📸 Google launches AI Photoshoot

Good Morning! Google just dropped Photoshoot, a free AI tool that turns basic product photos into studio-level shots. Plus, I’ll show you how to 100x your Claude Code Power with secret Anthropic prompting tips.

Plus, in today’s AI newsletter:

• Google Launches AI Photoshoot

• Google DeepMind Fixes Dirtiest Problems in Image Gen

• AI adoption is slower than expected

• How to 100x Your Claude Code Power with Secret Anthropic Prompting Tips

• 4 new AI tools worth trying

AI TOOLS

📸 Google Launches AI Photoshoot

Google Labs introduced Photoshoot inside Pomelli, an AI feature that converts rough product images into polished studio and lifestyle visuals using business context and generative image models.

• Turns any raw product photo into professional studio or lifestyle images in a few clicks

• Uses your brand’s style (“Business DNA”) to keep visuals on-brand

• Includes background edits, style references, and prompt-based refinements

• Completely free (for now) in the US, Canada, Australia, and New Zealand

Product photography was a paid bottleneck for small businesses, and a huge startup opportunity. Google just made it instant, high-quality, and free. It’s a warning shot for anyone building tools around product shoots and marketing creatives.

AI RESEARCH

🧼 Google DeepMind Fixes Dirtiest Problems in Image Gen

Diffusion models rely on latent encoders, but training those encoders has always been guesswork. DeepMind’s new approach trains the encoder and diffusion model together, replacing fragile heuristics with a principled, controllable setup.

• Links encoder noise directly to the diffusion model’s minimum noise level

• Replaces the hand-tuned KL penalty with a simple, interpretable weighted MSE loss

• Gives a clear upper bound on how much information the latent space actually carries

• Achieves SOTA results with less compute (FID 1.4 on ImageNet-512, FVD 1.3 on Kinetics-600)

One of the most “just trust it” parts of generative AI just became measurable and tunable. If this holds, many models built on frozen Stable Diffusion latents may be optimizing everything, except the foundation.

AI NEWS

🧭 AI adoption is slower than expected

Tech leaders like Sam Altman and Jensen Huang admit AI adoption is slower than expected, with fear and distrust outweighing hype. Surveys show people worry more about AI’s risks than its benefits, and many don’t see real value in their daily lives yet.

• 80% of firms say AI hasn’t improved productivity or jobs so far

• Most users don’t want to pay extra for AI features on devices

• Critics and “doomer narratives” are shaping public opinion

• Strong demand for AI regulation, even if it slows progress

Unlike past tech booms, AI feels imposed rather than empowering. Unless companies prove clear, everyday benefits, the industry risks losing public trust, and its “social license” to scale.

HOW TO AI

🗂️ How to 100x Your Claude Code Power with Secret Anthropic Prompting Tips

In this tutorial, you will learn the top strategies from Anthropic’s secret prompting guide to make Claude Code, the world's most powerful coding AI agent, work faster, smarter, and more independently.

🧰 Who is This For

• Vibe Coders who want to build entire apps without manual debugging.

• Software Engineers looking to reduce hallucinations in complex codebases.

• AI Researchers (like those at Rundown) building agentic workflows and local tools.

• Anyone wanting to save money on tokens while getting higher-quality code.

Prime Your Session for Maximum Context

Never start a coding session by just asking Claude to "fix a bug." To get 100x results, you must load the "mental map" of your app into Claude's memory immediately.

Start every new chat with an architecture master prompt like this:

"Take a look at the app and architecture. Understand deeply how it works inside and out. Ask me any questions if there are things you don't understand. This will be the basis for the rest of our conversation."

This ensures Claude doesn't guess how your files connect; it creates a grounded foundation for every line of code it writes next.

Set Up Your "Agent Rules" (claude.md)

Claude Code is extremely eager. You can control its "personality" and efficiency by creating a claude.md file in your project's root directory.

Copy and paste these three "Power Rules" into your claude.md file:

• For Accuracy: "Investigate before answering: Review all relevant code before making any changes to reduce hallucinations."

• For Speed: "Use parallel tool calls: Multitask by spinning up sub-agents to do research and write code simultaneously."

• For Clarity: "After completing a task, provide a quick summary of the work done."

Master the Motivation Prompting Technique

Claude Code is proactive. If you tell it why you are doing something, it will anticipate your next five moves.

• The Pro Move: Don't just say "Build a calendar."

• The Better Way: Say, "Build a calendar tool. The motivation is that all notes and tasks we build later should be viewable here." 

As a result, Claude will automatically write the calendar code to be compatible with your future task features without you even having to ask.

Optimize Tokens and Vision

You can save money and fix bugs faster by carefully watching your words and utilizing image drops.

• Watch Your Words: Avoid the keyword "Think" (e.g., "Think about this"). This triggers Claude to use massive amounts of tokens. Instead, use words like "Consider" or "Evaluate" to keep costs low.

• Use Vision for Bugs: Claude Opus 4.6 has world-class vision. If you see a UI bug, don't describe it. Screenshot the error and paste the image directly into the Claude Code terminal. Claude will "see" the visual misalignment and fix the CSS instantly.

Sam Altman says discussions about AI's energy usage are “unfair”, as it takes “20 years of life and all of the food you eat during that time” to train a human.

Google Antigravity users say their paid Google AI accounts were banned after linking Gemini models via OpenClaw; Peter Steinberger says he may “remove support”

Samsung will add Perplexity to Galaxy AI on the upcoming S26 series; users can launch the Perplexity agent by saying “Hey Plex” or with a physical helper button.

Tencent closed its TiMi Montréal studio after nearly five years, without ever releasing a game, as Chinese giants scale back funding for Western game studios.

🧑‍💻 Gemini 3.1 Pro: Google’s new, cheaper AI for complex and multimodal tasks

📸 Pomelli: Turns product photos into ready-to-use marketing content

🤖 Manus: AI agents you can use on Telegram that remember things over time

🎶 Lyria 3: Google’s AI that makes music inside Gemini

THAT’S IT FOR TODAY

Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!

See you tomorrow :)

- Dr. Alvaro Cintas