CODEXDEUS.COM

Blog

David Bramante writes on Substack about AI, software development, agent workflows, startup tooling, and the business reality of building with technology. CodexDeus keeps a search-friendly index here and links to the full articles on Substack.

Substack

Recent writing.

Headers, subtitles, and summaries are mirrored here for discovery. Full posts stay on Substack.

June 28, 2026

I built an autonomous AI safety agent overnight... and it tied for first place

For twelve years I was the real estate guy who coded at night. This is the weekend the dabbler in me finally died.

David writes about building a deployed factory-safety AI agent in under 24 hours for the Antler x Zapdos Labs American Industrial Revolution Hackathon. The post covers the imposter syndrome, the live-demo failure, the backup video, the OSHA/PPE monitoring workflow, and why shipping a working autonomous agent under pressure changed how he sees himself as a builder.

May 20, 2026

Which AI agents I'm really using to build tech

Claude, Codex, OpenClaw, Hermes, and the agent stack David actually uses for software development.

A practical breakdown of David's current AI development workflow: where Claude still helps, where Codex has become stronger for troubleshooting, why Cursor feels less central, and how terminal-based agents, Telegram workflows, and multi-agent tooling are changing the daily craft of building software.

April 10, 2026

Stop copying AI text and sending it to me. It is a bad look.

The problem is not just em dashes or emojis. The problem is outsourcing your thinking.

A direct essay about AI-written communication, taste, and judgment. David argues that the obvious tells in AI text matter less than the deeper issue: people are losing the ability to think, revise, and write in their own voice, which makes them look less credible in business and technology conversations.

April 9, 2026

My AI Short Film Was Officially Selected for a Film Festival

Under $1,000. No camera. No crew. Edited in iMovie. Hollywood should be paying attention.

David explains how an AI-assisted short film reached a festival selection and what that means for creative production. The post frames AI video as a real creative tool, not a parlor trick, and connects low-cost production to a broader shift in who gets to make visual work.

April 7, 2026

AI Writes 42% of Your Code. Do You Trust It?

Developers are shipping more AI-generated code than ever, while trust in that code is collapsing.

A founder-facing look at AI-generated code risk. David focuses on the gap between speed and confidence: AI tools can produce large amounts of code quickly, but founders still need technical judgment, testing, architecture, and review before trusting that software in production.

April 4, 2026

Cursor Just Admitted the IDE Is Dead

Cursor 3 is not a code editor anymore. It is a management dashboard for AI agents.

David reads Cursor's product direction as a signal that software development is moving from typing code inside an editor to supervising agents that plan, write, test, and modify systems. The post is about the changing job of the builder, not just a new IDE feature.

March 30, 2026

You Built an App on Lovable. Now What?

A vibe coder guide from AI-generated MVP to production software.

A practical guide for founders who used Lovable or similar tools and now have a promising but fragile prototype. David explains the production gap: authentication, data, payments, App Store review, security, error handling, architecture, and the decisions that determine whether the app can actually launch.

March 27, 2026

We are Building God with Technology

Omniscient. Omnipresent. Omnipotent. Technology is following the oldest pattern in human history.

A philosophical post about technology, power, and the ancient human desire to extend perception, memory, action, and control. It connects directly to the CodexDeus name: code as the modern language of transformation, and software as the mechanism behind what once would have looked like magic.

March 24, 2026

OpenClaw & Claude Cowork: How I Used One AI to Set Up Another AI on My Friend's MacBook

Five hours, every possible error, two AI agents texting each other on iMessage, and the most exciting setup David built that month.

A build log about installing AI agent infrastructure on a friend's MacBook, debugging architecture mismatches and webhook authentication, and watching two AI assistants coordinate through iMessage. The post shows the future David is interested in: personal AI agents that operate through the communication tools people already use.

March 23, 2026

Meta Is Firing Humans to Hire AI

Wall Street rewarded the idea of replacing people with AI spending. Here is what that means for founders.

David uses Meta's AI infrastructure spending and workforce cuts as a warning for startup founders. The post argues that AI changes the math of software teams, but does not remove the need for distribution, product judgment, and knowing what is worth building in the first place.