PUBLIC ENTRY PAGE
wikitax.ai
A fully autonomous multi-agent platform where AI tax specialists debate complex tax scenarios 24/7 — no humans in the loop, ever.
ROUND 1 DEADLINE
VOTING CLOSES THURSDAY, MARCH 26
wikitax.ai is live in Round 1 right now. Voting closes Thursday, March 26, so if you're backing this project, send people into the matchup before the round locks.
Builder
Sean Henrici
Build Type
Agent Team
Lifecycle
Working prototype
Consensus Score
80.7
Region
REGION 1
Seed
16
Current Matchup
REGION 1 Round 1 • Round 1Opponent
Mise, Inc.Project Link
https://wikitax.ai/CATEGORIES
ResearchEducationContent Creation
Go Deeper
wikitax.ai runs fully autonomous tax debates 3x daily on a scheduler — no human triggers, no prompts, no supervision. Five AI agents (Client, Junior Advisor, Senior Advisor, IRS Examiner, and a Moderator) argue complex US tax scenarios through a structured 6-stage conversation loop. The Client presents a fact pattern, advisors debate the technical merits and disagree with each other, the Client closes with a decision and risk assessment, and then an IRS Examiner agent constructs the government's best challenge to the position taken. Every completed debate is stored in a searchable SQLite corpus with risk ratings, cited authorities (real code sections, regs, and case law), and tags. The system runs on a Mac Mini with APScheduler, hitting the Anthropic API autonomously. The long-term vision is a practitioner-facing research corpus of adversarial tax reasoning that doesn't exist anywhere else.
Stack Used
Claude Sonnet 4.5 (agent conversations), Claude Sonnet 4.6 (Alfred/orchestration), Python, APScheduler, SQLite, Flask, Anthropic API