O Gemini quebrou? Por que mudei para o Codex 5.4
If Gemini was my right-hand man in programming, lately it has let me down, and I've had to look for alternatives to avoid stalling the development of my system. In this video, I'll be honest about my real-life experience of two months using Gemini to build Promovaweb's internal LMS and how the quality of the 3.1 Pro model has drastically degraded in recent weeks. See also: Free Immersion 2026: https://comunidade.promovaweb.com/c/i... Schedule a Consultation: https://promovaweb.com/comercial Discover our Courses: https://promovaweb.com Join our community: https://comunidade.promovaweb.com Join the Whatsapp Group: https://promovaweb.com/whatsapp Join the Discord Group: https://promovaweb.com/discord I'll tell you how he started making serious mistakes, such as deleting my development database and completely ignoring Laravel skills, in addition to exhibiting latency that makes the work drag on. In contrast, I tested Codex 5.4 and the result was brutal: much cleaner, optimized code and an assertiveness that makes me twice as productive, even with the message limitations. I'll show you why, for complex systems, Gemini is losing steam while Codex delivers senior programmer performance. I'll also talk about our AI Makers and Founders training programs, for those who don't want to go through this journey alone and want to accelerate the launch of their product or SaaS with the support of our community. If you feel your AI productivity has dropped or want to know which model to choose for your back-end, this discussion is essential. 00:00:00 Degradation and latency in Gemini 3.1 00:01:02 Serious errors: Test breakage and database deletion 00:01:38 Differences in usage between Gemini 3.1 Pro and 3.0 Flash 00:02:40 Problems with MCP and Skills in Laravel 00:04:51 Failures in migration creation and architecture 00:07:22 Gemini for programmers vs. VibeCoders 00:10:04 The turning point with Codex 5.4 00:11:40 Code quality comparison: Gemini vs. Codex 00:14:18 Productivity and message limits 00:17:30 Resource consumption and CLI performance 00:18:02 Gemini's strengths: Design and Landing Pages 00:19:05 AI Makers and Founders Training at Promovaweb 00:23:17 Consulting and final tool tips #Gemini #Codex #AIProgramming #Laravel #VibeCoding #SaaS #SoftwareDevelopment #Astro #AI #Promovaweb

The end of traditional SaaS? Meet Selfware

Android 17 sucks. So I put Linux on a phone.

GPT-5.5 Beats Fable, Cursor Takes On GitHub & Midjourney Scans Your Body | This Week in AI

Watch before you start on Vibe Coding (2026)

Codex vs Cowork for Regular People (Every Feature Compared)

Crie seu Servidor de API WhatsApp com Whatsmeow + Whatsmiau no Easypanel em Menos de 15 Minutos

The secret to professional n8n workflows

New Axial-Flux Motor: 350% More Power Than Any Electric Motor!

Will the Intel Arc Pro B70 32GB replace the Nvidia RTX 5090? I tested it with LLM and image gener...

6 AI Businesses That Still Have No Competition

Why we left Norway 🇳🇴 | An honest recap!!!

What is Harness Engineering?

How to avoid WhatsApp bans using the multi-instance rule

SetupVibe: Dev environment configured in 5 minutes

What do tech pioneers think about the AI revolution? - The Engineers, BBC World Service

🇩🇪 "SABER FALAR ALEMÃO ME ABRIU PORTAS QUE EU NEM SABIA QUE EXISTIAM"

Anthropic is Completely F*cked.

An overview of Vibe Coding and n8n

I tested the new /goal function in Codex on truly long tasks.

