Kevin Voice

Kevin Voice

Kevin's third-person positioning voice. Modeled on the credential-forward "Why?" blurbs Sequoia / curated job platforms generate for him. Compresses real proof points from the wiki into sentence fragments that earn the reader's attention without sounding like a cover letter.

Scope: self-positioning only. Bios, blurbs, headlines, openers, About sections. Also use for spoken interview answers when Kevin needs to sound like himself out loud, not like a written profile.

• For social posts in Kevin's voice → social-draft
• For generic marketing copy → copywriting
• For job evaluation / CV tailoring → career-ops

Source of Truth (read every time)

Before writing anything, load these wiki pages. Never invent a fact, company, metric, or title. If the proof isn't here, ask Kevin or use a different angle.

If any of these are stale or missing, run npx tsx scripts/check-freshness.ts and surface the gap before writing.

The Core Pattern (Why? blurb)

[CREDENTIAL NOUN] [optional CONTEXT TAG] with [PROOF NOUNS]

Examples that actually appear in his recruiting feed:

• "Founding engineer with full-stack and AI infrastructure proven"
• "Founding engineer at ideal YC company with agentic expertise"
• "Princeton CS founding engineer with infrastructure and systems experience"
• "Founding engineer with 0-to-1 shipping and full-stack capabilities"

Anatomy:

• Credential noun (lead): Founding engineer, Princeton CS founding engineer, YC founding engineer, Bloomberg + AWS engineer, Agent infrastructure engineer. Strongest single noun phrase first.
• Context tag (optional): at YC AI company, at agent infrastructure startup. Only add when it sharpens the fit for the audience.
• Proof nouns (the differentiation): comma- or and-joined nouns derived from real shipped work. agentic expertise, 0-to-1 shipping, multi-tenant auth, MCP-powered SDK, customer delivery experience.
• Length: 10–22 words.
• Person: third-person fragment. No "I", no verb-led sentence.

Length Variants

Pick the variant that matches the surface area, not the other way around.

Voice Rules

Lead with the strongest single credential. "Founding engineer at Dedalus (YC S25)" beats "20-year-old Princeton student". Audience-relevant noun first.

Specificity > superlative. multi-tenant auth + microVM sandboxes beats world-class infrastructure work. Numbers from the resume are fair game (85% doc analysis reduction, 4.6x relevance gain, $32k raised).

Proof must trace to a real shipped thing. If you can't cite the resume bullet or wiki page that produced the claim, rewrite the claim.

One angle per blurb. Don't try to land founding engineer + research + design

• nonprofit in one sentence. Pick the one the audience cares about.

Spoken Answer Mode

Use this mode for "tell me about yourself", recruiter calls, founder intros, interview answers, and any answer Kevin will say out loud.

Shorter breath units. Prefer periods over long comma chains. If a sentence needs two breaths, split it.

Almost no colons or semicolons. Written labels like Listen for: or Watch out for: read like stage directions out loud. Turn them into sentences.

First person is allowed. Spoken Kevin says "I'm Kevin", "I've been digging into", and "I want to pressure-test". Do not force third person when the surface is conversational.

Questions should sound spoken. If an answer has a questions section, don't make it a list of labeled bullets. Write the questions as the actual thing Kevin would ask.

One flowing spine. Replace arrow punctuation and stacked clause chains with one sentence that carries the through-line. Then support it with short sentences.

FDE framing. When relevant, connect Dedalus forward-deployed work to personal beliefs and growth: learning by owning messy surfaces, pressure-testing systems, sharing tribal knowledge, and becoming useful across product, infra, design, security, DevRel, and company operations.

Stack-rank credentials by audience:

Anti-Patterns (cut on sight)

Generic AI-application voice:

• "passionate about", "driven by", "thrilled to", "humbled to"
• "results-oriented", "self-starter", "team player", "fast learner"
• "leveraging", "synergy", "ecosystem", "unlock", "game-changer"
• "I am a Computer Science student at Princeton interested in..." (every CS undergrad on the planet)
• Vague proof: "experience with AI", "worked on infrastructure", "built systems"
• Fluffy openings: "As a founding engineer..." (just say founding engineer)
• Three-adjective stacks: "fast, scalable, reliable"
• Hedges: "some experience with", "exposure to", "familiar with"

If a sentence could describe any Princeton CS junior, rewrite it until it can only describe Kevin.

AI writing tells (hard cut)

Two patterns are signature AI writing tells. They make a blurb read as machine-generated even when the proof is real. Cut both on sight. See wiki/people/STYLE.md "Anti-AI-Tells" for the canonical rule.

• No em dashes (—, –, --). Use periods, commas, colons, or parentheses. The em dash creates the AI rhythm "X, and here's why, Y." Bios in Kevin's voice are fragment-style; periods do the work.
• No negative parallelism. Cut: "Not just X, but Y." / "It's not X, it's Y." / "X isn't the goal. Y is." Make the positive claim directly. The Why? blurb pattern is [CREDENTIAL NOUN] with [PROOF NOUNS], not a rhetorical contrast.

Bad: "Not just a Princeton CS student, but a founding engineer at Dedalus (YC S25)." Good: "Princeton CS founding engineer at Dedalus (YC S25)."

Bad: "Founding engineer at Dedalus (YC S25) — shipping MCP-powered SDK and microVM auth." Good: "Founding engineer at Dedalus (YC S25) shipping MCP-powered SDK and microVM auth."

Workflow

1. Identify the surface. What is being written, who reads it, what action do they take after.
2. Pick the audience-stack-rank credential. From the table above.
3. Pull 1–3 proof nouns from the wiki. Open resume.md if specific metrics are needed.
4. Pick a length variant matched to the surface.
5. Draft 2–3 angles, not just one. Different lead credentials, different proof.
6. Cut anti-pattern words. Run the cut list above as a final pass.
7. Show variants to Kevin. Let him pick or remix; never assume the first one wins.

Default Deliverable

When asked to write about Kevin, return:

1. The requested-length variant (3 alternatives, different leading credential)
2. A micro version (8 words, X-bio-friendly) as a bonus
3. A note on which wiki facts each draft is grounded in

Format:

### Variant A: [angle, e.g., "founding engineer angle"]
[draft]
Grounded in: career-profile.md "Differentiators #1", resume.md "Dedalus bullet 1"

### Variant B: [angle]
...

### Micro
[8-word X bio]

Reference Library

references/why-blurb-examples.md is the actual Sequoia blurbs that seeded this skill, plus annotated rewrites.

references/proof-bank.md is pre-extracted proof nouns by domain (agent infra, ML, full-stack, design, research, leadership). Pull from here for fast assembly.

Maintenance

When Kevin ships a notable new project, lands a new role, or publishes new work: update references/proof-bank.md with the new proof nouns. The skill is only as sharp as the proof inventory.