Latest
All posts →Go-to-market with Claude Code
Launch day: the first 48 hours
A plan for the first 48 hours: email your waitlist first, follow Product Hunt's actual rules, post where your people already are, reply to everyone, and track the few numbers that matter.
Oct 27, 2026 · 6 min read
Oct 21, 2026Keep agents cheap and safeHow the context window drives cost, when to use /compact and /clear, how to pick a model and check spend, and how to handle prompt injection, permissions and secrets in Claude Code.Oct 14, 2026Cold email that gets replies (and doesn't get you banned)Research first, write short, keep personalization true, set up SPF, DKIM and DMARC, ramp slowly, respect the law, and let a Claude Code skill draft while you do the sending.Oct 7, 2026Claude Code for people who don't codeUse Claude Code on documents, CSVs, research folders and transcripts: set up a safe workspace folder, try six good first tasks, and learn where it falls short.
Topic 01
Claude Code, the easy way
Skills, hooks, subagents and MCP, explained without the jargon.
All 8 posts →Oct 7, 2026Claude Code for people who don't codeUse Claude Code on documents, CSVs, research folders and transcripts: set up a safe workspace folder, try six good first tasks, and learn where it falls short.Sep 16, 2026MCP in plain English: connect Claude Code to your toolsWhat MCP servers are, how to add one with a single command, how sign-in works, and five official servers worth connecting if you run content or a small business.Sep 2, 2026Permissions: stop clicking Allow all day (safely)How Claude Code decides when to ask, which permission mode to pick, and a starter allow, ask and deny list you can paste into a content or ops project today.Aug 19, 2026Plan mode: make Claude think before it typesHow to use Claude Code's plan mode: getting in and out, when it's worth it, how to review and edit a plan, and how to pair it with a short written brief.
Topic 02
Go-to-market with Claude Code
Strategies for finding customers with agents doing the legwork.
All 8 posts →Oct 27, 2026Launch day: the first 48 hoursA plan for the first 48 hours: email your waitlist first, follow Product Hunt's actual rules, post where your people already are, reply to everyone, and track the few numbers that matter.Oct 14, 2026Cold email that gets replies (and doesn't get you banned)Research first, write short, keep personalization true, set up SPF, DKIM and DMARC, ramp slowly, respect the law, and let a Claude Code skill draft while you do the sending.Sep 23, 2026Turn one video into a week of posts with a skillGet a clean transcript, then let a Claude Code skill with your style guide and real examples draft LinkedIn, X, newsletter and clip ideas that still sound like you.Sep 9, 2026Price it before you build itWhy a price is the strongest validation signal, which simple pricing model fits a first product, how to set up a founding price in Stripe, what to say on the page, and how Claude helps with objections.
Topic 03
Building agents that last
Systems, evals and the boring parts that make agents dependable.
All 5 posts →Oct 21, 2026Keep agents cheap and safeHow the context window drives cost, when to use /compact and /clear, how to pick a model and check spend, and how to handle prompt injection, permissions and secrets in Claude Code.Sep 30, 2026Run Claude Code on a scheduleUse headless mode to run Claude Code from launchd, cron or GitHub Actions, with a worked weekly-report job, tight permissions, a spending cap, and drafts you review before anything goes out.Aug 26, 2026Evals for the rest of us: test your agent like a checklistWhat an eval is in plain English, how to build a small test set from real examples, and how to run it by hand, then with Claude Code, so you know when a change made things worse.Aug 5, 2026Subagents: parallel research without the messWhat subagents are, where their files live, and a worked example: a company researcher that profiles ten leads in parallel using one fixed template.
About
I'm Dan, an agentic engineer. I write down what actually works with Claude Code, so you can skip the expensive experiments.
The world at the top of this page is one of the low-poly scenes I build in Three.js. Everything here is written for people who want results from agents, not a new hobby.