Posts by Ali Murtaza (17)
Claude Code Commands, Part 1: Audit Your Prompts Before You Blame the Model
Before you say the model got worse, audit the prompts it's reading. Part 1: the slash-first rule, which version you need, and /claude-api prompt-audit with a re-runnable baseline.
Claude Code Commands, Part 2: /effort, /simplify and /frontend-design
Three commands for the build-and-polish part of the job: /effort to set how hard Claude thinks, /simplify to tidy what changed, and /frontend-design to build a page you can check.
Claude Code Commands, Part 3: ultracode, /batch, /compact and /rewind
When one chat can't hold the job: ultracode and /batch to scale out, /compact and /clear to keep context sane, and /rewind for undo, plus everything it won't undo.
Claude Code Commands, Part 4: build-eval and hillclimb for Your Claude App
"It feels better" isn't a metric. Part 4: let /claude-api build-eval build the eval, check whether it's lying, then hillclimb one change at a time on held-out cases.
Claude Code Mods: Rewrite the Agent From the Inside
Claude Code now takes mods: a few lines of JS or TypeScript that run inside the agent, redraw its UI and step into tool calls. Your first mod, guard rails, and the security fine print.
Harness vs Loop vs Graph Engineering: Which Layer Is Broken?
Harness, loop or graph? Three layers, three different fixes. A symptom-first guide to which one your agent actually needs, anchored to LangChain, Anthropic, OpenAI and Microsoft docs.
Claude Code Agent Teams, Gemini 4 Argon and the Loops-and-Graphs Playbook
Claude got a team lead, Gemini 4 Argon can write a million tokens in one go, and a ~8M-view X article explains how to stop babysitting agents. The week's hot items, sourced.
Jev by TypeSafe: The System One Model That Answers in Types
Bro, just say true. You asked for true or false; the LLM wrote five paragraphs. Jev answers in types, with a confidence number. Here it is on spam and on vibe-coded PRs, receipts included.
Claude Code Effort Levels: Low, Medium, High or Max?
Low took 1.5 minutes. Max took 67. What Claude Code's effort dial actually changes, where it pays off on Terminal-Bench 3.0, and the one failure that gets worse at max.
Claude Opus 5.5: Fable-Level Work at 40% Lower Cost Than Opus 5
Anthropic's first Claude 5.5 model performs at the level of Fable 5.1 on most work and costs 40% less to run than Opus 5. The pricing, the benchmarks and the caveats, in one place.
CLM-8B: How Contrastive Language Models Speed Up AI Agent Decisions
An open model that doesn't write a single word: CLM-8B scores an agent's candidate actions and returns probabilities, fast. How it works, what the numbers say, and where it fits.
What Is Hugging Face Used For? A Guide for AI Engineering Teams
Models, datasets and apps in Git repos anyone can build on — plus one API for hundreds of hosted models. How the Hub shortens the loop from research release to working product.
What Is Harness Engineering? Agent = Model + Harness (2026)
The model is the brain; the harness is everything that makes it reliable. What goes into a harness, why feedforward and feedback both matter, and a starter setup for your team.
What Is Graph Engineering? Designing Multi-Agent Systems as Graphs (2026)
When one agent loop is not enough, the design question becomes who talks to whom, who owns what and who can say no. How to design that topology, govern it and build a first graph.
Coolify vs Openship: Self-Hosted PaaS on a VPS or Dedicated Server (2026)
Two open-source ways to run your own deployment platform. One is established with a big template catalog; the other is young, zero-config and can build off the server. Here is how they differ in practice.
How to Set Up Coolify on a VPS or Dedicated Server (2026 Guide)
From a fresh Linux server to your first HTTPS deployment: the requirements, ports, one-line installer, the admin-account step you must not skip, and DNS that actually resolves.
How to Set Up Openship on a VPS or Dedicated Server (2026 Guide)
Install the CLI, run the guided setup, and ship your first project — plus headless installs for automation, service management commands, and the hardening steps a young platform needs.