Blog
Notes on running coding agents.
Guides, comparisons and field notes from the team that builds Fredrin.
Codex vs Claude Code: the verdict from running both
Codex vs Claude Code, judged by a tool that runs both as runtimes: where each one wins, what they really cost, and the bottleneck neither of them fixes.Claude Max for coding: is it worth it? The real math
Claude Max is $100 or $200 a month. Here is the arithmetic against metered API pricing, what the 2026 OAuth rules actually allow, and when Pro is enough.How to use Claude Code: setup, context, and parallel work
Install Claude Code, log in, write a CLAUDE.md that survives, set permissions you can live with, and run several sessions at once without them colliding.Git worktrees: the complete guide to working in parallel
What a git worktree is, the four commands that matter, the gotchas nobody mentions — env files, ports, hooks — and why AI agents made worktrees mainstream.Vibe Kanban is community-maintained now. Here's how to move your board.
Bloop wound down in April 2026 and Vibe Kanban became community-maintained open source. What changed, and how to move a board across.What is an AI coding agent? Definition and honest limits
An AI coding agent is a model with tools, a workspace and a stopping condition. What that buys, how agents differ from assistants, and where they fail.The FDE tool stack in 2026: build, deploy, iterate, and the missing work surface
Discover, build, deploy, iterate — and the layer nobody built. A survey of the 2026 FDE stack, competitors included, and where the real gap is.AI code review: tools, workflow, and the human gate
Every major AI code review tool comments and none of them blocks the merge. The tools, honestly compared, and the workflow for reading diffs at volume.Agentic coding: what it actually means in practice
Agentic coding is an AI agent reading your repo, editing files and running commands until a check passes. What it is, how it differs, and where it fails.Shipping AI-generated code a client will accept: the review-first workflow
Clients reject AI-written code for the diff, not the author. The review-first workflow that makes a change acceptable regardless of who wrote it.How FDEs run five client engagements in parallel with coding agents
Not by working five times faster. By isolating each engagement so switching costs a glance, and making review — not typing — the thing you optimize.Forward deployed engineer interview questions (2026)
What an FDE loop tests that a standard engineering interview doesn't: scoping ambiguity, building on the customer's product, and owning the outcome.FDE vs solutions engineer vs sales engineer: what actually differs
A sales engineer is measured on the deal, a solutions engineer on adoption, an FDE on whether the software works in production. What actually differs.cmux solved spawning. Nobody solved shipping.
If you're reading this, there's a decent chance you have six terminals open right now, each running Claude Code or Codex in its own git worktree.Why Fredrin — the unified context layer for vibe coding
I built Fredrin because I kept losing the plot.