Building an AI stock analyst with Kimi K3 + GPT-6 Astra: $450 does $60,000 of research
nevrekaraishwa2 · x · 2026-09-21
A walkthrough of an 'AI market analyst' built on Kimi K3 and GPT-6 Astra, claiming $450 of compute does what would cost $60,000/day in research.
The key trick is a funnel, not one giant model:
- Step 1: plain code does the math first, zero LLM cost
- Step 2: rules + a light classifier cut 10,000 stocks to 150
- Step 3: Kimi K3 reads eight quarters at once, keeps the 40 that don't add up
- Step 4: GPT-6 Astra runs bull-vs-bear debates, every claim needs a source
- Step 5: a human confirms anything that executes
The output deliberately issues no BUY ratings or predictions — only what changed and what contradicts it. Full architecture with formulas linked.
More from coding & agent
- LlamaIndex launches DocJev: OSS doc classification 6x faster than gpt-5.6-luna — multiply_matrix · 2026-09-21
- Ante v0.2.0 ships a stream protocol decoupling agent engines from interfaces — JaynitMakwana · 2026-09-21
- Rethinking classifiers: conversation+tool templates as special cases of inference-time specs — austinvhuang · 2026-09-21
- Analyst predicts Meta will pay Amazon for agent access, spawning a new agent business model — signulll · 2026-09-21
- Arena: coding-agent harnesses show up to 5x cost differences at similar success rates — thione · 2026-09-21
- Recap: Anthropic merged Claude chat and Cowork, redesigned Claude Code Projects — thione · 2026-09-21