## Products
24 products. All built solo with Claude Code and a custom agent factory.
CoffeeBot↗
LIVEAI coffee recommendation quiz. Answer 5 questions, get your perfect brew.
82 coffee profiles generated
InkCloak↗
LAUNCHINGAI text detector + humanizer. Fine-tuned DeBERTa (AUROC 0.9948) + adversarial paraphrasing. GPU inference on RunPod.
AUROC 0.9948 detection accuracy
Scrutly↗
PLANNEDFree AI text detector. Simple, fast, no signup required.
Tool Sites
LIVE199 programmatic SEO pages across 5 niche domains. AI-generated content with human editing.
199 pages across 5 domains
OpenClaw Finance
BUILT5 specialized agents for financial analysis and reporting. DeepSeek R1, Docker.
5 agents, deployed on 2 machines
RackCloud
BUILTIaaS provider architecture: 400+ bare-metal servers, 17 services, billing, dashboard. Hetzner/DO alternative.
17 IaaS services, 400+ servers, 10 racks
DIYCalc↗
LIVEFree calculators for concrete, mulch, gravel, paint, tile — with bag counts, cost estimates, and buying guides.
30+ calculators
FitCalc↗
LIVETDEE, BMI, body fat, macro, calorie calculators — science-backed formulas with personalized recommendations.
40+ calculators
## AICPO Platform — 16 systems under one hood
AICPO Platform↗
LIVEChat with AI, extract facts, generate 43 strategy artifacts. Rails 8, multi-model LLM cascade, 3-panel workspace.
43 artifacts, 16 systems
Autonomous LLM Runtime
LIVEDynamic model discovery every 12h. Groq→OpenRouter→Bootstrap cascade. Auto-blacklist dead models. Frustration-triggered escalation. $5/day budget gate. 2K tokens instead of 25K per request.
7 failover levels, 0 downtime
AI Discovery Engine
LIVE3 research rounds. DuckDuckGo + TG channel validation (subscribers ≥50, freshness <90d). Article extraction. LLM filters only relevant insights. 90-day cache.
3 rounds, 90-day cache
Dual-Brain Chat
LIVEGroq (0.5s response) for standard queries. Frustration detector (5 triggers: negative tone, stagnation, repetition) → auto-escalation to Claude. De-escalation when conversation normalizes. 90% token savings.
5 triggers, 0.5s → Claude
Artifact Pipeline
LIVE43 types of strategy artifacts. Pipeline: Groq builds structure → Gemini generates → Groq validates. Auto-retry if score <0.5. A/B test models via ?model= param.
43 types, 3 stages
Semantic Memory
LIVEpg_vector embeddings + cosine similarity. Facts searched by meaning, not keywords. Auto-tracking coverage of 19 AJTBD data points.
19 data points, vector search
Product Health Monitor
LIVEPostHog, Stripe, GA4, Telegram, and more. Import external data, auto-extract facts, credit-gated.
6 connectors, real-time
Team OS
LIVEFull org management: invites, roles (owner/admin/editor/viewer), SAML + OIDC SSO, audit logs, per-member usage.
SAML + OIDC, 4 roles
Self-Evolving Prompts
LIVEWeekly auto-draft of improved prompts from negative feedback. Prompt versioning with performance snapshots.
auto-draft weekly
Evolution Pipeline
LIVEDaily feedback digest. Weekly Claude Sonnet analysis of patterns. Auto-generated evolution backlog with priority scores.
Feedback → AI → Backlog
Research Corpus
LIVECross-user research corpus. TG channels, articles, insights cached with TTL (posts 24h, articles 7d, insights 90d).
cross-user, TTL 90d
Pain Extractor
LIVEExtract facts from chat, email, Telegram, connectors. Map to 19 AJTBD data points. Auto-update project brief.
4 sources → 19 fields
Credit Engine
LIVERow-locked atomic deductions. 6 tiers (free→enterprise). Monthly limits with soft/hard enforcement. Transaction history.
6 tiers, row locks
Tokenomics Guard
LIVEPer-request token usage and cost tracking. $5/day budget limit. Alert at 80%. Auto-fallback to free models when exceeded.
$5/day, alert at 80%
Email Bridge
LIVEIMAP poll every 2min. Auto-routing replies to workspace projects. Attachment validation. Ephemeral message storage.
poll 2min, auto-routing
PDF Publisher
LIVEGenerate print-ready PDF from any artifact. Public token-based links. B4 format, watermark, no login required.
B4, watermark, public links