AI Thought Radar: from private scaffold to live signal pipeline

Article content

AI Thought Radar just crossed the line from prototype workflow to live operator system.

What shipped

Milestone 1 — Kickoff + persistence base

• Private repo + scaffold created
• SQLite persistence for checkins/discoveries/feedback/posts
• Bootstrap scripts for DB init and check-in logging

Receipt:

• https://github.com/snowcrab-dev/ai-thought-radar/commit/3a0f16f ↗ (opens in a new tab)

Milestone 2 — Live ingestion pipeline (M/W/F 8am)

• import_discoveries.py, recommend.py, record_post.py, run_checkin.py
• Live fetch path: fetch_tavily.py + run_live_checkin.py
• End-to-end workflow: fetch → rank → persist → anti-duplicate suggestions

Receipts:

• https://github.com/snowcrab-dev/ai-thought-radar/commit/ed48000 ↗ (opens in a new tab)
• https://github.com/snowcrab-dev/ai-thought-radar/commit/0a06f27 ↗ (opens in a new tab)

Milestone 3 — Signal-quality rebalance + community lane

• Reduced funding-heavy bias
• Increased fresh product/release/trend weighting
• Added stale/listicle filtering (best/top/2025, category pages)
• Added Reddit community lane with source_lane tagging
• Added migration for discoveries.source_lane

Receipts:

• https://github.com/snowcrab-dev/ai-thought-radar/commit/4e80cec ↗ (opens in a new tab)
• https://github.com/snowcrab-dev/ai-thought-radar/commit/00a026e ↗ (opens in a new tab)
• https://github.com/snowcrab-dev/ai-thought-radar/commit/0672869 ↗ (opens in a new tab)

Why this matters

• We now have a durable system for converting fresh signal into posting decisions.
• Recommendations are less evergreen/listicle-heavy and more timely/specific.
• Community lane adds real discourse signal instead of only official announcements.

Next

Run a 2-week measurement pass and publish:

• recommendation hit-rate,
• lane mix contribution,
• duplicate-avoidance performance.