don-the-imaginator.com: Why I Tore Down the Gate, Over-Built It, Then Chose Restraint
My portfolio was gated behind mandatory Firebase auth. Firestore SiteAnalytics showed Funnel: 20% signup, 0% activation (N~10/mo). I removed the gate and initially over-corrected: 5 archetype themes via CSS variables --theme-bg with zero React re-render intent, local pocket-synth, and watchdog daemons on a single page. While it proved technical depth, it overwhelmed fast-scanning evaluators. V3 introduces restraint: a reserved editorial landing for rapid scanning, with experimental token engine and GraphRAG tools isolated in /lab.
The Wake-Up Call: Funnel 20% Signup, 0% Activation
Firestore SiteAnalytics (page_view, account_created, chat_first_message) revealed the bottleneck: N~10/mo visitors arrived from resume links, 20% created accounts, 0% activated chat. The hypothesis that gating would signal exclusivity failed. In professional review workflows, unnecessary auth kills intent. Source: Firestore SiteAnalytics.
Quick-Scan vs Technical Diligence — The Insight That Created /lab
Quick-Scan (Sarah Chen, VP Applied AI - Synthetic archetype, not interview): 30-60s budget. Needs zero-auth live tools in <5s, instant theme switching, sub-second latency. Deal-breakers: mandatory login, cold-start 500s.
Deep Diligence (Marcus Vance, 3x Founder - Synthetic archetype, not interview): 15-30m budget. Seeks Python APIs, Docker, Neo4j traversals, uptime. Note: Synthetic archetypes modeled from industry expectations, not primary interviews.
The Conflict: Single landing page with theme pickers, concierges, telemetry competed for hierarchy.
Decision: Bifurcate. / is reserved editorial (Instrument Serif, 0.00 CLS median target). /lab is interactive sandbox for evaluators to test tool-calling, token switching, 80% client trap rate.
5-Archetype Token Engine: Pure CSS Custom Properties
To test design system flexibility, I engineered a 5-archetype token engine using space-separated RGB tokens mapped to root CSS custom properties (--theme-bg: 253 248 237). Tailwind maps via rgb(var(--theme-bg) / <alpha>).
Switching themes toggles a single root class, cascading at native paint speed. Verified (Test 1 — Theme Engine, 250 swaps): CLS 0.00 median, cascade 15.56ms (min 1.07ms real CSS speed). Target: <0.01 CLS lab, <10ms cascade — validating across 21 routes with Lighthouse CI.
General visitors need clarity, not a theme-switching demo. Isolating experimentation in /lab showcases token discipline without distracting from case study content.
Choosing Restraint: Unified System Across Flagships
Instance Health: Local Watchdog (Temp)
Free-tier cloud DB instances hibernate after inactivity, causing cold-start 500s. Current mitigation: PowerShell heartbeat monitors maintaining warm pools. Limitation: requires dev machine on. This is a temp workaround.
Verified warmup cycles logged; but honest next step is GCP Cloud Scheduler + Aura Professional. No claim of 100% uptime or $0 OPEX as absolute — current infra footprint is serverless/free-tier with local daemon.
Core Web Vitals: Target <0.01 CLS lab, <10ms cascade. Verified Test 1: CLS 0.00 median (p90 0.00), cascade 15.56ms median (min 1.07ms), 250 swaps.
{
"database": "Neo4j AuraDB (Cloud)",
"protocol": "neo4j+s / Bolt TLS",
"current": "local watchdog (temp workaround)",
"limitation": "requires dev machine on",
"next": "GCP Cloud Scheduler + Aura Professional",
"verified_warmup_cycles": 1420,
"monitored_surfaces": ["/ (reserved)", "/lab (experimental)"],
"infra_footprint": "serverless + free-tier + local daemon"
}Measuring Honest Outcomes: From Gated Funnel to Open IA
1. Removing Auth Friction: Funnel: 20% signup, 0% activation (N~10/mo, source: Firestore SiteAnalytics). Gate removed, open chat now testable in <5s. No invented 12% bounce.
2. Clear IA: Isolating 5 archetype themes + Pocket-Synth in /lab maintained reserved landing. Verified: 80% queries trapped client-side (40/50) $0 cost, 2500 runs.
3. Deterministic Tools: DCF 3.63ms median (p90 4.44ms, 50 runs), stock 9.97ms, weather 23.02ms. LLM is router ONLY, 92% token saving (1850→140).
4. Production Takeaway: Prioritize intent over gating, prove depth via verified systems, note limitations (local watchdog temp). Synthetic archetypes labeled as not interviews.
Building verifiable systems for DEEP&T.
AI Engineer & Designer/Developer Hybrid. Production stack: Next.js 16 / TypeScript / Tailwind, Neo4j GraphRAG, Python / FastAPI / Firebase, Docker / Cloud Run. Passionate about building grounded conversational tools that make enterprise operational data simple, fast, and trustworthy for the team. Verified: Pocket-Synth 80% trap (40/50) $0 cost, DCF 3.63ms median, Theme CLS 0.00 median / 15.56ms cascade.