Ben Nadel

Want to understand your app? Think in events, not state. In many CRUD apps we're used to thinking about behaviour in terms of state stored. Customers, invoices, addresses, quotes. This is great... until we need to reason about what happened in the past. It's difficult to reason why our apps are slow, have downtime or exhibit weird behaviour without seeing what happened in the past. When we collect an event log and query this, I start to understand strange behaviour and reason about my app on a different level. If you're keen to learn how to switch from state querying to event querying, I've got a free observability course. DM me OBSERVE for more details.

1

1 Comment

Remove Yourself as the Middleman The typical agentic coding workflow (if you have really embraced it) might look something like this: - New task appears in Jira or Linear - You review the task and hand it off to Claude Code (via MCP or CLI) - Claude reviews the task and creates a plan - You approve the plan - Claude works on it, pinging you along the way until it's done - You review the code (hopefully you asked for a changelog and docstrings) - You approve or request changes - You're still the bottleneck at every step. I wanted something closer to managing an actual development team. Given what's already possible with tools like OpenClaw (Clawdbot), none of this felt out of reach. Here's what I mapped out as the ideal state: - Create a task in Linear with clear scope and a defined outcome - An agent picks up the work and implements it - That agent opens a PR and submits it for review - A separate agent handles code review, addresses feedback, and approves the work - Another agent runs QA using Playwright or similar - The work is approved and queued for deploy So I built it. Version 0 is a local server that receives webhooks from Linear, reviews the issue, then orchestrates and assigns agents on my machine to do the work. When it's done, a GitHub webhook triggers a review cycle to catch and resolve any remaining issues. Everything is tracked and visible on a dashboard. It works. But I'm still too involved. I've been iterating since then and am starting to prove out v3 this week. We're heading toward a world where agents ship features to production with minimal human involvement. The key is what happens before any code gets written: clear planning and tight specs up front. Get out of the syntax. Go back to the basics of architecture.

4

1 Comment

Stop stuffing your Quarto repo with figure files like an overpacked suitcase... If you're storing generated figures in your Quarto project's main repository, you're creating a mess. Slow clones, bloated repos, unhappy teammates. Git submodules = lean main repo + organized assets + happy developers See it working in Quarto projects: 📂 Main repo: https://lnkd.in/gUtitFyW 🖼️ Figures repo: https://lnkd.in/gHViUFCH Implementation guide: https://lnkd.in/gzgf3tMg #quarto #git #submodules

62

3 Comments

The truly egoless builders are invisible by design. They're not at conferences, not on podcasts, not writing thought leadership. They're pushing code at 2am because the feature should exist. Their GitHub might have three stars. They don't care. Be authentic in everything you do. When you see something broken, don't hedge with "perhaps we should consider" or "it might be better if." Say "This is broken. I'm fixing it now." Then fix it. Clarity cuts through bullshit. The people who attack you for being direct are projecting their own fear of standing behind something real. When someone says "you should post this on HackerNews" or "why didn't you use React?" - stop. See the pattern. They're really saying "perform for us." They want you to play the game they're playing. Your refusal to engage threatens their whole framework. Note it, then go back to building. The best products you use every day were probably built by someone you've never heard of. Some database library handling millions of requests. Some protocol that just works. No Medium articles about "My Journey Building X." Just the thing, doing what it's supposed to do. Building for everyone means building for no one in particular. Not for VCs, not for Twitter, not for your ego. The product exists because it should exist. The idea that you saw it first is just luck or timing - someone had to. Might as well be you. Comments sections are where ego goes to feast. "Actually, you should've..." "Have you considered..." Stop when you see it. Everyone's performing their intelligence, their superiority. Identify the pattern - it's always the same performance. The builder who sees it, notes it, and just keeps building is the only one actually building. Group think disguised as wisdom is everywhere. "Best practices" that are just collective ego protection. "Industry standards" that mean "don't make us look bad." Real thinking means stopping when advice comes, recognizing the pattern of control, then choosing your own path anyway. Think different means think for yourself. Go build. Stop wasting your time on others who are lost and adrift.

4

2 Comments

Here's a simple framework to build systems that scale. Track: Notice what you're doing repeatedly Note: Add to your automation wishlist Just Do It: Start your day with twenty minutes of system building Today I built a UI testing loop in Claude Code that runs iterations automatically. No more sitting there, running the same tests manually, waiting for results. Five minutes saved per test. Twenty tests a day. That's 100 minutes back. But here's what most founders miss about systems: They wait for the "right time" to build them. When things calm down. When they have bandwidth. When the crisis passes. They're still waiting. I keep a simple list in my notes app. Every time I catch myself doing something twice, it goes on the list: "Manually copying customer data between systems" "Rewriting the same email response" "Running the same code tests" Then every morning, before the chaos starts, I pick one thing. Twenty minutes. That's it. This morning's automation will save me hours this week. Last week's Gemini CLI setup now transcribes thousands of calls while I sleep. Monday's Superwhisper setup means I speak emails instead of typing them. Small wins compound. Six months ago, I was manually formatting every email. Now Superwhisper handles it while I talk. A year ago, I was copying data between systems. Now APIs do it while I sleep. The compound effect is real. But it only works if you start. What's one thing you'll automate in the next 20 minutes?

4

2 Comments