Sean Traynor
Portfolio 2026

AI-native operator. 3 companies. 18 months of agentic-SWE practice. 1,006 tickets merged in 18 days on Claude Max.

Website
Email
LinkedIn
9 git repos 51 agents deployed 1,006 tickets merged 3 active companies Building agent-first since Sept 2024
Sean Traynor — Portfolio 2026 Slide 1 of 12
02 Bio

Background.

Education
Tufts University — B.S. Computer Science, May 2024. Dean's List. Two-year captain of the Men's Varsity Soccer team. 2019 NCAA National Champion. 3x Academic All-Conference · 2023 Academic All-District.
Career
MakenaAI (Sep 2024 – Jan 2026) — sole AI engineer; AI, mobile, and property tech. Built internal RAG chatbot from requirements to production. Architected subscription system that onboarded 2,000+ clients in one week. Led "Digital Clone" AR pilot for remote property inspection.

Columbia University (Dec 2022 – present) — Full-stack iOS engineer, Language & Memory Lab. NIH grant contributor. Published in The European Journal of Neuroscience. Advised research teams across Africa, Europe, and Asia (4,000+ patients).

Scale AI (Feb–Jul 2024) — Lead AI Consultant. Evaluated 500+ LLM code responses across 8 generative AI projects.
Sept 2024 – Present — 18 months of practice → 18-day burst
Started building with Cursor in September 2024 — GPT-4o and Claude 3.5 Sonnet era, before agent frameworks existed. At MakenaAI, shipped sole-AI-engineer features while inventing internal patterns for agent orchestration, testing harnesses, and git protocols. Left MakenaAI in January 2026; sketched the Master OS architecture during part-time consulting work, spanning DevOps, frontend, backend, security, and end-to-end deployment — all through an agentic-SWE lens.

18 days ago, turned on the Claude Max plan and built the entire Master OS — 9 repos, 7 brands, 51 agents, 1,006 tickets merged. The 18 months of practice was the prerequisite; Claude Max was the unlock. One operator. $200/month flat. Zero additional headcount.
Sean Traynor — Portfolio 2026 Slide 2 of 12
03 Thesis

The operating thesis.

Human-forward AI
specialists directing agents — not replaced by them
principle
The highest-value humans in the next decade are those who can direct agentic teams with judgment, taste, and accountability. AI doesn't eliminate the operator — it removes the friction between intent and execution.
Deterministic scaffolds for nondeterministic AI
trust boundaries enforced at the code layer, not by model compliance
architecture
Every merge is gated by automated hooks. Agent actions are either explicitly permitted or explicitly blocked — no implicit trust. Verifier sign-off required before anything reaches main.
10–100x force multiplier
individuals overseeing agent teams at enterprise output scale
leverage
A specialist operator running a well-designed agentic system compresses work that once required 5–100 people. The ceiling is not "save one job" — it is "run a company at enterprise output with startup headcount." Proof: 1,006 tickets merged across 3 active companies, 1 operator, 1 Claude Max plan, $200–400/mo in API costs.
Spec-driven execution
structured execution pipeline · parallel agent dispatch
engineering
Every ticket must specify scope, acceptance criteria, and verification commands before an agent can pick it up. 1,458 tickets processed in 18 days. 69% merge rate. Agents run in parallel with automated conflict detection and structured review at each stage.
Sean Traynor — Portfolio 2026 Slide 3 of 12
04 Portfolio Overview

3 companies. 1 operator. 1× Claude Max plan.

Sandbox
ai-services · ai-os · sandbox university
Active
AI-native services agency + education platform. Signed client: Chill'n Kendall ($4K build + $300/mo retainer). University launching summer 2026.
Labs custom builds University $49/mo Studios $9/mo
Blockwise Intelligence
autonomous ai-native quantitative research
Active
Autonomous AI-native quant research lab. Studies markets, builds quantitative tooling, and tests strategies one block at a time — research, platform, and strategy testing stages with deterministic gates.
Daily research loop LLM-council critique
Violet Studios
ai production tools · bloom music platform · c++ plugins
Active
AI-driven music analysis platform (Bloom POC delivered in one weekend via agentic workflow). Real-time C++ audio plugins using JUCE. Developer tools for producers and artists.
Bloom POC shipped Productionization in progress
AI-native ops cost line
$200/month flat — one Claude Max plan covers all development across every company. That's the primary cost; all three businesses run on top of it. Zero engineering headcount beyond the operator. Equivalent staffing: 5–100 people per company × 3 companies = impossible without the OS. Built on 18 months of agentic-SWE practice — custom agent framework + testing harness + git protocols invented at MakenaAI — then executed in 18 days once Claude Max activated. The practice was the prerequisite; Claude Max was the unlock.
Sean Traynor — Portfolio 2026 Slide 4 of 12
05 Sandbox — Strategy

3-product architecture.

Sandbox Studios — $9/mo
Community + brand-voice content. Danny-led. Skool. UGC / AI marketing for entrepreneurs. Entry point to the funnel.
Sandbox University — $49/mo
Self-updating AI-native curriculum. 5 courses: Skills 101, Agents 101, Workflows 101, Personal OS Build, Business OS Build. Credentials stack.
Sandbox Labs — Custom
Done-for-you AI OS builds. Day rate: $1,200. Build: $4K base + $2K/additional location. Retainer: $300 Starter → $750 Standard → $1,500 Pro.
Funnel flow
Studios $9 → upcharge → University $49 → upcharge → Labs custom → offboard → University $49
Every Labs SKU bundles 1 year of University access at $1 (card-on-file, auto-converts to $49/mo at year 2). Labs → University offboarding is the steady-state retention vehicle — clients who graduate to self-serve remain subscribers indefinitely.
Labs pricing rubric
$1,200 day rate · -25% friends & family on build
$300/mo Starter retainer · 10% channel cut (Year 1)
+$2K build + $200/mo per additional location
Personal OS tier — Labs SKU #1
$5K flat build for a done-for-you Personal OS deployment.
Bundles 1 year of University at $1, auto-converts to $49/mo.
Built for high-leverage individual operators.
Sean Traynor — Portfolio 2026 Slide 5 of 12
06 Sandbox U — Proof + GTM

AI-native curriculum. Near-zero overhead.

Jun 15
Community live
Jun 30
Coursework v1
$49
Standard /mo
$39
Early-bird /mo
Curriculum v1 — 5 courses
Skills 101 Agents 101 Workflows 101 Personal OS Build (capstone) Business OS Build (capstone)
Capstone courses produce a deployable artifact — a working Personal or Business OS — not a certificate of completion. Micro-credential → Operator (5 courses) → Architect (10 courses).
Near-zero overhead model
Agentic research pipeline ingests top AI creators daily, applies Sandbox brand perspective, and updates coursework within the week. No permanent teaching staff. Curriculum stays current by construction.
vs. $200K / 4-year degree
Traditional CS programs teach frameworks from 2018. Sandbox U teaches the operating system that runs today — and updates weekly. $49/mo vs. $200K / 4 years.
Scaling mechanism
Each University graduate installs the OS framework on their own business. Sean doesn't become the bottleneck. Labs delivery scales through a trained contractor pool using the same substrate.
Launching summer 2026 Skills 101 Ch.1 public now Enterprise B2B via Labs
Sean Traynor — Portfolio 2026 Slide 6 of 12
07 Violet Studios — Strategy

Four-product AI music-tech business.

Bloom — AI Mixing & Mastering Analysis
core
Deterministic audio analysis engine. Gives producers actionable, explainable feedback on loudness, frequency balance, and section dynamics — not a black-box result.
Loudness normalization, frequency cross-section, dynamic range analysis
Reference-track A/B: "your mix vs. your target"
Empowers the producer — vs. Suno/LANDR which replace them
Production Services
revenue now
Warm-lead pipeline from Bloom users who want a producer in their corner. Services-led entry to the VS ecosystem.
Mix + master, session production, artist development
Converts software trial users into paying clients
Plug-in Suite
roadmap
VST/AU plug-ins built on Bloom's deterministic analysis core. Brings the feedback loop directly inside the DAW.
Real-time metering + reference-track deviation overlay
Studio-grade, explainable — not generative
Sample Packs + Blog GTM
distribution
Curated sample packs seeded through an SEO-optimised blog. Passive top-of-funnel for Bloom and Production Services.
Blog: AI music production technique, mixing guides, reference-track analysis
Sample packs priced at $0–$29; upsell to Bloom subscription
Personal OS / Sean Traynor Slide 7 of 12
08 Violet Studios — Technical Proof

Empowerment vs. black-box — the Bloom thesis.

Analysis layer 1
Loudness
LUFS-I, LUFS-S, true-peak, dynamic range vs. platform targets (Spotify −14 LUFS, Apple −16 LUFS, streaming normalisation map).
Analysis layer 2
Frequency
FFT spectrum cross-section across sub/bass/mid/hi-mid/air. Per-section deviation from reference track. Masking identification.
Analysis layer 3
Reference A/B
User uploads a commercial reference. Bloom diffs their mix across all three layers and surfaces ranked deltas — biggest problems first.
Why deterministic beats generative
Suno generates audio. LANDR replaces the engineer. Bloom audits the engineer's work and explains the delta in plain language. Producers keep creative control; the system surfaces what their ears can't catch in the mix session.

Every insight is reproducible: same audio file always produces the same reading. No hallucinated "vibes" feedback — just signal-level maths the producer can act on.
Bloom — current status
Core analysis engine built. UI shipped. Entering closed beta with production-services clients, then public launch with the Plug-in Suite.
Engine: built Beta: Q3 2026 Launch: Q4 2026
Personal OS / Sean Traynor Slide 8 of 12
09 Blockwise Intelligence

Your AI-native quant research lab.

An autonomous AI finance research system that studies markets, builds quantitative tooling, and tests trading strategies — one block at a time. Co-built with partner Finn Westerink.

Stage 1 — Research
→
Stage 2 — Platform
→
Stage 3 — Strategy Testing
Autonomous daily loop
Observe → Research → Build → Test → Critique (LLM council) → Report → Improve. Founder receives daily update — not daily chores.
Blockwise Research Dashboard
Internal OS coordinating agents, research artifacts, data pipelines, backtests, simulations, and decision records. Core modules: research inbox, strategy registry, backtest engine, paper-trading monitor, LLM council panel, audit log.
No guesswork — deterministic gates
Hypothesis → data validation → backtest → bias audit → walk-forward → paper trading → LLM council critique → human approval. No strategy is trusted until it survives independent validation layers.
Partnership — Finn (NYC)
Co-founder partner based in New York City brings quant finance domain expertise and institutional-market context. NYC-coded brand identity: city blocks, density, infrastructure, markets, street-level intelligence.
Product direction
01 Internal research lab — self-education and infrastructure
02 SaaS research platform — AI quant tooling for individuals
03 Education — AI-native quant finance courses and labs
04 Strategy marketplace — tested frameworks, not guaranteed returns
Posture — research-only
No live capital today. Substrate first: build the autonomous research stack, prove the gates, ship critique-loops with the LLM council. Live trading is a downstream decision — only after the validation chain is auditable end-to-end.
Personal OS / Sean Traynor Slide 9 of 12
10 AI-Native Operations

1 operator × 3 active companies × $200/mo = 10–100x leverage.

live — tickets.db
—
tickets merged
—
30d merge rate
—
merges / day
—
brands managed
—
sessions run
—
active now
1
operator
×
3
active companies
×
$200
/ mo Claude Max
×
1,006
tickets merged
The margin thesis
Deterministic scripts + nondeterministic AI = 10–100× leverage at fractional headcount cost. The OS encodes repeatable processes into ticket-driven pipelines; AI handles the judgment layer.
What this unlocks
Run a portfolio of companies in parallel — not sequentially — while compounding institutional knowledge into a persistent, queryable wiki. This is the Sandbox Labs product: deploy this architecture inside your business.
Behind the throughput
300+ automated tests with strict type-checking and linting — no bypass. 80-metric health registry with daily snapshots. Self-improvement and learning loops run nightly. LLM council critique runs on-demand.
Friction-removal patterns
Verifier agents auto-file bugs on test failure. Parallel dispatch eliminates sequential bottlenecks. Auto-deploy on merge keeps everything current. Operator stays on strategy; the OS handles the rest.
Personal OS / Sean Traynor Slide 10 of 12
11 Client Engagement — Max (Chill'n)

Inventory management system for a restaurant group.

The engagement
Building a custom inventory management and shift-accountability system for a friend's restaurant business — four locations, daily operations. The system runs on the same OS substrate: ticket-driven, agent-executed, live on Vercel. Max receives automated morning reports on each location's shift performance.
Phase 1 — shipped
Portal live at chilln-kendall-portal.vercel.app. Four-location dashboard, shift logs, daily accountability reports to Max. Build fee: $4K. Retainer: $300/mo.
Phase 2 — mid-June pitch
Travelling to Florida mid-June to present Phase 2 in person: full inventory system, ordering workflows, multi-location analytics. Referral path: Max's 37-location network (Juicy Patties, Club Pilates).
Why this matters
Proof-of-concept for Sandbox Labs: a non-technical operator gets an enterprise-grade OS deployment for $4K. Every referral from Max's network is a new Labs client at the same price point.
$4K build + $300/mo retainer chilln-kendall-portal.vercel.app 4 locations live Phase 2 pitch: mid-June 2026
Personal OS / Sean Traynor Slide 11 of 12
12 Why Now

18 months of practice. 18 days of Claude Max. One operator OS.

The 18-day burst
1,458 tickets processed across generation, verification, pruning, and merge workflows in the last 18 days. 7 brands managed. 51 agents deployed. 9 repos in sync. One operator. $200/month flat. Not a prototype — multiple real business lines running simultaneously with safety hooks, verifier gates, auto-bug-filing on test failures, and an auditable wiki that compounds daily. The 18 months of practice was the prerequisite. Claude Max was the unlock.
The inflection
Before mid-2025, building a custom AI operating system required a full engineering team. Claude Code changed the build economics: a single operator with the right orchestration framework can now run production-grade AI ops for $200–400/month. The competitive window is the next 12–18 months before the space fills in.
Get in touch
Email
sptraynor2001@gmail.com
Live demo
chilln-kendall-portal.vercel.app
Building agent-first since Sept 2024 1,006 tickets merged $200/mo proprietary OS infrastructure 7 brands · 1 operator · compounding
Engineered: Structured ticket execution Slot-pool dispatcher 9-repo direct-to-main Auto-deploy on Vercel LLM-council critique loops Automated quality gates
Personal OS / Sean Traynor Slide 12 of 12
1 / 12
Overview — press ESC to close