Self-hosted copy (sessions + GitHub evidence). A sessions-only variant is also available. Not submitted to Overflow.
David Strawn
Ten years of solo-built voice, video, game, and health-data products — implemented by hand through 2024, now run by directing agents.
01Project role
The evidence most resembles a solo product builder and operator with two distinct eras. From 2016 through 2024, GitHub shows sustained hand-written implementation: a voicenotes iOS app carried across 16 months of Swift commits (transcription, background downloads, AI summaries, folders), a video-cleanup tool iterated over 18 months, a browser game over 14 months, and earlier Java, Go, and JavaScript work including a message broker and client-facing web and integration projects. From 2025 onward, the same person runs comparable projects by directing coding agents: agent co-authorship appears in recent repos, and session history shows delegation across build, deploy, production triage, and marketing, with verification done personally against the finished artifact — listening to edited audio for dropped words, comparing rendered frames against reference screenshots.
It resembles a platform or infrastructure engineer much less — operations evidence stops at a single small server — and it does not resemble an ML engineer or researcher; AI use is application-level (transcription, summaries, captioning, agents), not model work.
02Tools & working environment
claude-codeprimary · 12 sessions
Default agent across all seven recent local projects. Delegation spans building features, debugging from pasted console logs, deploying, resolving divergent git history, and triaging production incidents on a live server.
githubprimary · 34 repos
Continuous personal use since 2016; 30 source repos with the person as sole committer, four public. Multi-month arcs are normal: five repos each carry 70+ commits over a year or more.
browser & computer usefrequent · 4 sessions
Directed in-browser work to reconfigure a Reddit-marketing SaaS and capture its guides; two earlier desktop computer-use sessions delegated shopping-cart automation. Screenshot-based visual QA in a game-UI project.
custom agent skillsfrequent
Custom skills present and exercised in two different agent CLIs: a research-paper corpus toolchain (ingest, browse, full-page fetch) and a customer-prospecting workflow. Skill authorship is likely agent-assisted.
production vpsfrequent
Operates a small production server hosting a video product, with a monitoring agent running on it and a managed Postgres instance behind it; directed triage of a 504 export timeout, a database connection drop, and a ~2-hour render that needed to get faster.
ios toolchainfrequent · 2016-2024 era
Two voicenotes iOS apps built directly in Swift/TypeScript (2023–2024); no iOS activity observed since September 2024.
codex clioccasional · 1 session
One retained session, invoking the prospecting skill and planning customer outreach. Retention may hide earlier use.
dockeroccasional
Dockerfiles present in two hand-written repos; no orchestration or CI pipeline authoring observed anywhere.
03Languages, frameworks & platforms
javascript web appsvery familiar
The dominant language across a decade: a client-facing web page maintained for a year (2020–21), a 100+ commit productivity app (2022), an 18-month video tool (2024–25), a 14-month browser game, a public medical-taxonomy visualizer, and the current agent-built products. Hand-written through ~2024, agent-directed since.authorship: mixed — direct historically, directed-reviewed recently
video / media pipelinesvery familiar
Three generations of video tooling: a hand-built cleanup/captioning tool iterated 18 months, a production captioning product (WebAssembly FFmpeg in-browser, server-side render pipeline, color-accuracy tuning, A/B-tested pipeline changes), and a transcript-verified multi-clip editing workflow.authorship: mixed
agent-directed deliveryvery familiar
The consistent recent mode of work: plain-language task definition, mid-flight correction, and personal acceptance review that catches agent mistakes (duplicate clips, truncated sentences, render artifacts). Coordinates multiple agents; one session fanned out a 100+ subagent research workflow.authorship: direct
swift / iosfamiliar
95 hand-written commits over 16 months on a voicenotes app with transcription, background downloads, AI summaries, and folder organization; a second, week-long voicenotes build in 2023. Stale since 2024-09.authorship: direct
typescriptfamiliar
Cluster of direct 2023–24 work: an HTML editor, a voicenotes app, a Node backend, and agent-protocol experiments (one public).authorship: direct
pythonsome
An audio-cleanup tool (2023–24, direct), data scripts inside health-informatics repos, and an agent-written secrets-routing utility.authorship: mixed
java / gosome
Direct but dated: a message broker and a data-sorting program in Java (2016); one booking-integration service in Go (2020).authorship: direct
postgressome
A managed Postgres instance behind a production app; directed triage of a connection drop. No schema or query work visible.authorship: directed-reviewed
geospatial datasome
Directed a mapping tool layering zoning, appraisal, water, and pollution data for regional land evaluation.authorship: directed-reviewed
04Subject-matter experience
video & audio productssubstantial
Recurring theme since 2023: two voicenotes apps with transcription and AI summaries, an audio-cleanup tool, an 18-month video tool, a production captioning product, and a transcript-verified editing pipeline. Repeated across seven repos and multiple sessions.
biomedical literature & health informaticssubstantial
A connected arc since late 2024: a deployed consumer guide on household allergen remediation, a research corpus organized around explicit research questions, a PubMed/MeSH meta-science analysis of an allergen-associated disease space, and a public interactive MeSH taxonomy explorer.
games & interactive experimentssome
A 14-month browser game, a 4X prototype, a terminal MUD, game-UI experiments, and a factory-game planning tool. Hobby-scale but persistent.
early-stage marketing & customer discoverysome
Configured a Reddit-marketing SaaS, archived its playbooks, ran a prospecting workflow that ranks potential first customers, planned outreach.
client-facing web & integrationssome
Earlier era (2020–22): a guest-entry web page maintained for a year, a booking-platform integration in Go, and a productivity/secretary app. Commercial context cannot be established from repos alone.
land & real-estate datasome
Regional land evaluation combining zoning, appraisal, water, and environmental overlays. One project.
05Relative profile
Versus a product-oriented full-stack engineer: the decade of single-author repos establishes real direct implementation (JS, Swift, TS, Python, Java, Go), but always solo and at personal-product scale — no PR review, no team codebase, no test suites visible in commit history. Breadth of ownership (idea → build → operate → market) is the stronger axis.
Versus a consultant/operator: a close match for the 2025+ era: plain-language outcomes, pragmatic tradeoff decisions (disabling a payment flow, deferring scaling), and verification against the real user-facing artifact rather than the code.
Versus an automation/integration specialist: shares the tooling instincts — custom skills in two agent CLIs, browser automation, standing server-side agents — and has one real third-party integration in Go, but the automation predominantly serves their own products.
Versus an ML engineer/researcher: no overlap in evidence. Applies AI (transcription, summaries, captioning, agent direction) in products; no training, evals, embeddings, or model infrastructure work appears.
06Project match
Good fit
Agent-directed delivery of small web products from idea through deploy
Hands-on JavaScript implementation at solo-project scale, with agents or without
Acceptance review and QA of agent-produced work, especially media output
Biomedical-literature tooling and consumer health-education content
Early-stage marketing and customer-discovery experiments run through agents
Bring a specialist
Performance and scaling beyond a single small server
iOS work targeting current frameworks (direct Swift experience is real but stale since 2024)
Database and schema architecture
Team codebases requiring PR-based review discipline (no observed history of it)
Not established here
ML model training, evals, or research
Infrastructure at scale (Kubernetes, IaC, multi-service platforms)
Android development, security engineering
Collaborative engineering practice — every analyzed repo is single-author
Commercial release outcomes: deployment and operation are observable; customers and revenue are not
07Limits & methodology
Two evidence sources. Local agent sessions: 15 retained (Claude Code 12, 2026-06-05 → 07-15, one being this profile run; Codex 1; Cowork 2, April, low-fidelity titles only). GitHub: 34 repos (2016-01 → 2026-07) analyzed at the subject's request via read-only API — commit metadata, authorship, languages, and sampled commit messages; not a line-level code review.
Commit counts are not quality. Sustained multi-month arcs and sole authorship are established; code quality, testing habits, and architecture are not assessed.
Recent authorship is blurred. Post-2025 commits without agent co-author trailers may still be agent-written; pre-2025 repos predate coding agents and are treated as direct.
Session retention is rolling. Codex and Cowork volumes almost certainly understate use; one personal session was excluded from all examples for privacy.
Solo bias. All 30 source repos are effectively single-author, so collaboration, review, and team-process evidence is structurally absent — absence of evidence is not evidence of absence.
Cadence (metadata only): 11 sessions in the last 28 days, 7 active weeks of the last 12, most recent session 2026-07-15.