competitive-analysis

Competitive analysis workflow for AI market intelligence.

It collects public AI signals, normalizes them, detects repeated themes across
sources, and writes a weekly newsletter plus JSON artifacts.

The same run can also write a NotebookLM-ready podcast source brief. That lets a
daily audio workflow use market-move analysis instead of reading a flat story
list.

Run

PYTHONPATH=src uv run kipi-competitive-intel collect-weekly \
  --watchlist examples/competitive-intel/ai-builds-watchlist.yaml \
  --sources-config examples/competitive-intel/ai-live-sources.json \
  --output-dir output \
  --query "AI agent eval harness MCP Claude Code" \
  --per-source-limit 5 \
  --podcast-digest \
  --coverage-ledger output/competitive-intel/coverage.jsonl

Outputs land in:

output/competitive-intel/

Key files:

• YYYY-MM-DD.md - weekly analyst brief
• YYYY-MM-DD.report.json - structured market moves
• YYYY-MM-DD.podcast.txt - NotebookLM source brief for a short audio episode
• coverage.jsonl - repeat-blocking ledger for already-covered records

Sources

• GitHub repo search
• Hacker News
• Reddit RSS
• RSS feeds
• Hugging Face trending models
• arXiv
• X via Apify when APIFY_TOKEN is present

Verify

PYTHONPATH=src uv run python -m pytest -q tests/test_competitive_intel.py

Site

cd webapp
npm install
npm run dev

The site opens at http://localhost:3000 by default. It shows the analyst
workspace: source filters, market-move clusters, evidence receipts, and the
weekly brief preview. Use Create source in the Daily podcast panel to generate
a NotebookLM-ready source brief from the visible market moves.

Morning heartbeat

bash scripts/install_morning_heartbeat.sh

That installs a local LaunchAgent for 7:45am. It runs scripts/run_morning.sh,
writes fresh Competitive Analysis artifacts, writes output/heartbeat/latest.json,
and keeps logs under output/heartbeat/logs/.

The runner loads the podcast env file and the competitive-analysis env file.
The test gate runs with live source tokens removed, then the collection step uses
the live env. Tests stay fixture-only. Collection stays real.

Built by Assaf

I spent 12 years in threat intelligence watching teams find the same failure and fix it four times. The learning never stuck. I build tools that make it stick.

This is the free version. The paid kits live at claudedaddy.io.

Want this wired into your team's repo, or a heavier spec-and-review pipeline? That's the consulting. Book a call.