cd /news/artificial-intelligence/our-philosophy-for-coding-with-ai · home topics artificial-intelligence article
[ARTICLE · art-72736] src=axiacore.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Our philosophy for coding with AI

Axiacore CEO Camilo Nova argues that adopting AI for coding is a spectrum, not an all-or-nothing process, and outlines a five-concept philosophy for using tools like Codex and Claude Code: plan and adjust before implementing, read every line of generated code, simplify further, prioritize fewer lines for more value, and ensure each line has a purpose.

read2 min views1 publishedJul 24, 2026
Our philosophy for coding with AI
Image: Axiacore (auto-discovered)

Blog

Camilo Nova

CEOWe are all learning to adopt AI when building software. Some people claim you can code anything; they say whatever software you need can be built without knowing how to code, and that's it. I think that's wrong for building real products.

Adopting any AI tool isn't a 0-to-100 process; it's more of a spectrum, like adjusting the temperature. How much has to be according to what feels right for what you want.

Think of it like adding salt when cooking: there isn't an exact amount; it's totally up to who's cooking.

At Axiacore, we have been using Codex over Claude Code. We are being cautious of the token spend. It's easy to add more lines of code, and hard to remove them. The incentive of any LLM is to produce as many tokens as possible, if that's their business model.

After many weeks working with those tools, our philosophy can be summarized (not by AI) in five concepts:

Plan. Adjust. Implement. #

For any feature, start by writing a plan. You can do it by changing the mode to /plan and thinking hard about the feature, making it simple, and avoiding over-engineering. When the plan is written, go back and read it. Find where to simplify, and run the plan again. Adjust until it feels right (two or three times). Then, and only then, implement the code.

Read. Every. Line. #

Whatever the output you get from the last step, make sure to read every single line. This is important; the code is meant to be understood by a person, not just by the LLM. It's dangerous to have code that no one knows how it works. Remember: read every single line.

Simplify even more. #

After the last step, continue simplifying further. The feature has to fit in your mental model; you are the only one who can make connections and have the whole context, that's the fuel of the "feels right" part.

Fewer lines. More value. #

More code doesn't mean better; it's the opposite. Work through the simplest and most elegant implementation you can. This is the part where your genius shines, and you defeat the laziness most people have.

Each line must have purpose. #

Finally, be aware that each line has a purpose. Don't be the guy who adds files or code that you don't have any idea what they do. That's just sad. You can do better.

Written by Camilo Nova

Axiacore CEO. Camilo writes thoughts about the intersection between business, technology, and philosophy

Subscribe to our newsletter here:

Learn how to use technology to get back your time and enjoy an empty calendar on a work day.

We respect your inbox. Privacy policy

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @axiacore 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/our-philosophy-for-c…] indexed:0 read:2min 2026-07-24 ·