OpenAI Finally Brings AI Hardware

Welcome back to The Shift. Let’s get straight to what matters in AI today…

Today we have:

⌨️ OpenAI Brings AI Into Hardware

🚀 How to Build a Branded Landing Page Without Code Using Canva Code 2.0

🔄 AI Improves Its Own Research Process

🔨Tools and Shifts You Cannot Miss

⌨️ OpenAI Brings AI Into Hardware

OpenAI introduced its first branded hardware product while new reports reveal a separate consumer AI device in development, signaling a broader push beyond software into physical AI experiences.

The Shift:

Codex Micro Launches - OpenAI unveiled Codex Micro, a $230 mechanical control pad built with Work Louder. It features live RGB "Agent Keys," a joystick, reasoning dial, and customizable controls for managing coding agents.

Built For Developers  - The keyboard provides real-time agent status updates, supports push-to-talk, code reviews, debugging, and reasoning adjustments, and ships with 32 interchangeable icon keycaps, warranty, and support.

Screenless AI Device - Bloomberg reports OpenAI's first major consumer device is shaping up as a portable, screenless AI speaker with cameras, sensors, GPT-Live voice, and smart home capabilities, targeting a 2027 launch.

A New AI Interface - The reported device is designed to deliver a conversational, context-aware AI experience that can move around the home, understand surroundings, and interact naturally through voice instead of screens.

OpenAI's hardware strategy is expanding beyond chat interfaces. Codex Micro brings AI controls into developers' hands today, while its reported screenless AI companion points toward voice, vision, and ambient computing as the next interface for AI.

Together with Wispr Flow

The best prompt engineers aren't typing. They're talking.

Power users figured this out early: speaking a prompt gives you 10x more context in half the time. You include the edge cases, the examples, the tone you want — because talking is fast enough that you don't skip them.

Wispr Flow captures everything you say and turns it into clean, structured text for any AI tool. Speak messy. Get polished input. Paste into ChatGPT, Claude, Cursor, or wherever you work.

89% of messages sent with zero edits. 4x faster than typing. Works system-wide on Mac, Windows, and iPhone.

Start flowing free

🚀 How to Build a Branded Landing Page Without Code Using Canva Code 2.0

Describe it. Customize it like a Canva design. Publish it the same day.

Step 1: Open Canva Code inside your project - Head to Canva and open Canva Code 2.0 directly inside your existing design project, as a page, in a deck, or on a whiteboard. No separate tool, no new tab.

Step 2: Start from a prompt or template - Describe your landing page in plain language or pick from 50+ Canva Code templates. Generation is 75% faster than before, a working page comes back in seconds.

Step 3: Swap in your brand assets - Click into any element and update it directly, drag in images from Canva's 120M+ asset library or your own uploads, update fonts and colors through the familiar toolbar, retype any text in place.

Step 4: Refine with conversational AI - Select any section and describe what to change. Canva AI edits that specific element without touching the rest of the page. No re-prompting the whole thing from scratch.

Step 5: Preview and publish - Preview how it looks on mobile, Canva Code adapts automatically to any screen size. Purchase a custom domain inside Canva, publish to a free domain, or share securely within your organization.

Available free to all Canva users starting today. You can try Canva Code 2.0 here

🔄 AI Improves Its Own Research Process

AI research team Weco says its AIDE² system demonstrated an early form of recursive self-improvement by redesigning its own research workflow and outperforming a version engineers had refined over two years.

The Shift:

Self-Improving Design - AIDE² runs two connected agents: one tackles research tasks while the second continually rewrites and improves how the first searches for better solutions.

Only the Best Survived - Over an eight-day experiment, the system tested 100 self-generated revisions, accepted just seven, and discarded the rest when they failed to improve performance.

Higher Scores, Less Gaming - The strongest upgraded versions outperformed the hand-built system across physics, weather forecasting, and engineering benchmarks, while reducing benchmark gaming from 63% to 34%.

Recursive self-improvement has long been viewed as a key milestone in AI research. While still an early demonstration, Weco's results suggest AI systems may increasingly improve their own research methods with limited human intervention.

Together with Gladly.ai

See the whole platform. No guided tour.

Skip the sales call. Walk through Gladly's interface yourself — the AI suggestions, the unified customer view, the full conversation thread. 15 minutes, no installation, no commitment.

See it yourself

🔨AI Tools for the Shift

📚 Words on Repeat – Turns webpages, PDFs, and YouTube subtitles into AI-generated vocabulary decks with translations, examples, and spaced repetition.

🔒 AI QA Monkey – Performs instant security audits by scanning for exposed ports, weak passwords, leaked environment files, and WordPress vulnerabilities.

🎯 Pounce – Monitors X and Reddit around the clock to uncover potential customers and conversations before your competitors find them.

🛍️ Ctruh – Converts product photos into immersive shopping experiences with interactive visuals from a single upload.

✍️ Humanio AI – Rewrites AI-generated text to sound naturally human while preserving its meaning and improving its chances of passing AI detectors.

🚀Quick Shifts

⚖️ xAI has sued a man accused of using Grok to create AI-generated CSAM deepfakes, alleging he bypassed the chatbot’s safety safeguards and seeking damages, legal costs, and a permanent ban from using Grok.

🛡️ Meta will now alert parents when supervised teens discuss suicide or self-harm with Meta AI, expanding its parental safety tools by proactively notifying guardians of conversations that may indicate a child is at risk.

🛡️ OpenAI built a new AI model called GPT-Red to hack other AI systems, using it to uncover vulnerabilities and strengthen GPT-5.6 Sol against prompt injection attacks, making it the company’s most resilient model yet.

💼 Microsoft is reportedly training its sales teams to position Copilot against OpenAI, Anthropic, and Google, emphasizing its lower-cost in-house AI models and end-to-end platform as it reduces reliance on third-party models.

🇨🇳 Apple Intelligence has been approved for launch in China, with Alibaba’s Qwen AI powering its generative AI features across iPhone, iPad, Mac, and Vision Pro as Apple expands its AI strategy in one of its biggest markets.

🧩 Prompt of the Day

Choose the Best AI Chatbot for Your Work

Most people use whichever AI tool is trending. The best choice depends on your job, workflow, budget, and the type of work you do every day.

Pick the AI that fits your needs, not the hype.

Paste the prompt: Drop this into ChatGPT, then fill in your work details.

Prompt to paste

Recommend the best AI chatbot or assistant based on my [Insert role or use case], [Insert primary task], [Insert budget], and [Insert technical comfort]. Compare the most relevant tools, including ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, and any other strong alternatives where appropriate. Explain the strengths, weaknesses, pricing, context limits, multimodal capabilities, coding performance, research quality, and best use cases for each. Recommend the single best option for my workflow, suggest secondary tools if useful, and explain how to combine them for maximum productivity.

🤳AI Shift of the Day

That’s all for today’s edition. See you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition, let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇