· 7 min read
Semantic Entity Resolution Without LLM: When Cosine > 0.92 Is Enough
How a three-level resolver cuts entity deduplication costs by 85% by reserving LLM calls for the 5% of cases that actually need them.
## Blog
Building in public. Lessons from shipping AI products solo.
· 10 min read
PostgreSQL as a Graph Database: CTE + pgvector + pg_trgm = No Neo4j Needed
How to build a production knowledge graph on PostgreSQL using recursive CTEs, pgvector, and RRF fusion — without touching Neo4j.
· 9 min read
Bi-Temporal Knowledge Graph: Why Facts Get Invalidated, Not Deleted
Why overwriting facts destroys institutional memory — and how bi-temporal storage preserves the full history of what you knew and when you knew it.
· 8 min read
Switch Formula From Data, Not From Assumptions
Why most teams build Switch Formula from guesses — and how to extract it from Knowledge Graph data instead.
· 9 min read
26 Data Points × Knowledge Graph = Autopilot for Product Research
How combining 26 JTBD data points with a Knowledge Graph turns manual research coding into real-time autopilot.
· 7 min read
Evidence-Traced Artifacts: Every Claim in Your PRD Has a Source
How a four-layer Knowledge Graph architecture makes every AI-generated statement verifiable — so you can trace any claim back to its source interview.
· 8 min read
The Best KG System Is Useless With One User: Tech vs PMF
I built 38 features into a Knowledge Graph system. One active user. Here's what went wrong and what I'm doing differently.
· 9 min read
Institutional Memory: When the Knowledge Graph Knows More Than Anyone on the Team
How a four-layer knowledge graph transforms scattered research into organizational memory that persists across teams and projects.
· 7 min read
Contradiction Detection: When AI Notices You're Contradicting Yourself
How a four-step pipeline using embeddings, semantic search, and bi-temporal storage automatically detects when new facts conflict with old ones.
· 8 min read
Predecessor Errors: 6 Bugs in One Session and How the Factory Learns
An AI agent skipped 147 lines of docs, triggered six cascading failures, and cost $50 to fix. Here is how Factory OS turns disasters into rules.
· 8 min read
71% Orphan Nodes: How I Built a Graph That Wasn't Actually a Graph
1083 nodes, 388 edges, and 769 nodes with zero connections. How I found the problem and got orphan rate down from 71% to 53%.
· 10 min read
38 Features in One Session: How AI Factory Builds Product
Knowledge Graph v2, Competitive Intelligence, Trend Monitoring — 38 capabilities, 22 commits, ~6500 lines of code, zero manual writing by the CEO.
· 7 min read
Don't Trust the Agent: Why Auditing Every Milestone Is Non-Negotiable
14 tests passed. 3 critical bugs survived. Here is the audit checklist that catches what tests miss.
· 8 min read
Smart Model Cascade: How to Not Die from Rate Limits on Free Models
50 concurrent users, free tier exhausted in 2 minutes. The architecture that keeps uptime at 100% for $2.40/month.
· 8 min read
Knowledge Graph for Product Research: Why RAG Is Not Enough
RAG is great for finding text. Product research requires understanding relationships. Here is why I built a methodology-native Knowledge Graph instead.
· 14 min read
AI Is More Human Than You Think
Forgets things. Skims long documents. Gets sloppy under load. Doubles down on mistakes. The cognitive failure catalog — and what to do about it.
· 12 min read
AI Reasoning Collapse: Why the Longer the Conversation, the Worse the Output
Every error correction adds tokens to the context. Those tokens bias the next response. The fix is architectural, not conversational.
· 11 min read
Why AI Agents Fail in Finance — and How I Fixed It
Five agents at 95% accuracy each. Joint accuracy: 67%. Every third report contains an error. The fix is verification loops, not better prompts.
· 13 min read
A Manager With Zero Code Will Outcompete Your Best Developer
The most powerful AI use case isn't for engineers — it's for experienced managers who can instantly spot when something is wrong.
· 6 min read
nevrai: LEGO for AI Products
Five independent API modules that handle the infrastructure every AI product has to rebuild from scratch — so you can build the actual product instead.
· 12 min read
How I Built a 31-Subsystem SaaS Platform Solo in One Month
Building AICPO — a product research platform with 31 interconnected subsystems — without writing a single line of code manually.
· 10 min read
Factory OS: How to Turn AI Agents Into a Reliable Engineering Team
A methodology for AI-native development — 15 specialized roles, 40+ explicit rules, and independent quality gates that make autonomous coding reliable.
· 10 min read
From Managing 150 People to Solo AI Builder
Career path: PUNKT E (revenue x5.7) to Mafin (scaling) to why I went solo and what Factory OS enables.
· 9 min read
Knowledge Graph: How AI Sees Connections and How to Verify Every Claim
Why 'the AI thinks so' is the worst argument in product research — and how a knowledge graph makes every recommendation traceable to specific user words.
· 11 min read
AI-Native Development: Not 'Use ChatGPT', But Architecture
The difference between AI-assisted and AI-native development — and why restructuring how work gets done matters more than which model you use.
· 8 min read
Dual-Brain: How to Cut AI Costs 90% Without Losing Quality
Most AI chats route every message through the same expensive model. A two-tier routing system cuts costs dramatically while keeping response quality the same.
· 8 min read
I Built 6 Products in 3 Weeks as a Solo Founder
How I shipped 6 AI products with zero employees, $15/month infrastructure, and an AI agent factory that does most of the work.
· 10 min read
I Built 9 AI Products Solo — Here's Every Single One
A breakdown of all 9 products: what each does, the tech stack, and what it costs to run and maintain.
· 7 min read
Self-Evolving Prompts: A System That Improves Without You
Most AI products run on static prompts that never change. Here's how to build a feedback loop that automatically improves prompt quality every week.
· 9 min read
Factory OS vs Cursor vs Copilot: What Actually Works in Production
Three levels of AI in software development — autocomplete, AI-IDE, and AI team — and when each one makes sense.
· 6 min read
The $15/Month Stack
Every tool and service I use to run 6 AI products. Total infrastructure cost: $15/month. Here is what works, what breaks, and when to upgrade.
· 10 min read
How to Actually Implement AI Agents in a Corporation
87% of Russian companies plan to adopt AI. 10% have. The gap isn't technology — it's approach. A practical framework for going from pilot to transformation.
· 8 min read
My AI CEO Runs While I Sleep
Inside Factory OS: a custom AI agent orchestration system with 15 roles that builds, tests, and deploys code autonomously.