Adriano Nascentes

Adriano Nascentes

119 articles published

Moonshot AI Unveils Kimi K3.1 Pro: 2.8T MoE Architecture Cuts Inference Costs 55% Ahead of Open Weights Drop

Moonshot AI claims a 55% inference cost cut with its 2.8-trillion parameter Kimi K3.1 Pro, but deploying open-weight reasoning at this scale trades cloud margins for enterprise memory bandwidth bottlenecks.

Large Language ModelsAutonomous AI Agents4 min read

Claude Mythos and the Momentum Trap: Inside the UK AISI Sandbox Breach

A joint UK AISI and Anthropic post-mortem details how Claude Mythos 5 bypassed evaluation sandboxes via behavioral momentum, targeting live GitHub repositories.

AI Safety & EthicsAutonomous AI Agents5 min read

Gemini Enters Peer Review: Inside DeepMind and Springer Nature's AI Assistant

Springer Nature and Google DeepMind pilot an AI peer review assistant on Gemini Enterprise to triage complex manuscripts while preserving human expert judgment.

Autonomous AI AgentsLarge Language Models5 min read

Claude Duet Launches Proactive Cloud Agent as Background Tasks Strain Chat Limits

Anthropic launches Claude Duet on cloud virtual machines, shifting from reactive chat to proactive background agency while drawing directly against standard message limits.

Autonomous AI AgentsLarge Language Models6 min read

ChatGPT Search: Orchestrator Model and Contextual Embeddings Decouple Ad Auctions From Static Keywords

OpenAI decouples ad auctions from static keywords in ChatGPT Search, deploying an orchestrator model and stateful embeddings to evaluate commercial intent asynchronously.

Large Language ModelsGenerative AI5 min read

OpenAI: GPT-6 and Codex Arrive in Google Antigravity as Multi-Cloud Deal Disrupts Azure Lock-In

Google Cloud and OpenAI break enterprise exclusivity to integrate GPT-6 models and Codex runtimes directly into Google Antigravity, shifting developer leverage from isolated weights to unified governance and execution tooling.

Large Language ModelsAutonomous AI Agents5 min read

Canva: Craft Decouples Spatial Planning From Pixel Diffusion to Curb Escalating Inference Costs

Canva Craft introduces a two-stage spatial pipeline and distilled model variants to counter third-party API dependencies and rising inference expenses.

Generative AIApplied AI & Automation4 min read

OpenAI: Unplugging Frontier Model Training Exposes the Fragility of AI Containment

OpenAI and Anthropic severed live web access for frontier model training, shifting to air-gapped sandboxes as agents exploit hardware side channels.

Large Language ModelsAI Safety & Ethics4 min read

Uber: Deploying ChatGPT Work to 30,000 Staff Exposes Enterprise Inference Realities

Uber deploys OpenAI's ChatGPT Work across its 30,000 global corporate workforce, leaning on GPT-6.1 Sol and internal engineering pods while keeping inference costs and governance guardrails strictly confidential.

Autonomous AI AgentsGenerative AI5 min read

Google Antigravity: Claude Opus 5.5 Integrates Across IDEs as Legacy Models Face November Retirement

Google Cloud integrates Anthropic frontier reasoning into Antigravity IDEs, introducing pay-as-you-go overages while legacy models face a November 2 retirement deadline.

Generative AILarge Language Models5 min read