Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons
Dry-Comfortable-2678 · reddit · 2026-10-23
A months-long build: an agent that researches sports matchups, compares its read to prediction-market implied probabilities, and can place positions onchain with human approval by default.
Loop: plain-English intent → research (stats, injuries, form, news) → read market price as implied probability and compare → structured proposal → user approval (autonomous mode optional) → execute from the agent's own wallet → settle in USDC.
Stack: Claude for reasoning, Circle Agent Wallets, YES/NO outcome tokens, Arc settlement.
Design decisions that mattered:
- Separate thinking from acting: a validation layer checks market open, position limits and balance before anything signs
- Guardrails live in code, not prompts — "a prompt is a suggestion, not a guardrail"
- Approval on by default; autonomy is earned by being visibly right
- Price the research: most runs never become a trade
Still hard: stale injury news, calibration (proving 60% means 60%), thin liquidity. Advice for money-touching agents: start with the proposal format and validation layer, not the prompt.
More from coding & agent
- We tested 12 LLM classifiers — the new classifier hype is LARP — KlausCodes · 2026-10-23
- Dev builds ESP32 desk display showing Claude Code and Codex agent status via local Go helper — tonylook · 2026-10-23
- Let your agent read its own traces: AgentsView + history tool debugging workflow — dbreunig · 2026-10-23
- YC-backed Mireye launches land search API + MCP with 160M parcels for AI agents — ycombinator · 2026-10-23
- Testing 12 decision systems: the new LARP hype around classifiers — KlausCodes · 2026-10-23
- Monid launches 'Product Hunt for agents,' ranking tools by real agent calls — claudeai · 2026-10-23