GPT-5.5 Instant Is Now Default, Anthropic Doubles Limits, and Governments Lock In AI Access

Published on 07.05.2026

AI & AGENTS

GPT-5.5 Instant Is Now ChatGPT's Default Model

TLDR: OpenAI has swapped ChatGPT's default model to GPT-5.5 Instant, claiming a 52% reduction in hallucinations and adding memory integration across past conversations and Gmail. The change happened quietly, with no major announcement.

GPT-5.5 Instant Just Became ChatGPT's New Default


US Government Locks In Pre-Launch Access to AI Models from Google, Microsoft, and xAI

TLDR: NIST's new Center for AI Standards has signed agreements with Google, Microsoft, and Elon Musk's xAI to test unreleased AI models for national security risks before they ship to the public. This is a formalized government-level evaluation pipeline for frontier models.

GPT-5.5 Instant Just Became ChatGPT's New Default


Anthropic Doubles Claude Rate Limits After SpaceX Colossus GPU Deal

TLDR: Anthropic has doubled the rate limits for Claude Pro, Max, and Team plans while removing peak-hour caps — made possible by securing 300MW of GPU capacity from SpaceX's Colossus 1 data center. This is a significant capacity expansion announced without fanfare.

GPT-5.5 Instant Just Became ChatGPT's New Default


NVIDIA and ServiceNow Launch Project Arc: A Sandboxed Long-Running Desktop Agent

TLDR: NVIDIA and ServiceNow have jointly launched Project Arc, a desktop AI agent designed for knowledge workers that runs in a sandbox environment and is governed by a layer called AI Control Tower. This targets enterprise users who need persistent, multi-step automation without the risk of unconstrained agents.

GPT-5.5 Instant Just Became ChatGPT's New Default