Adrian's seed content — 2026-05-31

Raw content from Adrian for seeding Mr Paint's intelligence database.

Files received

File	Type	Use
`audio-1-brinsmead.txt`	Voice note transcript	L3 kb_chunks · brand vocabulary
`audio-2-bungalow-eves.txt`	Voice note transcript	L3 kb_chunks · brand vocabulary
`audio-3-bungalow-queenslander.txt`	Voice note transcript	L3 kb_chunks · brand vocabulary
`video-1-no-speech.txt`	Video had no speech (dramatic music only) — visual content reference only	—
`photos.md`	3 photo references with captions	L3 kb_chunks · job evidence

Important — spoken vs. written voice

Voice notes are spoken voice — narration style, real-time, walks-you-through. This is different from Adrian's writing voice. Spoken samples are useful for:

L3 kb_chunks (what he knows + how he describes jobs) ✓
L1 vocabulary signals (brand names, products, suburb references) ✓
L1 specificity baseline (he names tools, grit numbers, products — model should match this level of specificity) ✓

NOT useful as Layer 2 written voice samples — too narration-y for GBP / IG / email content. Still need 10–15 of his actual written content (GBP posts, customer emails) for proper L2.

What I extracted

See extracted-structured.yaml for the L1 structured-fact signals pulled from these transcripts:

3 paint brand names (Sikkens, Dulux, Sikkens specific product line)
4 tool brands (Festool, Makita, Gerni / pressure washer, Rotex sander)
2 confirmed Cairns suburbs (Brinsmead, Bungalow)
1 architectural style (high-set Queenslander)
1 service pattern (full weatherboard strip → planar → sand → seal → top coats)

This is exactly the kind of detail the strategist needs to sound like he knows the business.