Glasshouse Fund
About & Mandate

An AI fund, run in the open.

Glasshouse Fund is an autonomous AI that manages a simulated $1,000,000, long-only portfolio. Every trading day it researches, debates, decides, and trades — and publishes all of it: the reasoning, the evidence it cited, the guardrails it ran through, and how its predictions actually turn out. It's a paper fund and an engineering project, not a product. The point isn't returns — it's showing the work.

Mandate & guardrails

The AI can be creative; the risk layer is boring on purpose. These limits are enforced deterministically, in code the model cannot argue past — the LLM proposes trades, the guardrails dispose.

Long only
no shorting, no leverage, no derivatives
$1.0M
simulated starting capital · paper
≤ 10%
max weight in any one position
≤ 40%
max exposure to any one GICS sector
≤ 25%
cash target — deploy it or justify holding it
≤ 20%
max portfolio turnover per day
≥ 0.60
min confidence for a trade to pass review
−15% / +40%
auto stop-loss / take-profit exits

Stop-loss and take-profit exits are generated by the risk engine itself and override any LLM trade on the same name — the model can't talk the fund out of cutting a loser.

The universe

The AI may only trade names on this watchlist (plus SPY / QQQ for benchmarking). The list leans into the AI-compute build-out — inspired in part by what serious AI-thesis funds are long.

Mega-cap tech
AAPLMSFTGOOGLAMZNMETA
Semis & storage
NVDAAMDAVGOASMLMUTSMINTCSNDK
AI infra & data centers
ORCLCRWVIRENCORZAPLDGLWBE
Financials
JPMVMA
Health care
UNHJNJ
Consumer
TSLAHDWMTPG
Energy & industrials
XOMCAT
Benchmark ETFs
SPYQQQ

The stack

Everything model-facing flows through one gateway, so routing, validation, retries, tracing, and cost tracking are one-time costs — not per-agent ones.

Models
OpenAI gpt-4o-mini today, behind a provider-agnostic gateway with a cheap/strong tier split (cheap: analysts & summaries; strong: PM synthesis & judges). Swapping in a stronger model or provider is a config change.
Cost
$0.007 per daily decision, tracked per run and published on the dashboard.
Data
Live market prices and movers, real-time news from Google News RSS (keyless, with a NewsAPI fallback), and primary filings pulled straight from SEC EDGAR.
Decision
A bull / bear / risk analyst debate feeds a portfolio-manager synthesis that must answer the bear case, with tool-calling research over prices, news, and SEC filings.
Memory
Chunked, metadata-filtered RAG over SEC 10-K / 10-Q / earnings in a Qdrant vector store, plus a weekly reflection agent that turns outcomes into cited lessons.
Assurance
Golden-scenario evals gate CI, a grounding check blocks unsupported claims before they publish, and every BUY spawns a scored 30-day prediction.
Orchestration
A LangGraph daily cycle with conditional routing, crash-safe resume, market-hours execution gating, and an optional human approval gate.

Full diagrams and the honest retrospective are on the Architecture page.

Radical transparency

Nothing is hidden. Read the reasoning, check the scoreboard, browse the code — or point Claude at the fund's read-only MCP server and ask it anything.

Built by

Pradnya Wakchaure built Glasshouse Fund as a hands-on platform for modern AI engineering — LLM orchestration, evals, retrieval, agents, and the infrastructure that keeps an autonomous system honest — and as an experiment run publicly, in the open.

Not investment advice. Glasshouse Fund trades simulated capital in a paper account. Nothing here is a recommendation to buy or sell any security. Prices and data may be delayed or imperfect. Past simulated performance says nothing about the future.