Victoria Lo - Tech Blog

How I rebuilt my blog into an end to end publishing system my AI agent can safely drive: drafts on their own branch, scheduled publishing with no cron, Cloudinary uploads, a gated newsletter, and an MCP server with structural guardrails.

Debugging cron job timeouts caused by session bloat. Learn how sessions accumulate state, why that breaks reliability, and how to design robust scheduled tasks with isolated sessions.

Three practical automation tools that compound together. Schedule multi-platform social posts via Buffer's GraphQL API, generate Techybara comics in a font cloned from my own handwriting, and add consistent blog callouts without touching an image model.

I ran my blog through Cloudflare's agent readiness checks and fixed what I could: markdown for agents, Content Signals, Link headers, an API catalog, skills and WebMCP. Also why I deliberately skipped OAuth and MCP.

From GPT-2's 1.5B parameters to Kimi K3's hybrid memory architecture — a deep dive into 7 years of LLM evolution, and why the biggest breakthroughs had surprisingly little to do with making models bigger.

Automating deliberate personal work with AI. Build adaptive language learning with spaced repetition, data-driven weekly reviews, and automated blog cover generation.

Building a compound knowledge system for AI agents. Learn how to implement structured memory with entity pages, provenance tracking, and semantic search that actually scales.

System design lessons from building a personal AI assistant. Learn patterns, best practices, and what to do differently when building agentic systems that run 24/7.

Complete guide to securing your OpenClaw AI assistant. Learn how to implement firewall rules, file permissions, tool allowlisting, and automated security monitoring for production deployments.

Deep dive into three advanced OpenClaw workflows - local RAG knowledge base, LinkedIn post generation with Buffer, and blog idea generator. Plus understanding OpenClaw's memory architecture.

Set up automated workflows and scheduled tasks with OpenClaw. Learn how to organize with Telegram topics, configure cron jobs for daily summaries, backups, and proactive assistant behaviors.

Connect OpenClaw to Google Workspace with the gog skill. Set up Gmail, Google Calendar, and Drive integration with proper OAuth scopes and security best practices for your AI assistant.

Learn how to build your own personal AI assistant with OpenClaw. Step-by-step guide to deploying on Oracle Cloud's free tier and connecting Telegram for a persistent, context-aware second brain.

Reading Hannah Arendt's The Human Condition on a work trip: five reasons not everything needs fixing immediately, and why slowness is a form of strength.

Karen Hao's Empire of AI read as a study of one mind: how Sam Altman engineered his original team, used myth-making as control, and where his blind spot is.

Only 10-15% of resumes reach a human reviewer in Singapore, and most roles are filled through referrals. Why online applications fail, from a hiring tech lead.

"AI bubble" has become the default framing for every generative AI story. What the real adoption and workforce data says, and where the gap actually is.

Why boundaries at work feel so hard to set, what crossing one quietly costs you, and what setting them professionally actually looks like.

What Anthropic's Claude Skills actually are, the types available, how to use them, and how they differ from MCP — context engineering, not magic.

Five things that actually grow a blog in 2026 — an email list from day one, your own domain, and refusing to split "technical" from "personal".