Building Clickfolio — A Full-Stack App on Cloudflare Workers
How I built an AI-powered portfolio generator using Workers, D1, R2, Queues, Durable Objects, and WebSockets.
Thoughts on software development, technology, and building products.
The best posts if you want to understand how I think about systems, product tradeoffs, and shipping real software.
How I built an AI-powered portfolio generator using Workers, D1, R2, Queues, Durable Objects, and WebSockets.
Real-time log streaming, OTLP-compatible ingestion, and PostgreSQL full-text search — without running a giant observability stack.
Lessons from building and scaling Alita Robot, a community automation bot used by large public Telegram groups.
A practical comparison of current AI models — what's good, what's not, and what I'd actually use for real work. Last updated April 2026.
Upload a PDF resume, get a live web portfolio. I built clickfolio.me entirely on Cloudflare's edge stack, D1 for data, R2 for files, Queues for async processing, Durable Objects for real-time WebSockets, and Gemini for AI parsing. Here's every technical decision, including the ones I regret.
I built an open-source Telegram bot that grew to 300k+ real users, 200+ GitHub stars, and 300+ forks. Here's the brutally honest story of how—including the tactics I'm not proud of.
Every logging tool felt like overkill—ELK's operational complexity, Datadog's pricing, Loki's LogQL learning curve. So I built Logwell: one Docker Compose, PostgreSQL-native, real-time streaming. Here's what I learned.
Ten flagship AI models. Five months of daily use. One spreadsheet of benchmarks. Here's the honest breakdown — and the winner wasn't what I expected.
180K+ GitHub stars. Security researchers panicking. $10-25/day in API costs. I spent 2 weeks with OpenClaw trying to find the killer use case. The product isn't ready — but the paradigm it's building might define the next decade of AI.
Everyone's talking about vibe coding like it's the future. After building a prototype in hours and then drowning in an unmaintainable mess, here's what the hype gets wrong.
The authenticity paradox explained: why your online persona diverges from your real self. Psychology research + my experience with 500k users and daily metrics obsession.
I've spent $50+/month on AI coding tools for a year. Here's what actually works, what's overhyped, and when to use each tool (including free local LLMs).
How I shipped 30+ projects while getting my MS in Computer Science at ASU. The stack choices that let me build fast, deploy faster, and actually sleep at night. Real lessons from PickMyClass, Alita Robot, and projects serving 250K+ users.
Zero hosting costs, global edge deployment, and sub-50ms response times. Here's how I run divkix.me on Cloudflare Workers with Next.js 15 and why you should consider ditching Vercel.
How I run powerful AI models locally on my M4 Pro MacBook with LM Studio. Privacy, speed, and vision models that can see images - all offline.
How I built a class notification system that watches for open seats and instructor updates, so students never miss their dream classes due to constant refreshing.