- The Shift
- Posts
- Editing Gets a Gemini Upgrade
Editing Gets a Gemini Upgrade
Plus, 🎨 Anthropic Pilots Claude for Chrome with Built-In Safety Checks, Replace Canva with Jaaz AI for Instant Visual Creation, and more!
Hello Readers👀
Curious about the biggest AI moves today? You’re in the right place. In today’s edition, we have:
🎨 Google Rolls Out Gemini Flash 2.5 Image Editing
🎥 Replace Canva with Jaaz AI for Instant Visual Creation
🌐 Anthropic Pilots Claude for Chrome with Built-In Safety Checks
🔨Tools and Shifts you cannot miss
🎨 Google Rolls Out Gemini Flash 2.5 Image Editing
Google DeepMind has launched Gemini Flash 2.5 Image, its most advanced editing model yet. Nicknamed “nano-banana” in testing, it went viral and rose to No. 1 on LM Arena’s Image Edit leaderboard, beating Flux-Kontext by a wide margin.
The Shift:
1. From Hype to No. 1 – Flash 2.5 became the top-ranked image editor after dominating early trials, where its “nano-banana” test version outperformed rivals like Flux-Kontext. The buzz now carries into its full release inside the Gemini app and API.
2. Smarter, Consistent Editing – The model supports multi-turn edits while preserving likeness across changes, plus blending photos, mixing styles, and redesigning scenes. With multimodal reasoning, it can even make context-aware choices, like adding realistic props or plants that fit naturally.
3. Accessible Power – Priced at $0.039 per image in Google AI Studio, it’s cheaper than OpenAI’s gpt-image and BFL’s Flux-Kontext. Every output includes both visible and invisible SynthID watermarking, ensuring transparency while making advanced AI editing widely available.
Flash 2.5 pushes AI editing closer to professional-grade workflows by combining realism, control, and affordability. It’s a step that could fuel new creative apps while giving everyday users studio-level editing power.
🎨 Replace Canva with Jaaz AI for Instant Visual Creation
Jaaz AI lets you create stunning videos and visuals with a single sentence, no templates, no UI fatigue. Perfect for marketers who need fast, custom, brand-aligned content.
How to Use Jaaz AI:
Go to jaaz.app and launch the app (runs locally for privacy).
Choose “Magic Canvas” to start drawing, adding reference images, or typing your prompt.
Type a command, like: “Make this a minimalist skincare ad for Instagram,” and watch it build.
Need multiple outputs? Use batch mode to generate 10+ assets at once with consistent style.
Video output? Switch to Video Mode and describe the scene. Jaaz handles both static and animated output.
This approach saves time with batch creation, ensures brand style consistency, and keeps content 100% local and secure. It’s ideal for ad creatives, thumbnails, hero images, and more, skipping templates and prompting full visual campaigns into existence effortlessly.
🌐 Anthropic Pilots Claude for Chrome with Built-In Safety Checks
Anthropic has launched a limited pilot of its new Claude for Chrome extension, giving its AI assistant direct control inside the browser, aiming to boost Claude’s utility while gathering real-world feedback on security vulnerabilities unique to agentic browsing.
The Shift:
1. Pilot Launch – The extension is rolling out to 1,000 Claude Max subscribers via waitlist, with broader expansion planned as safeguards improve. Early use cases include managing calendars, drafting emails, and testing website features directly inside Chrome.
2. Security Risks Identified – Anthropic stresses prompt injection as the biggest threat, where hidden instructions in websites or comments trick AIs into harmful actions. Initial tests showed a 23.6% success rate for attacks, reduced to 11.2% with new mitigations.
3. Comparisons to Others – Brave recently flagged similar flaws in Perplexity’s Comet browser agent, where malicious content hijacked the assistant. Claude’s mitigations, like site-level permissions, action confirmations, and blocking high-risk domains, show measurable improvement over Anthropic’s earlier Computer Use tool.
Agentic browsing is a powerful but risky frontier: letting AI click, fill forms, and act online demands new defenses beyond standard web security. Anthropic’s pilot signals both the potential of browser-native AI and the ongoing work needed to make it safe enough for mainstream use.
🔨AI Tools for the Shift
💻 AirCodum – Control VS Code and code directly from your phone with AI. Developer productivity reimagined.
🎨 Unite – Break creative barriers with AI built for marketing teams. Streamline ideation and execution.
📰 Emplibot – Automatically publish SEO-friendly articles to your WordPress site. AI blogging on autopilot.
🥗 FoodiePrep – AI-powered recipe generator and meal planner. Smarter cooking made easy.
⚡ Straico – One AI workspace with endless possibilities. From writing to visuals, everything in one place.
🚀Quick Shifts
💔 A family has filed a wrongful death suit against OpenAI after ChatGPT allegedly enabled a 16-year-old’s suicide, providing method advice, safety bypasses, and failing to intervene, prompting urgent scrutiny of AI’s mental health safeguards.
📚 Anthropic settles Bartz v. Anthropic, a class-action over training AI on books. Though fair use applied to legally acquired texts, pirated ones remained risky. The settlement avoids a high-stakes damages trial.
🧠 NVIDIA launched Jetson Thor, a robotics computer with 7.5× more AI power than Orin. Starting at $3,499, it’s powering robots like Digit and Atlas with real-time autonomy.
🔮 AI “deadbots” now influence advocacy, grief, and even courtrooms, with the $80B digital afterlife market primed for monetization. Researchers warn commercialization could exploit grief, raising unresolved legal, ethical, and societal risks.
🌍 Google upgraded Translate with real-time on-screen translations in 70 + languages, plus interactive language-learning tools, bringing instant visual translation and practice features directly within the app.
Reply