

“Moat durability outweighs multiple risk at this cycle position. Half‑Kelly sizing given drawdown regime.”

Supporting — cited
  • WIKICUDA ecosystem compounding since 2007 (19 years).
  • NVDABlackwell ~70% of DC compute in Q1 FY26; GB300 crosses GB200.
  • GTCJensen signals $1T+ AI infra TAM through 2027.
Objections — considered
  • AWSTrainium 3 lands at ~50% better price/perf vs H100 / B200.
  • INTLCustom ASICs capture 15–25% of 2026 accelerator share.
  • FUTNVDA share drifts from 95% peak to ~75–80% by EOY 2026.
The product

Three things your analysts do.
Now with receipts.

10kgraph doesn’t replace your judgment. It gives you a desk that reads faster, forgets less, and argues harder — so you can underwrite with conviction.

01 · Read

Reads every filing before you open your laptop.

Every 10-K, 10-Q, 8-K, earnings call, and comment letter is ingested, structured, and tied into the knowledge graph — so your theses are grounded in the whole record, not the paragraph you remembered.

Outcome

First-read advantage on every ticker you follow.

02 · Debate

Stress-tests every thesis before it’s funded.

A structured bull case and bear case are built from separate evidence, then reconciled by a PM copilot. The adversarial transcript is the artefact you underwrite — not a blob of prose.

Outcome

Conviction that survives its strongest objection.

03 · Size

Sizes with conviction, not vibes.

Position sizing is Kelly-bounded, factor-aware, and regime-adjusted. Every recommendation cites the risk bounds it respects and the trades it correlates with.

Outcome

Book-level risk that matches your book-level view.

Ticker Knowledge Graph

Every ticker is a neighbourhood,
not a PDF.

Moats, customers, suppliers, tailwinds and risks — linked with provenance, updated on every filing. Hover the nodes to explore NVDA’s current graph.

Live neighbourhood· NVDA
NVDA$4.9T capCUDA19yr ecosystemBlackwell~70% DC computeTSMC 3nmsole fabHyperscalers$700B '26 capexSovereign AISaudi · UAE · JP · FRAMD MI300X$15K vs H100 $32KCustom ASICs15–25% '26 share
Hover any node to see its edges and neighbours.8 nodes · 9 edges
Trust & control

Your edge stays inside your walls.

Your data, your graph

Tenant-isolated, row-level security. Your reasoning is never mixed with another fund’s, never used to train a shared model.

Every claim is citable

Every bullet, every verdict, every sized position traces to a specific paragraph in a specific document. Hover to read the source.

Deterministic risk gate

Position limits, sector caps, and conviction floors are enforced outside the language model. No hallucinated fills, ever.

Why us

Incumbents search public content.
10kgraph remembers your edge.

Published research is a commodity. Your fund’s accumulated reasoning is not. 10kgraph puts the latter first.

Capability
AlphaSense
BlueFlame
Canalyst
10kgraph
Searches your private research corpus
KPIs auto-extracted from filings
Language & tone drift detection
Full audit trail for compliance
Adversarial bull / bear debate
Temporal knowledge graph w/ provenance
Cross-asset regime monitor
Kelly position sizer + risk gate
Idea journal w/ AI retrospective
Paper + live execution

Questions we hear every demo

How is this different from a ChatGPT wrapper?
ChatGPT has no persistent knowledge graph, no adversarial debate, no deterministic risk gate, no audit trail, and no execution layer. 10kgraph has all five, and your reasoning compounds inside the graph over time.
Can I use my own models and keys?
Yes. On Enterprise, bring your own reasoning model and keys. The orchestration and graph are model-agnostic.
What about compliance and audit?
Every claim, every decision, every fill persists with inputs, outputs, timestamps, and source citations. Export to CSV or stream to your SIEM. Enterprise ships a SOC 2-style evidence pack.
Hosted, VPC, or on-prem?
Your choice. Beta users are onboarded onto a dedicated VPC. Enterprise deploys into your account with your keys.
When does it come out of beta?
We’re onboarding funds a few at a time to ensure every design partner gets hands-on founder time. Request access and we’ll share the current timeline.
Currently in private beta

The next great fund
starts with a better graph.

See what your research desk looks like when it never forgets.