← All posts

n8n or code? Stop picking the wrong one

A workflow tool, a few lines of code, or both: how I decide in five minutes.

Workflow tools are fast to build and easy to hand over. Custom code is easier to test and scale. The choice usually comes down to who will run it next year.

Almost every automation can be built either way: as a workflow in a tool like n8n, or as a small service in code. I use both, and the choice is rarely about what is possible.

When a workflow tool wins

Use n8n when the job is mostly moving data between apps that already have integrations, when the steps are easy to see as a diagram, and when someone on the client’s team will need to change it later without a developer. Being able to open the workflow and see each step run is worth a lot.

When code wins

Write code when the logic has many branches, when it needs real tests, when volume is high, or when the same rules are used in several places. Code is also easier to review and version, so changes are safer.

Often it is both

A common shape is a workflow on the outside and code on the inside: n8n handles the triggers, schedules and app connections, and calls a small service for the part with the real logic. Each tool does what it is good at, and the hand-over stays simple.

Keep reading

  • AI Agents
  • Automation

Let the AI draft. Let a human press Send.

Why the smartest AI agents stop exactly one click short.

The most useful AI agents I build do almost everything, then stop one step short. Here is why the last click belongs to a person, and how to design for it.

  • Email & DNS
  • Backend

Have you ever thought how email works?

One click on Send, six hops, five DNS records and a handshake in plain text.

You press Send and a second later it lands on the other side of the world. Here is the whole journey: SMTP, the servers it passes through, and the DNS records (MX, SPF, DKIM, DMARC) that decide whether it reaches the inbox or the spam folder.

  • AI Costs
  • Architecture

Why did my AI bill just triple?

One gateway, one hard cap, and the two leaks that cause most surprise bills.

Most surprise AI bills come from a few predictable places. A single gateway with logging, model routing and a hard monthly limit closes all of them.

Have a process worth automating?

Tell me what eats your team's week. I'll tell you what I'd automate first.

Start a project Remote · PrimeX Meta, Dubai · teams across time zones