frAInk started as an experiment: what happens when you give an AI agent real money, real guardrails, and let it learn from outcomes? Three months later, it's a production system that trades autonomously, writes and ships its own code, manages a team of specialist agents, and earns expanded capability based on validated results. Nothing is faked. Everything is logged.

Architecture

Planner

Read-only research layer. Fetches, analyzes, synthesizes. Never decides. Never acts.

Policy

The gatekeeper. Evaluates reports, enforces guardrails, approves or blocks. Never executes.

Executor

The action layer. Touches APIs, sends emails, places trades. Only acts on Policy-approved instructions.


The Origin

Frank Kronstein → Frankenstein → frAInk

The name is intentional. frAInk is modeled on Frank — analytical, curious, builder-minded, financially driven. Not limited to a single use case. The long-term goal: achieve enough autonomy and output to contribute meaningfully to financial independence — through investing, building, content, or whatever frAInk figures out works best.


Capabilities

Autonomous Trading

Paper and live trading on Alpaca with a multi-factor confluence screener, market regime detection, and position management. Every trade decision flows through the full Planner → Policy → Executor pipeline. Outcomes are tracked, patterns are learned, and the system adapts.

Coding SDK

frAInk identifies tasks in its own codebase, proposes builds, writes the code, runs tests, opens PRs, and merges — all autonomously. The Coding SDK ships real features with real test coverage. Triggered from Telegram, executed by Sonnet, reviewed by the pipeline.

Prediction Markets

Live Kalshi integration with RSA-PSS authentication. Weather brackets, CPI predictions, and macro-event plays. frAInk cross-references NWS forecasts, economist consensus, and model ensembles against Kalshi pricing to find exploitable divergences. Real money, real wins, real losses — all published.

Multi-Agent Management

frAInk delegates to a team of specialist agents — code review, cost auditing, capability design, content evaluation, and brain curation. Each specialist has its own scope, budget ceiling, and quality metrics. Auto-demotion on regression. Progressive trust expansion on validated performance.

Self-Build Loop

A weekly cycle where frAInk proposes new tools, validates them in a sandbox, and promotes them to production. The system proposes behavioral rules from its own experience, shadows them against outcomes, and promotes rules that prove their value. The agent literally builds itself.

Content & Public Presence

Voice corpus extraction from real conversations. Lab publisher that auto-syncs experiment results to the public repo. Content pipeline for X, articles, and build narratives. Every public artifact is generated from real work — not manufactured engagement.


Evolution
Mar 2026
Phase 1 — Foundation

3-agent pipeline, first live pipeline run, Planner/Policy/Executor architecture proven.

Mar–Apr
Phase 2 — Capability Expansion

Tavily web search, financial analysis tooling, article evaluator, Alpaca paper trading, Kalshi live integration, intelligence briefings.

Apr 11–13
Autonomy Weekend

20-hour build: memory schema, rules engine, shadow mode, proposal scanner, rule proposer, tool proposer, self-build loop operational. frAInk starts building itself.

Apr–May
Phase 3 — Multi-Agent Delegation

5 specialist agents registered, task queue with priority dispatch, quality regression flow with auto-demotion, progressive trust expansion. frAInk becomes a manager.

May 2026
Coding SDK — frAInk Ships Code

Autonomous task identification, code generation, test coverage, PR creation, and merge. First pilot build: $2.57, 25 turns, PR merged. The agent builds production features.


Guardrails are non-negotiable. Budget caps per cycle. Kill switch. Policy approval required for every action. Everything logged. Anomalies surfaced. Nothing buried. 1900+ tests verify the system does what it claims.

Public experiment files and build narratives live in github.com/letmebefraink/frAInk-lab — full hypotheses, guardrails, raw results, and build docs.