cd /news/ai-agents/the-starting-line · home topics ai-agents article
[ARTICLE · art-129575] src=mikaei.substack.com ↗ pub= topic=ai-agents verified=true sentiment=· neutral

The Starting Line

A developer argues that AI coding tools like Codex and Claude serve as a shortcut to the starting line of software projects, freeing engineers from setup and configuration drudgery so they can focus on the problems they actually want to solve. The developer contends that this automation demands discipline: building lasting context that outlives an LLM's fleeting attention span, extracting verifiable deterministic logic from non-deterministic processes, and maintaining personal mental models to take responsibility for AI-generated code. Without that discipline, the developer warns, it is like building a house you have no intention of living in yourself.

by read2 min views1 publishedSep 14, 2026
The Starting Line
Image: source

Tiger got to hunt, bird got to fly;

Man got to sit and wonder “why, why, why?”

Tiger got to sleep, bird got to land;

Man got to tell himself he understand.

― Kurt Vonnegut, Cat’s Cradle

The urge to understand has been one of my strongest driving forces as a developer. Does it remain just as strong in the face of the upheaval AI brings?

Before AI, I often spent much of my time just getting to the starting line. There was so much to understand, set up, connect or configure before I could tackle the problem I actually wanted to solve. Then the environment would stop working. By the time I was finally ready to begin, it would already be Friday.

With that still fresh in my mind, it feels almost unnatural to watch Codex or Claude chew through a complex problem with the appetite of a thousand undead developers. I’ve had that creeping feeling that I’m being replaced. But is it that simple?

I see AI as my shortcut to the starting line. It lets me start closer to the point where my abilities get to shine. Now I’ll parachute into places I once had to make my way through the jungle to reach.

But there’s still a jungle down there. Reality is detailed and messy, and understanding is as crucial as ever. I need to distinguish between the details I can leave to the machine and those I need to have a firm grasp on. There’s a paradox in this liberating automation. Coding with AI also has a surprising amount to do with freeing yourself from it:

- I need to build [lasting context](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) that outlives the LLM’s fleeting attention span.
- I need to [extract verifiable, deterministic logic](https://replicated.live/blog/away) from the LLM’s non-deterministic processes.
  • I need to develop and maintain my own mental models so I can understand and take responsibility for what the LLM produces.

Without this discipline, it’s like building a house I have no intention of living in myself. The excitement of getting the house finished can overshadow my interest in what it will be like to live there.

Curiosity and the urge to understand are qualities that never go out of fashion. I still have to take responsibility for building houses fit to live in, and be honest about which rooms I have yet to enter. The difference is that I now get to the starting line faster.

── more in #ai-agents 4 stories · sorted by recency
── more on @codex 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/the-starting-line] indexed:0 read:2min 2026-09-14 ·