Welcome back!

GitHub Spark lets you turn natural language into functional "micro apps," making app creation accessible for non-programmers and taking on rivals like Replit and v0. And there's more: Copilot now supports Claude and Google Gemini alongside OpenAI. Let’s unpack...

Today’s Summary:

• ⚡ GitHub Spark simplifies app creation

• 📐 Autodesk turns videos into 3D scenes

• 🔲 OpenAI announces custom chips for 2026

• 🔎 Meta's AI-powered search engine in the works

• 🌠 Recraft V3 tops image charts

• 🎧 OpenAI expands Realtime API voices

• 🛠️ 2 new tools

TOP STORY

GitHub Spark lets you build apps using natural language

The Summary: GitHub launched Spark, a tool that creates working “micro apps” from natural language descriptions. Users can build custom applications by explaining ideas in plain English, with real-time previews and automatic deployment. The system manages all technical aspects including hosting, data storage, and AI features.

Key details:

• Apps deployed to desktop and mobile via PWA technology

• Features include managed key-value storage, built in theming, automated hosting and real time previews

• GitHub announced that Copilot can now integrate with Claude and Gemini, in addition to OpenAI

Why it matters: GitHub Spark opens up small-scale software creation to non programmers, challenging other AI coding tools like v0 and Replit. Microsoft expansion of Copilot to include models from Anthropic and Google also heats up competition, putting pressure on rivals like Cursor.

All announcements from GitHub conference

TOOLS

Autodesk Wonder Animation transforms videos into 3D scenes

The Summary: Autodesk Wonder Animation transforms regular videos into fully editable 3D scenes. The AI tool converts multi-camera footage into complete 3D environments with animated characters, all while giving artists full creative creative control. In beta, Wonder Animation lets users turn simple video shoots into professional grade 3D content.

Key details:

• Converts multi-angle footage into a cohesive 3D scene

• Reconstructs camera angles and character movements in 3D space

• Exports to industry tools Maya, Blender, and Unreal Engine

• Lets creators film in any setting, transforming it into a virtual environment

Why it matters: While AI has made major progress in 2D image and video generation, 3D modeling has remained a challenge. Wonder Animation shows that AI companies are beginning to target the 3D domain, suggesting a new wave of AI tools aimed at spatial computing.

Try it in Wonder Studio

OPENAI

OpenAI plans first in-house AI chips for 2026

The Summary: OpenAI is building its first custom chip with Broadcom and TSMC, targeting 2026 for production. The company has assembled a team of 20 engineers, including former Google TPU designers. The project aims to reduce reliance on Nvidia amid a projected $5 billion loss for 2024, primarily due to compute expenses.

• Team includes Thomas Norrie and Richard Ho, key architects behind Google's TPU chips

• Custom chip will focus on inference tasks rather than training, with Broadcom handling design and TSMC manufacturing

• OpenAI dropped plans for its own chip factories for now due to cost constraints

• Company projects $5 billion loss in 2024 despite $3.7 billion revenue, with compute costs as the main expense

Why it matters: This move could shake up AI chip dynamics as leading AI companies seek alternatives to Nvidia 80% GPU market dominance. It has the potential to modify the economics of AI deployment and add fresh competition in specialized chip manufacturing.

QUICK NEWS

Quick news

• The mysterious AI image generator topping charts was Recraft V3

• OpenAI expands Realtime API with 5 new voices

• Meta is working on an AI-powered search engine

TOOLS

🥇 New tools

• Omnigen - Image Generator with consistent visuals

• Magic Notepad by TimeOS - If Apple Notes and ChatGPT had a baby