Criador do bridgecode

Marcelo Santiago

Developer and creator of bridgecode, an agentic development environment for orchestrating squads of AI agents. Helps people going from vibe coding to professional development move from idea to shipped product.

GitHub Twitter / X LinkedIn Instagram Threads Website

Artigos publicados

252 artigos por Marcelo Santiago

Opera WebAI panel on a dark background shows a connected engine, selected Gemma models and an embeddings option.

Newsletter

Oct 27, 20266 min

Opera WebAI brings local models to the browser

Opera launched WebAI in preview in Opera One, with local models, opt-in activation and per-site permissions for AI in pages and extensions.

#Opera#WebAI#IA local

Marcelo Santiago Read

Light, out of focus flowers in the foreground against a blue sky, official image associated with the MAI Image 2.7 announcement.

Benchmarks

Oct 27, 20266 min

MAI Image 2.7: 2nd on Artificial Analysis and 3rd on Arena

MAI Image 2.7 placed 3rd on Arena and 2nd on Artificial Analysis; the announcement highlights advances in scene editing, text and restoration.

#MAI Image 2.7#Microsoft AI#Image Edit Arena

Marcelo Santiago Read

Promotional image of Genie 4 with a long legged bird playing piano inside a sphere on the snow, accompanied by the text Genie 4, Where next?

Newsletter

Oct 27, 20266 min

Genie 4 builds interactive worlds with native audio

Google DeepMind announced Genie 4 on October 27, with real time interactive simulation, native audio and limited initial access.

#Genie 4#Google DeepMind#modelos de mundo

Marcelo Santiago Read

Dicas

Oct 27, 20265 min

How to make money with AI: validate a service

Validate a real pain point, sell an AI-powered service before automating everything, and turn the workflow into a product only when there is demand.

#ganhar dinheiro com IA#serviço de IA#produto digital

Marcelo Santiago Read

Grid with ten examples of XR apps evaluated by GenXR-Bench, including volume visualizer, voxel garden, virtual brush, fruit slicer and rainbow pen.

Benchmarks

Oct 27, 20265 min

GenXR-Bench: Claude Opus 5 scores 50.9% on XR

GenXR-Bench brings together 112 XR tasks. In paper v2, Claude Opus 5 reached 50.9% Pass@1, with 29.2% on SDK feature tasks.

#GenXR-Bench#XR#coding agents

Marcelo Santiago Read

Acc-Cost Pareto chart for LongMemEval-s with Qwen3.6-27B, showing accuracy, token cost per corpus and storage size.

Newsletter

Oct 27, 20265 min

Memory Tax exposes the hidden cost of agent memory

Preprint on arXiv compares memory systems and RAG across three benchmarks and reports cases where memory reduces accuracy.

#Memory Tax#memória de agentes#RAG

Marcelo Santiago Read

Newsletter

Oct 27, 20265 min

Simplismart raises US$14 million in Series B

Simplismart announced a US$14 million round led by DVC, with NVIDIA and Accel, to expand its AI inference platform.

#Simplismart#inferência de IA#infraestrutura de IA

Marcelo Santiago Read

Newsletter

Oct 27, 20265 min

GPT-6.1 Terra joins Agent Arena, scores still pending

Arena has added GPT-6.1 Terra to long tasks with web, filesystem and terminal access, and the official account said scores will still be released.

#Arena#Agent Arena#GPT-6.1 Terra

Marcelo Santiago Read

Illustration of Cloudflare with a magnifying glass over a blue droplet inside a circuit, next to the title about KV cache exposure in GLM-5.3 Workers AI.

Newsletter

Oct 27, 20266 min

Cloudflare patches isolation flaw in GLM-5.3

Cloudflare says 155 GLM-5.3 responses included data from other requests; the KV cache and SGLang flaw was fixed in September.

#Cloudflare#Workers AI#GLM-5.3

Marcelo Santiago Read

Official GitHub Copilot artwork with a code screen, chat panel and blue mascot under the title “AI that builds with you”.

Dicas

Oct 27, 20266 min

Free AI coding agents: 3 options

Compare GitHub Copilot Free, OpenCode and Cline. See the limits and learn how to avoid model costs and privacy surprises.

#agentes de IA#programação#GitHub Copilot

Marcelo Santiago Read

Official Code2Video-AA artwork on two film strips, with the names Artificial Analysis and Code2Video.

Benchmarks

Oct 27, 20266 min

Code2Video-AA benchmarks 27 models on code to video

Code2Video-AA evaluated 27 models across 114 briefs; see the leaderboard, render rate and benchmark limits.

#Code2Video-AA#Artificial Analysis#benchmark

Marcelo Santiago Read

Meta and NVIDIA logos above two stylized characters, one light and one with a black jacket, in the partnership announcement image.

Newsletter

Oct 27, 20265 min

Meta and NVIDIA expand partnership on Muse models

Meta and NVIDIA will optimize the Muse family and expand infrastructure for agents, including Vera Rubin, CPUs Vera, Spectrum-X and BlueField-4.

#Meta#NVIDIA#Muse

Marcelo Santiago Read

The biggest AI newsletter in the World

Follow the evolution of AI squads and multi-agent systems.

In-depth technical curation: agentic architectures, real token benchmarks, MCP servers and open-source tools in your inbox and on the blog.