Gemini 3.1 Pro, OpenAI Hardware Dreams, and the Billion-Dollar AI Bets Shaping 2026

Published on 23.02.2026

AI & AGENTS

Google Gemini 3.1 Pro: Multi-Step Reasoning Gets a Dedicated Model

TLDR: Google released Gemini 3.1 Pro, a model purpose-built for complex, multi-step reasoning tasks. Instead of bolting reasoning onto a general-purpose model, they built one from the ground up for sophisticated problem-solving.

Gemini 3.1 Pro Is Here (and it's built for multi-step thinking)


OpenAI GPT-5.3-Codex-Spark Hits 1,200+ Tokens Per Second on Cerebras

TLDR: OpenAI's coding model GPT-5.3-Codex-Spark now runs 30% faster on Cerebras hardware, crossing the 1,200 tokens-per-second threshold. This makes real-time AI-assisted coding dramatically more responsive.

OpenAI GPT-5.3-Codex-Spark Speed Boost


OpenAI Goes Physical: Jony Ive-Designed AI Hardware Starting 2027

TLDR: OpenAI is developing AI-powered physical devices designed by Jony Ive, including a $200-$300 smart speaker launching in 2027, with potential smart glasses and a lamp in the pipeline.

OpenAI AI Hardware Plans


Anthropic Ships Claude Code Security: AI-Powered Vulnerability Hunting

TLDR: Anthropic released Claude Code Security in preview, an AI reasoning tool that scans codebases for vulnerabilities and suggests patches, targeting gaps that traditional static analysis tools miss.

Claude Code Security Preview


YouTube Tests Conversational AI Assistant on Smart TVs

TLDR: YouTube is testing an AI assistant on smart TVs that lets viewers ask questions about videos they are watching in real time, bringing conversational AI to the lean-back entertainment experience.

YouTube AI Assistant on Smart TVs


AI Investment Roundup: Record-Breaking Rounds Signal Market Conviction

TLDR: OpenAI is closing a $100B+ round at $850B+ valuation, World Labs raised $1B for spatial AI, and Ineffable Intelligence is raising what could be Europe's largest-ever seed round at $1B.

The AI Break - Investments Section