Jonathan Hawkins/Builder of AI-native products
I turn frontier model
capabilities into products people can actually use.
20+ years shipping games. Two-time Game of the Year winner on God of War and Eclipse: Edge of Light. Now solo-founding Aligned Tools at the edge of agentic AI.
01—Featured work
Three case studies.
Projects that show the through-line: take a frontier capability, ship a product around it, fast.
01Case study
Claude API
Aligned Tools
Your company's brain — listens to every meeting, remembers what your team decides, ships finished work.
Integrates with:JiraLinearAsanaMondayNotionGitHubZoomTeamsSlackGmailCalendar
Most software is the same on day 1,000 as day 1. Aligned isn't. It reasons over patterns no human can hold in their head (recurring blockers, scattered expertise, decisions that keep getting re-litigated), so the longer a team uses it, the smarter their company becomes.
- Problem
- Engineering managers lose 30 to 60 minutes per meeting turning decisions into tickets, then spend the week reconciling the same task across Jira, Linear, Asana, Notion, and Monday. Action items slip, ownership is ambiguous, and the same decisions get re-litigated three sprints later because nobody remembers.
- Constraint
- Solo founder, B2B SaaS, in active enterprise sales. Must clear CIO-level security review and slot into the customer's existing stack. Switching tools is off the table.
- Move
Aligned listens across the surfaces work actually happens on (Zoom, Meet, Teams, Slack, Gmail, Calendar), then runs every transcript through an AI pipeline that extracts decisions, commitments, and ownership into a diff-style review UI where every human edit becomes training signal.
The shape of the system:
- Company Brain. A pgvector institutional-memory layer that synthesises across meetings so Aligned can surface “you decided this three sprints ago” instead of letting it get re-litigated.
- Agent frameworks in production. A CrewAI sprint planner for capacity-aware task prioritisation, running on a user-scoped Mem0 memory layer because CrewAI's default memory backend is process-local with no tenant partitioning, meaning shared FastAPI workers leak one customer's retrieval context into the next. CI greps every commit to make sure nobody silently flips it back. A streaming email-triage pipeline handles inbound work in real time. Computer-use cloud-desktop agents (OpenClaw + VNC) operate full Ubuntu desktops for tasks that need to drive a real computer: terminal, files, any GUI app, not just a browser.
- Real-time voice + avatar. A LiveKit cloud agent with a LemonSlice video avatar joins live calls. A 7-layer auth chain (service token + HMAC + single-use nonce + ±30s timestamp window + session-freshness TTL + heartbeat watchdog + unique-active-session index) means a leaked service token alone can't impersonate a user.
- Bidirectional ticket sync. Jira, Linear, Asana, Monday, Notion, and GitHub: per-provider mapping for project, status, priority, and issue-type, with a transactional honesty guardrail that only reports “synced” when zero errors fired. Round-trip sync is 5× the integration complexity of write-only, but the customer keeps their tool of choice and Aligned becomes the connective tissue underneath.
- SOC 2-aligned from day one. SSO-only, per-tenant isolation, AES-256-GCM versioned token encryption with key rotation, AWS Secrets Manager for per-tenant webhook secrets, hash-chain audit logs, fail-closed rate limiting, and dual TS+Python environment validators that block startup on missing or malformed config.
- Outcome
- MSA and two SOWs in negotiation with a name-brand AAA game studio after passing CIO-level security review on the first pass. >99% sync fidelity across 500+ tasks moving bidirectionally in live tenants. Median meeting-to-tickets dropped from ~45 min to under 3, with transcript provenance on every shipped ticket.
Claude APIMulti-provider syncBidirectional OAuthMCP serverNext.jsPostgresB2B SaaS
The Company Brain5 surfaces · 11 integrations
Aligned Tools
02Case study
WebSpatial
VibeView
Spatial multi-agent orchestration for Claude Code
- Problem
- Running a swarm of Claude Code agents in parallel is powerful but illegible. Output streams interleave in a flat terminal, work-in-progress is invisible until something breaks, and there is no way to compare what 5 agents are doing against the same plan.
- Constraint
- 28 hours, solo, at the SenseAI Hackademy. Cross-platform spatial UI from scratch.
- Move
- VibeView turns your editor into a room. Each Claude Code agent gets a floating glass window with live tmux output and parsed model/token/cost telemetry; a shared spatial kanban above them is fed by Claude Code's own ~/.claude/tasks/ files. Hold-to-talk voice routes commands to any agent, and when one transitions idle a Python bridge auto-summarizes its terminal with GPT-5-mini and speaks the summary back through a per-agent ElevenLabs voice. The hard call was WebSpatial. It ships to Apple Vision Pro, desktop browsers, and PICO 4 Ultra from one bundle, but the docs are sparse, so I had to reverse-engineer the renderer from runtime behavior.
- Outcome
- Shipped end-to-end in 28 hours, demoed live driving a 5-agent swarm against a real codebase. Won the WebSpatial track at SenseAI Hackademy and Best Technical Excellence at TechWeek 2026. Open-sourced post-event.
WebSpatialClaude Code SDKVoice agentMulti-agentvisionOS
VibeView
VibeView
Describe what the team should build…
Spawn Team2projects
6agents3active98done14pending$365.54cost
vibe-view:buildidle
> bash curl -s /dev/null -w "%{http_code}" http://localhost:3333 200> npx webspatial-builder run --base http://localhost:3333 ─── build end ─── xcrun simctl list devices "Apple Vision Pro" Apple Vision Pro (AC03C78A-4143-…) (Shutdown) +5 lines (ctrl+o to expand)✓ Built & launched on the AVP simulator.* Baked for 1m 28s · 1 bg task running (↓ to manage)> push all the latest to github* Stashed (auto-restores after submit)> bash curl -s /dev/null -w "%{http_code}" http://localhost:3333 200> npx webspatial-builder run --base http://localhost:3333 ─── build end ─── xcrun simctl list devices "Apple Vision Pro" Apple Vision Pro (AC03C78A-4143-…) (Shutdown) +5 lines (ctrl+o to expand)✓ Built & launched on the AVP simulator.* Baked for 1m 28s · 1 bg task running (↓ to manage)> push all the latest to github* Stashed (auto-restores after submit)
master · Opus 4.6+27 / -50 · 321k
vibe-view:pitchactive
Update src/components/PitchDeck.tsx 102 + 103 + <a href="#live-demo" 104 + className="flex items-center gap-1.5" 105 + <svg viewBox="0 0 16 16" fill="currentColor" 106 + className="w-3.5 h-3.5 opacity-60" 107 + <path d="M6.22 8.72a.75.75 0 0 0 1.06 1.06 108 + l5.22-5.22v1.69a.75.75 0 0 0 1.5 0" 109 + L6.22 8.72z" /> 110 + </svg> 111 + <span>WebSpatial Hackathon 2026</span> 112 + </a>* Cooked for 40sUpdate src/components/PitchDeck.tsx 102 + 103 + <a href="#live-demo" 104 + className="flex items-center gap-1.5" 105 + <svg viewBox="0 0 16 16" fill="currentColor" 106 + className="w-3.5 h-3.5 opacity-60" 107 + <path d="M6.22 8.72a.75.75 0 0 0 1.06 1.06 108 + l5.22-5.22v1.69a.75.75 0 0 0 1.5 0" 109 + L6.22 8.72z" /> 110 + </svg> 111 + <span>WebSpatial Hackathon 2026</span> 112 + </a>* Cooked for 40s
master · Opus 4.6+105 / -8 · 129k
vibe-view:bridgeactive
Requirements for this to work: - Your Mac on, running: pnpm dev:flat, pnpm bridge - Tailscale Funnels stay active (they are currently) - Frontend :3333 → Tailscale Funnel HTTPS 443 - Bridge :8765 → Tailscale Funnel :8443Note: to go back to local dev, reset env vars to localhost.* Sautéed for 1m 28s · 3 bg tasks still running* Background "Start dev server" completed (exit 0)* Old dev server replaced w/ Tailscale URLs. All good.> Bash(lsof -ti:8765 && echo "bridge OK") bridge OK✓ Bridge running. All three services up.Requirements for this to work: - Your Mac on, running: pnpm dev:flat, pnpm bridge - Tailscale Funnels stay active (they are currently) - Frontend :3333 → Tailscale Funnel HTTPS 443 - Bridge :8765 → Tailscale Funnel :8443Note: to go back to local dev, reset env vars to localhost.* Sautéed for 1m 28s · 3 bg tasks still running* Background "Start dev server" completed (exit 0)* Old dev server replaced w/ Tailscale URLs. All good.> Bash(lsof -ti:8765 && echo "bridge OK") bridge OK✓ Bridge running. All three services up.
master · Opus 4.6+8 / -3 · 254k
aligned-tools:planidle
Plan: implement the Q&A panel 1. useQAStream() hook → /meetings/:id/qa 2. <QAList /> inside <MeetingChat /> 3. Optimistic posts via Server Action 4. Unit tests: empty / loading / error 5. Migrate "questions" (idx meeting_id, ts)─ ready to apply (yes / no / a) ─> proceed with step 1Edit src/hooks/use-qa-stream.ts +47 lines, +0 imports export function useQAStream(meetingId: string) { const [items, setItems] = useState<QA[]>([]); // subscribe to SSE …Plan: implement the Q&A panel 1. useQAStream() hook → /meetings/:id/qa 2. <QAList /> inside <MeetingChat /> 3. Optimistic posts via Server Action 4. Unit tests: empty / loading / error 5. Migrate "questions" (idx meeting_id, ts)─ ready to apply (yes / no / a) ─> proceed with step 1Edit src/hooks/use-qa-stream.ts +47 lines, +0 imports export function useQAStream(meetingId: string) { const [items, setItems] = useState<QA[]>([]); // subscribe to SSE …
feat/qa-panel · Opus 4.6+12 / -4 · 273k
aligned-tools:testsactive
> npm test --filter @aligned/webPASS src/lib/transcript-parser.test.ts (1.2s)PASS src/lib/ticket-extractor.test.ts (0.9s)PASS src/lib/diff-review.test.ts (2.1s)PASS src/components/QAList.test.tsx (1.7s) ✓ renders empty state ✓ renders streaming questions ✓ optimistic post then settles ✓ retries on transient SSE dropTest Suites: 4 passed, 4 totalTests: 21 passed, 21 total✓ Suite green in 5.9s> npm test --filter @aligned/webPASS src/lib/transcript-parser.test.ts (1.2s)PASS src/lib/ticket-extractor.test.ts (0.9s)PASS src/lib/diff-review.test.ts (2.1s)PASS src/components/QAList.test.tsx (1.7s) ✓ renders empty state ✓ renders streaming questions ✓ optimistic post then settles ✓ retries on transient SSE dropTest Suites: 4 passed, 4 totalTests: 21 passed, 21 total✓ Suite green in 5.9s
feat/qa-panel · Opus 4.6+5 / -1 · 188k
aligned-tools:commitidle
> git diff --stat origin/main components/QAList.tsx | 84 ++++++++ components/MeetingChat.tsx | 12 +- lib/use-qa-stream.ts | 47 +++++ app/api/qa/route.ts | 31 ++++ 4 files changed, +168 / −6> git add -pStage this hunk [y,n,q,a,d,e,?]?> y components/QAList.tsx: 3 hunks staged✓ Working tree clean except for staged.> git diff --stat origin/main components/QAList.tsx | 84 ++++++++ components/MeetingChat.tsx | 12 +- lib/use-qa-stream.ts | 47 +++++ app/api/qa/route.ts | 31 ++++ 4 files changed, +168 / −6> git add -pStage this hunk [y,n,q,a,d,e,?]?> y components/QAList.tsx: 3 hunks staged✓ Working tree clean except for staged.
feat/qa-panel · Opus 4.6+0 / -0 · 72k
Spatial Swarmvibe-view·aligned-tools
VibeView
03Case study
Tauri
SkillVault Desktop
Making Claude Code's invisible config visible
- Problem
- Claude Code's power lives in env vars, hooks, slash commands, and skills, but they're scattered across files, undocumented in-product, and impossible to compare across projects. New users hit a wall; power users build the same scaffolding over and over.
- Constraint
- Consumer-facing desktop app, must feel native on macOS, ship a marketplace from day one.
- Move
- SkillVault Desktop is a single pane that reads your Claude Code config across every project, surfaces what is actually wired up, and lets you install vetted skills/hooks/commands from a marketplace with one click. The harder design call was the diff: when a user installs a skill, they see exactly which files change, with rollback. I built it on Tauri so it stays a 12 MB binary instead of a 200 MB Electron app, and the marketplace is just signed manifests in a public repo so anyone can publish without a backend.
- Outcome
- Public beta launched with 40+ community-contributed skills in the first week. Average user activates 6 skills in their first session.
TauriRustReactClaude CodeMarketplace
SkillVault Desktop
My Skills
All ▾Scan
Claude Code
Published by you 4
Statusline
v1.0.3code · 7d
Plan → Build
v1.0.0code · 5d
Team Bundle
v1.0.0auto · 4d
Global skills 10
bd-to-br
LOCAL8f · 2 sub
br
LOCAL3f · ref
bv
LOCAL3f · ref
team-debug
LOCAL5f · sub
review
LOCAL2f
init
LOCAL1f
Recent activity
✓Installedteam-debugjust now
✓Updatedbv → v1.212s
↻Scanning41 / 41 skillslive
MarketplaceTauri · 12 MB
SkillVault Desktop
02—Supporting work
More things I've shipped.
Side projects, weekend builds, and open-source experiments. Most ship a real product, even the small ones.
- 01
LabFork
A GitHub for agentic research. Fork labs, run agents on papers.
AgentsResearchNext.js - 02
voxherd
Voice-first multi-instance Claude Code controller for iOS.
iOSVoiceClaude Code - 03
TinkerSchool
Open-source AI homeschool platform for K–6 with M5StickC.
EducationHardwareOpen source - 04
Patina
A C++ → Rust conversion of the Godot engine.
RustGame engineGodot
03—About
A little context.
20+ years in games. Started as an intern at Sony Santa Monica and rose to Lead Level Designer over a decade, shipping God of War 1, 2, and 3 along the way. Was one of four hand-picked leads on a new unannounced AAA IP, managing 15+ designers.
In 2014 I founded White Elk Studios and recruited a small team of God of War franchise veterans to build Eclipse: Edge of Light, a VR adventure that won three Mobile VR Game of the Year awards (UploadVR, Daydream District, VR Fest 2018) and shipped to 8 platforms including PS VR, Oculus Quest, Steam, and Nintendo Switch.
In 2011 I founded GameDevDrinkUp, a monthly mixer that virally scaled from LA to 20+ cities worldwide, sponsored by Twitch.
Now solo-founding Aligned Tools. I take frontier model capabilities the week they ship and turn them into products non-engineers can use the same week.