Claude is now getting good at Robotics

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

Today we have:

🤖 Claude Is Getting Surprisingly Good At Robotics

🗣️How to Design a Custom AI Voice With Rumik Silk Mulberry

🎨 Adobe Wants AI To Remember Your Creative Work

🔨Tools and Shifts you cannot miss

🤖 Claude Is Getting Surprisingly Good At Robotics

Anthropic's latest Project Fetch experiment tested whether Claude Opus 4.7 could independently complete robotics tasks that previously required teams of humans. The results suggest AI is rapidly moving from assisting robot operators to directly controlling physical systems.

The Shift:

Humans Couldn't Keep Up - Tasks that took the Claude-assisted human team 181 minutes were completed by Opus 4.7 in just 9 minutes and 35 seconds. Across all shared tasks, the model was nearly 19 times faster.

Less Code, Better Results - Despite outperforming both human teams, Opus 4.7 generated only 1,045 lines of code compared to over 10,000 lines produced by the original Claude-assisted team during the first experiment.

Robotics Without Training - Anthropic says these gains weren't the result of robotics-specific development. The improvements emerged from general model scaling, similar to how AI unexpectedly became stronger at coding and cybersecurity tasks.

Physical Precision Remains Hard - Claude successfully connected sensors, controlled the robot, and detected objects, but struggled with the final challenge of precisely moving and retrieving a beach ball through continuous physical feedback.

Project Fetch highlights a familiar AI pattern: first models help humans, then humans help models, and eventually models begin handling tasks on their own. While robotics remains far from solved, Anthropic believes AI is entering the early stages of physical agency, where models can increasingly operate real-world tools and machines with limited human involvement.

Together with MarketBeat

7 Stocks to Buy Before the Robots Take Over

The next AI trade may not be another chatbot.

It may be surgical robots, automated warehouses, smart factories, and machine vision systems already reshaping how companies operate.

MarketBeat’s new 7 Stocks to Buy Before the Robotics Revolution report reveals seven companies positioned across the automation boom, from robot builders and AI chip leaders to machine vision providers and factory automation giants.

This is where AI gets a body.

And as labor shortages, wage pressure, and supply chain stress push more companies toward automation, these stocks could move before the robotics story becomes impossible to ignore.

The report normally sells for $29.97, but it is free for a limited time.

Send My Free Report

🗣️How to Design a Custom AI Voice With Rumik Silk Mulberry

Most TTS tools give you a dropdown. Rumik lets you describe the voice you want and builds it. Here's how.

Step 1: Open the Rumik Playground - Head to playground.rumik.ai. No complex setup, log in and you're in.

Step 2: Describe your voice - Type a plain-language description of what you want. Age, gender presentation, accent, tone, energy level, speaking style. The more specific, the closer the output lands.

Step 3: Generate and listen - Silk Mulberry 1.5 builds the voice to match your description. It handles pauses, whispers, laughs, and emotional range, not just neutral speech delivery.

Step 4: Iterate on the description - If the output isn't right, adjust one variable at a time, make it warmer, younger, more energetic, and regenerate. You're designing, not selecting.

Step 5: Deploy via API - Once you have the voice you want, use it in production via API for real-time voice applications, companions, or any audio output you're building.

On instruction-following benchmarks, it hits 74% overall consistency, strongest on descriptive style prompts at 82.7%, making it most powerful when your brief is specific and detailed. You can try the Rumik Playground here.

🎨 Adobe Wants AI To Remember Your Creative Work

Adobe is pushing AI beyond one-off prompts and into long-term creative workflows. The company unveiled a redesigned Firefly experience alongside specialized AI assistants across its biggest creative applications.

The Shift:

Reusable Creative Assets - Firefly's new Elements feature lets creators save characters, locations, and objects for future projects, making it easier to maintain visual consistency without repeatedly rewriting detailed prompts.

Context Stays Intact - The new Projects workspace keeps generations, assets, references, and creative history together, allowing users to pause and resume work without losing context or organization.

Specialists For Every App - Adobe is rolling out dedicated AI assistants across Photoshop, Premiere, Illustrator, InDesign, and Frame.io, with each assistant trained to handle the unique tasks and workflows of its application.

From Editing To Automation - Premiere's assistant can organize footage, rename clips, identify spoken topics, and build rough timelines, while Photoshop's version can perform editing tasks through simple natural-language instructions.

Adobe's focus is shifting from generating content to managing entire creative workflows. Instead of starting from scratch every session, creators can now build persistent projects, reuse assets, and delegate repetitive production work to AI assistants.

Together with Wispr Flow

Your best prompts are the ones you'd never bother typing.

The detailed ones. The ones with examples and edge cases. Wispr Flow lets you speak them instead — clean, structured, ready to paste into any AI tool. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

🔨AI Tools for the Shift

📧 Mara – Writes welcome emails, win-back campaigns, and lifecycle messages in your founder voice while learning from every send.

🌐 OpenToon – Translates comic and manga chapters from a simple URL, helping readers access content without waiting for official translations.

🎬 Seavid AI – Converts images and ideas into engaging AI-generated videos with minimal effort.

🌍 GeoAxis – Identifies where a photo was taken using visual clues alone, without relying on GPS, EXIF, or metadata.

🎵 Freesong – Generates original music from prompts, making it easy to create songs without production experience.

🚀Quick Shifts

🎵 The Atlantic launched a searchable database of music used in AI training, exposing millions of tracks linked to major datasets and giving artists a way to check whether their work may have been used to train AI models. 

🧬 Nobel Prize-winning AI researcher John Jumper is leaving Google DeepMind for Anthropic. Jumper co-created AlphaFold, the protein-structure prediction system that earned him and Demis Hassabis the 2024 Nobel Prize in Chemistry. 

🎓 Norway is restricting AI use in schools, banning it for most students aged 6–13, allowing limited supervised use for ages 14–16, and focusing on responsible AI literacy for older students. 

🧠 AWS is exploring sales of its Trainium AI chips to external data centers, signaling ambitions beyond internal use. Amazon believes Trainium could support a $50B annual business across AWS and third-party customers. 

🧩 Prompt of the Day

Rubber Duck Debugging That Helps You Find the Bug Yourself

Most debugging gets messy because you jump straight to fixing before fully understanding the logic. Rubber duck debugging slows the process down so the mistake becomes visible.

Turn confusion into a guided debugging conversation.

Paste the prompt: Drop this into ChatGPT, then paste your code and issue details.

Prompt to paste

Act as my rubber duck debugging partner for [Insert programming language]. My code should [Insert what the code should do], but instead it [Insert what actually happens]. Review my pasted code by asking me step-by-step questions about what each section does, what output I expect, and where the behavior changes. Help me trace the logic, identify likely failure points, and suggest the smallest safe fix. Do not jump straight to rewriting everything unless necessary.

🤳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 👇