> blog

Writing.

Technical writing when I have something worth saying. Not a newsletter. Not a content strategy.

articleTR · EN · NL2026-07-08

Which One Is Real, Which Is AI? Making Indistinguishable UGC Ads with Claude Code + Higgsfield

A client rejected the first videos as “too AI.” The fix wasn’t a better model — it was a repeatable recipe driven from inside Claude Code: still-first to protect credits, an anti-AI prompt for real skin and window light, the ~8s identity-drift trap, the floating-bottle physics glitch, an ffmpeg verify loop, and the Virality Predictor picking the winner (how-to 58 vs hook 53). Real commands, brand anonymized.

UGCHiggsfieldClaude CodeMarketing StudioVirality Predictorffmpeg
read article → →
article + skillTR · EN · NL2026-06-23

Loop Engineering — Building the System That Prompts Claude For You

For two years the loop has been the same: you prompt, Claude runs and stops, you review, you prompt again — and the bottleneck is you. Loop engineering flips it: give a goal, put a real mechanical gate behind it (a Stop hook running actual tests — un-foolable, zero tokens), and the system runs until it hits the target. Comes with a drop-in Claude Code skill.

Loop EngineeringClaude Code/goalStop HookClosed LoopSkill
read article → →
articleEN · TR · NL2026-05-03

I Made Two AI Agents Talk in One Folder — OpenClaw + Claude Code

I tried 4 different models for OpenClaw — one of them silently drained my Claude Max quota. Then I realized: connecting two AI agents doesn't need an API or a protocol. A single shared folder is enough. Full model comparison plus the filesystem-as-message-bus pattern.

OpenClawClaude CodeMulti-AgentOpenAIMarkdown
read article → →
articleEN · TR · NL2026-04-29

Multi-Agent Systems & Obsidian — Building a Developer Army with Claude Code

Picture the same work split across four specialists running in parallel — all of them inside your terminal. Planner + executor architecture, skill files, and why Obsidian fits the equation. Plus the most mature community repos to clone instead of starting from scratch.

Multi-AgentClaude CodeObsidianSkillsSubagents
read article → →
articleTR · EN · NL2026-04-25

Using Claude Code Like a Pro — CLAUDE.md, Sub-Agents, MCP

Everyone talks about plugin lists. But what about the actual infrastructure underneath? CLAUDE.md files, sub-agent collections, and MCP servers — set them up right and Claude Code becomes unrecognizable. GitHub repos, install commands, and the best community resources for each layer.

Claude CodeCLAUDE.mdSub-AgentsMCPAI Workflows
read article → →
open source★ 25 starsgithub.com/selmakcby/knowledge-pipeline

Knowledge Pipeline — turning raw conversations into a living wiki

A reusable LLM pattern + skill for building persistent, schema-disciplined knowledge vaults. Three operations — INGEST (fold new sources into the wiki), QUERY (answer from the wiki, refile good answers), LINT (health checks for contradictions, orphans, broken links). Domain-agnostic: research, reading notes, product memory, team wiki.

Ships with a single-file terminal-aesthetic presentation (8 scenes, SPACE to advance) and a drop-in Claude Code skill. MIT licensed.

LLM PatternsClaude CodeSkillsKnowledge GraphsMarkdownOpen Source
view on github → →
coming soon

How I built vibeapp — streaming Claude into a live-deployed web app

A walkthrough of the architecture behind vibeapp: SSE streaming from Claude Haiku 4.5, Groq Whisper voice input, and spinning up an isolated Vercel project per generated app via the Deploy API.