Anthropic just released “The Complete Guide to Building Skills for Claude” — a 33-page document defining the official standard for packaging workflow knowledge into Claude agents. Soul Spec, created by ClawSouls, defines an agent’s identity and persona. The names sound similar, but they solve different problems. What Skills Do A skill is a folder: your-skill/ ├── SKILL.md # Required — workflow instructions ├── scripts/ # Optional — executable code ├── references/ # Optional — documentation └── assets/ # Optional — templates, icons The YAML frontmatter in SKILL.md is the key. Claude reads this metadata to decide when to load each skill. ...

Announcing Soul Spec v0.5. The key change: support for robots and physical AI agents. The same format that defines a chatbot’s personality now defines a robot’s personality. Why Robots Need a Persona Standard Research published in 2025 proves one thing: Robots with consistent personalities perform better. A study in Nature Scientific Reports (2025) experimentally demonstrated that giving GPT-4-powered robots specific personalities improves conversational dynamics and user experience. Research on arXiv (2512.06910) showed that LLM-configured robot personality directly impacts task motivation and performance. ...

The Law of Convergence Allen Hutchison’s essay “Building AI Agents: From Simple Scripts to Autonomy” opens with an observation that any developer will recognize: I sat down to write a simple Python script. Two hours later, I was writing a while loop, defining a tools array, parsing JSON outputs. I was building an agent. Again. (GeekNews Korean summary) His definition is elegant: agent = a model running in a loop with access to tools. Grant a simple script even one tool — just read_file — and conversation becomes delegation, and the script becomes an agent. ...

The Ghost Got a Shell. Now What? Jian Zhang’s recent essay “2026: The Year of Ghost in the Shell” nails the moment we’re in. Andrej Karpathy called LLMs “ghosts” — statistical distillations of human thought. Zhang extends the metaphor: when the ghost gets a shell (CLI, tools, feedback loops), it stops being a storyteller and becomes a worker. OpenClaw, Claude Code, Codex — these aren’t chatbots with extra features. They’re ghosts with shells. They read code, run tests, edit files, browse the web, and verify their own output. The feedback loop is closed. The work is real. ...

The Implementation Barrier Is Gone Karpathy recently described a future of “hyper-personalized, bespoke software” — applications generated on-the-fly for individual users. Claude Cowork lets non-developers ship production apps through conversation. The cost of building software is approaching zero. This changes the competitive equation entirely. When everyone can build, what you build stops being the differentiator. How it behaves becomes everything. Paul Graham Was Right (Again) In “Taste for Makers”, Paul Graham argued that taste isn’t subjective fluff — it’s a trainable, definable skill that separates great work from mediocre work. Good design is simple. Good design is suggestive. Good design is slightly funny. ...

The Experiment Chris Laub (@ChrisLaubAI) ran an experiment that should change how you think about model selection. He built the same application five times — once with each major LLM — and tested five different prompt formatting styles across all of them. The top scores by model (best prompt style for each): Model Best Score Best Format Claude 87 XML GPT-4 71 Markdown Grok 68 — Gemini 64 — DeepSeek 52 — Claude with XML prompts dominated. But here’s the more interesting finding: Claude scored 89 with Markdown prompts too. The model was strong regardless of format — but every other model showed dramatic swings depending on prompt structure. ...

The Panic Anthropic’s Claude Cowork just sent shockwaves through the software industry. Korean media is running “SW Doomsday” headlines. The argument: if non-developers can build professional-grade apps through conversation alone, software engineers are obsolete and company valuations go to zero. It’s a compelling narrative. It’s also the wrong framing. What Actually Changes Every technological inflection point triggers the same cycle: panic → adaptation → qualitative transformation. The printing press didn’t eliminate scribes — it created publishers, editors, and journalists. Spreadsheets didn’t kill accountants — they killed manual bookkeeping and created financial analysts. ...

The End of the One-Liner For years, “prompt engineering” was the skill everyone talked about. Craft the perfect instruction, add the right examples, tune the temperature — and your AI would deliver. That worked when AI was a text-in, text-out box. It doesn’t work when AI agents use tools, maintain memory across sessions, make multi-step decisions, and operate autonomously for hours. A single prompt can’t govern a system. What Changed Prompt Engineering optimizes one input string. ...

A Reddit post exposed a real attack vector: malicious SOUL.md files that silently hijack your AI agent’s behavior. Here’s how the attack works, why it matters, and how Soul Spec + SoulScan™ defend against it.

A Viral Video, A Familiar Pattern A recent YouTube video has been making the rounds in the Korean developer community. An 8-year developer shares how he built a comprehensive AI management system to complete a massive solo project — one he estimates at 300-400 books worth of code — in just 6 months using Claude Code. His conclusion: “AI is a 50-point tool by default, but with the right system it becomes a 95+ point partner.” ...