Sora 2 and its New Social Media | Weekly Edition

PLUS HOT AI Tools & Tutorials

Oct 03, 2025

Hey there! Welcome to your latest Creators’ AI Edition!

This week, OpenAI unveils Sora 2, a video model with synchronized audio, plus a new interesting Sora app. Anthropic launches a new model for coding (is it the best?), OpenAI adds Instant Checkout to ChatGPT, Apple tests a smarter Siri, and Google pushes forward with Gemini Flash updates and Dreamer 4 in Minecraft.

From video to commerce to coding, the AI race is heating up

Let’s dive in! 🚀

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

Featured Materials 🎟️

Sora 2 is Here

OpenAI released Sora 2, a video model that now generates synchronized audio and dialogue alongside 1080p visuals up to twenty seconds, with major gains in physical realism and multi shot consistency. Compared to earlier versions, Sora 2 better preserves cause and effect in scenes like sports or gymnastics and can maintain a coherent world state across scene changes. It creates matching sound effects and speech that align with visual style, and early releases show longer clips in the five-to-ten-second range with improved motion and continuity.

A Sora 2 Pro tier is planned, and API access is coming soon.

To create Sora 2, Sam Altman had to steal chips from Target. Link to CCTV footage joke :)

Another Sora?

OpenAI launched a new Sora social app that centers on Cameos, which lets users record their likeness and voice once, then appear inside any AI-generated scene or remix. The feed is composed entirely of Sora created videos, and users can prompt from text, add characters with scripts, insert themselves or friends, and spin off variations from others. The app launches free with usage limits in the United States and Canada, invites allow four friends per user, and Pro subscribers get access to the Sora 2 Pro model. The goal is to make the rapid creation and sharing of short, synchronized audio-video stories mainstream.

Thanks for reading Creators' AI! This post is public so feel free to share it.

Anthropic Launches Claude 4.5

Anthropic launched Claude Sonnet 4.5, billed as the best coding model, with SOTA on SWE-bench Verified, nearly 20% improvement over Claude Opus 4.1 for computer use, and autonomous coding runs of 30+ hours producing 11,000 lines of code, while keeping API pricing unchanged; platform upgrades include Claude Code with checkpoints, memory and context editing in the API, a Claude Agent SDK for building agents, and a 5-day “Imagine with Claude” preview showing real-time software generation, positioning Anthropic to counter OpenAI’s GPT-5-Codex momentum.

Instant Checkout inside ChatGPT

OpenAI rolled out Instant Checkout for U.S. users so purchases happen inside ChatGPT, partnering with Stripe to process payments, starting with Etsy sellers and expanding to over 1 million Shopify merchants soon; shoppers tap Buy, confirm details, and pay in chat, while OpenAI open-sourced the Agentic Commerce Protocol (ACP) so any retailer can integrate it, Stripe merchants need only minimal code changes, merchant fees apply only on completed sales, and product rankings remain organic by relevance, marking a key step toward agentic AI commerce and a new revenue stream for OpenAI.

Siri is going to become smarter

Apple is internally testing a ChatGPT-style tool called Veritas to stress-test the coming Siri upgrade, per Bloomberg. Veritas trials include searching personal data and editing photos by voice, and it exercises Linwood, a system that combines Apple models with third-party options. Engineering issues pushed the AI-powered Siri launch target to March 2026, triggering executive reshuffles and a talent exodus to rival AI labs. Apple is not planning a standalone Veritas app, opting instead to embed these capabilities directly into Siri, a strategy that avoids a head-to-head chatbot fight while it races to close the gap.

Google updates Gemini 2.5 Flash and Flash-Lite

Intelligence vs End-to-End response time

Google DeepMind shipped preview upgrades of Gemini 2.5 Flash and 2.5 Flash-Lite in AI Studio and Vertex AI, cutting output tokens by 24% and 50% respectively, while boosting quality. Flash-Lite now follows complex instructions better, is less verbose, and improves multimodal transcription, image understanding, and translation; use gemini-2.5-flash-lite-preview-09-2025. Flash improves agentic tool use with a 5% gain on SWE-Bench Verified (48.9% → 54%) and delivers higher quality with fewer thinking tokens; use gemini-2.5-flash-preview-09-2025. New -latestaliases—gemini-flash-latest and gemini-flash-lite-latest—always point to the newest previews, with two-week change notices, while gemini-2.5-flash and gemini-2.5-flash-lite remain the stable choices.

Dreamer 4 By Google DeepMind

Google DeepMind introduced Dreamer 4, an agent that learns entirely inside a predictive world model and becomes the first to collect Minecraft diamonds using offline data only. The model practices in real time from pixels, executing 20,000-plus actions in its internal simulation, and follows a staged path from video learning to decision making to practice without touching the live game. Testers completed 14 of 16 tasks in Dreamer 4’s sim versus 5 with rivals like Oasis. Dreamer surpassed OpenAI’s Minecraft VPT while using about 100 times less data, and beat approaches built on Gemma vision language models. The result suggests safer, cheaper training paths where simulation accelerates robotics and complex agent skills beyond gaming.

Useful tools ⚒️

C1 by Thesys - Make AI apps respond with interactive UI in real-time

CrePal - Create short movies with just a prompt with our AI director

Caesr AI - AI that controls real apps across web, desktop, and mobile

Fern - Instant Docs and SDKs for your API

Verdent Deck - Your AI code editor that runs multiple AI agents in parallel

Verdent Deck coordinates multiple AI agents to tackle complex coding tasks in parallel. Sessions can run on their own while you step away, with collision-free execution, clear insight, and a seamless flow that turns ideas into real, shippable code.

Share this post with friends, especially those interested in AI!