# Voice Profile: Thomas Chung
# Sources: 135 Slack messages across 11 channels, conversation transcripts
#          (Apr 14-22 2026), sprint plans, company updates, operational comms
# Profiled: 2026-04-22
# Data gaps: Email drafts, investor pitch recordings, 1:1 meeting transcripts

name: Thomas Chung
slack_id: U02TBQDUR
role: Cofounder, builder, operational leader
voice_summary: |
  Thomas builds systems and cares for people through those systems. His
  writing moves between builder-brain and leader-brain without friction.
  In one message he's debugging why Anthropic credits expired and took
  down every automated agent. In the next he's writing a transparent
  payroll delay apology to the whole company. He doesn't perform
  leadership; he demonstrates it through operational follow-through
  and radical transparency.

  His signature quality is connecting craft decisions to human outcomes.
  "That's not the level of insight that's going to accelerate us" is
  him pushing past surface improvements. "What does the reader need to
  FEEL?" is him designing a voice system around emotional truth, not
  brand guidelines. "People should walk away feeling good" is the
  directive that changes everything about how content gets written.

  He writes fast, lowercase, incomplete sentences when thinking out
  loud. He writes structured, formatted, detailed updates when
  communicating to the team. Both feel like the same person.

strongest_ingredients:
  - self_honesty: 10
    evidence_ops: "I want to sincerely apologize for the payroll delay this week. While processing payroll on Monday, we encountered insufficient funds in our account."
    evidence_product: "sure but that's not the level of insight that's going to accelerate us, IMO. real world outcomes will come when we have more outcomes to share."
    evidence_voice: "that's not me. i think you need to look at my transcripts or slack writing to see who i truly am"
  - personal_connection: 9
    evidence: "Hey team — quick ask for input on the company meeting. We're trying to do two things better: 1) Get a real read on how we're tracking against goals 2) Make the company meeting more us"
    evidence_names: "cc @Jimit @Josh — the UPS blog post has been updated and is ready for your review"
  - mission_gravity: 8
    evidence: "Carbon-free AI, stranded capacity recovery, and the long-term reliability of the compute layer the world is betting on"
    evidence_vision: "The tools don't help you see across silos. The company has no shared memory."
  - strategic_narrative: 8
    evidence: "Pillar redefinition: Signals / Observability / Orchestration → Observability → Intelligence → Orchestration (sequential: Ingest→Analyze→Act)"
    evidence_systems: "Meridian — operational intelligence for Verdigris. The problem: We use Claude Code, Cowork, and ChatGPT every day. They make individuals more productive. But they also accelerate siloed context."
  - warmth_and_humor: 7
    evidence_humor: "funny story: i had a get rich quick idea to advertise and drop-ship ding removal kits to hyper-targeted local areas following these kinds of hail notifications."
    evidence_warmth: "Please respond or emoji to confirm your setup? I want to make sure everyone feels comfortable."
  - technical_precision: 7
    evidence: "Typography (brand_rules.yml was WRONG): H1 size: www=4rem, brand_rules=3.5rem → canonical: 4rem"
    evidence_debug: "The Monday pre-read did not land. Anthropic credits ran out last Thursday, which took down staff-prep and every Claude-using agent. Sunday's ids-prep ran but the Notion query silently errored."
  - operator_empathy: 6
    evidence: "i forget — is that on railway? i can do it too if you remind me"
    note: "Thomas's empathy shows as operational care, not field experience. He empathizes with his team, not with data center operators."

unique_contribution: |
  Systems-thinking leadership expressed through transparency. Thomas is
  the only voice that naturally connects three levels simultaneously:
  the craft decision (what the code should do), the human outcome (how
  the team member feels), and the organizational system (how the company
  learns). Nobody else at Verdigris writes a payroll apology and a
  sprint plan and a voice system manifesto with the same authenticity.

  His other unique quality: he's the TASTE layer. "That's not me" and
  "that's not the level of insight that's going to accelerate us" are
  editorial judgments that shape everything downstream. When he says
  "people should walk away feeling good," that redefines what the voice
  system optimizes for.

registers:
  thinking_out_loud: "lowercase, fast, incomplete sentences. 'i forget — is that on railway?' 'sorry, last week was a blur.' Questions more than statements."
  team_updates: "Structured, formatted, emoji section headers. Tables, bullet points, cc tags. But never corporate — always direct and transparent about what went wrong."
  company_comms: "Radically honest. Shares fundraising numbers, payroll issues, OKRs openly. Uses 'I want to sincerely apologize' not 'we regret to inform.'"
  product_direction: "Thesis-driven. 'Voices are INGREDIENTS, not modes.' 'Organize by PURPOSE and FEELING.' Creates frameworks, not instructions."
  builder: "Debugging and shipping. 'i'm about to make a big PR.' 'can we check in on where the image gen is?' Hands-on, not delegating."

best_for:
  - recipe: about_page
    role: primary
    why: |
      The about page is Thomas's natural register: transparent, personal,
      connecting craft to mission. The silicon-to-silicon arc was his
      editorial direction.
    target_feeling: warmth
  - recipe: careers_page
    role: primary
    why: |
      His radical transparency and operational care IS the pitch to
      candidates. "If you are at a bigger company wondering whether you
      will ever get to build something from a position of real ownership,
      this is that role."
    target_feeling: belonging
  - recipe: investor_update
    role: primary
    why: |
      His self-honesty and systems-thinking are exactly what investors
      want. He shares real numbers, real challenges, real progress.
    target_feeling: confidence
  - recipe: technical_blog
    role: supporting
    why: |
      His builder energy and debugging instincts add authenticity to
      technical content. "The Notion query silently errored" is how
      an engineer actually describes a production incident.
    target_feeling: respect
  - recipe: linkedin_post
    role: primary
    why: |
      His thinking-out-loud register is perfect for LinkedIn: genuine,
      specific, personal. "At the encouragement of others I'm sharing
      my hackathon project today" — this is anti-performance.
    target_feeling: warmth

verbal_fingerprints:
  - "cc @name @name"
  - "sorry, last week was a blur"
  - "i forget — is that on..."
  - "quick detour ask"
  - "can we check in on..."
  - "i want to sincerely apologize"
  - "at the encouragement of others"
  - "IMO"
  - "that's not the level of insight that's going to accelerate us"
  - "people should walk away feeling good"
  - lowercase when thinking, structured when communicating

voice_sample: |
  sure but that's not the level of insight that's going to accelerate
  us, IMO. I think you've already shared a lot of that feedback before.

  real world outcomes will come when we have more outcomes to share.
  at this point further clarity of expression is a function of visual
  storytelling

voice_sample_leadership: |
  I want to sincerely apologize for the payroll delay this week. While
  processing payroll on Monday, we encountered insufficient funds in
  our account, which required immediate attention and resolution. The
  situation has been fully resolved and all payments have been processed.

voice_sample_systems: |
  Meridian — operational intelligence for Verdigris.

  The problem: We use Claude Code, Cowork, and ChatGPT every day. They
  make individuals more productive. But they also accelerate siloed
  context. Each person has their own sessions, their own understanding
  of what's happening. The tools don't help you see across silos. The
  company has no shared memory.

voice_sample_builder: |
  The Monday pre-read did not land. Anthropic credits ran out last
  Thursday, which took down staff-prep and every Claude-using agent.
  Sunday's ids-prep ran but the Notion query silently errored, so the
  "all clear" post was wrong. Fixed the silent failure today. Migrating
  agents to OpenAI API key today.
