Claude will develop the next vaccine

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

Today we have:

🧬 Claude takes on drug discovery

💪How to build more without burning credits using Replit Free Mode

🎓 Google turns Gemini into an AI study hub

🔨Tools and Shifts you cannot miss

🧬 Claude takes on drug discovery

Anthropic says Claude can now run substantial parts of protein-design research with limited human involvement. In testing, its AI-designed molecules worked across 14 of 15 biological targets, with reported success rates above typical industry levels.

The Shift:

1. AI Designs the Proteins - Anthropic gave Mythos Preview and Opus 4.8 one expert prompt, internet access, and research tools. The models then largely ran the protein-design campaigns themselves.

2. 14 of 15 Targets Worked - Claude-designed candidates produced working molecules across 14 of 15 targets. Anthropic reported binding success rates of 22% to 35%, compared with its cited industry norm of 10% to 15%.

3. Humans Still Ran the Lab - Claude did not physically create or test the molecules. Twist Bioscience and Adaptyv Bio manufactured the candidates and performed the laboratory measurements.

4. Four Days to 19 Minutes - Separately, Opus 5 analyzed raw instrument files without specialized lab software and measured a sample at 96.4% purity in 19 minutes. The lab's own report took four days.

Claude isn't replacing the physical laboratory, but Anthropic's results suggest AI can increasingly handle the reasoning between experiments, from designing candidates to interpreting complex lab data.

Together with Guardian Financial Publishing

SpaceX Is Overvalued. Here's the Smarter Play.

At $1.75 trillion, SpaceX is now worth more than Coca-Cola, Disney, and Netflix combined — a company with no public earnings track record and a $135 price tag set by the people selling it to you.

Do the math. To justify that valuation, SpaceX would have to grow into one of the largest companies on Earth before you ever see a profit. History is brutal on day-one IPO buyers who pay that kind of premium. Most are underwater within a year.

There's a smarter way to ride the wave. Our analyst found 3 stocks positioned to ride the SpaceX boom — without the bubble pricing. You get the tickers, entry guidance, price targets, a bonus 4th pick, and a 3-phase buy/sell playbook.

The window before Wall Street rotates in is closing.

Get the Free Report

💪How to build more without burning credits using Replit Free Mode

Replit just launched Free Mode, powered by OpenAI's GPT-5.6 Luna, so everyday tasks no longer use your credits. It is the default for all Core ($20/month) and Pro ($100/month) subscribers as of August 19.

Step 1: Open Replit, Free Mode is already on - Free Mode is now the default setting for Core and Pro users. No setup needed. Every chat, ideation session, and simple task runs on Luna automatically without touching your credit balance.

Step 2: Describe your app in plain language - Type what you want built. "Build me a full stack web app with a database, user auth, and a clean homepage." Luna handles chats, ideation, and straightforward builds entirely within Free Mode.

Step 3: Let Replit route heavy tasks automatically - When a task requires a more advanced model, Replit automatically routes it to GPT-5.6 Sol, then returns to Free Mode once the task is completed. No manual switching, no surprise charges.

Step 4: Iterate without watching your credits - Describe changes in plain language. Add pages, connect a database, adjust the UI. Anything Luna can handle runs free. You only spend credits when the task genuinely needs frontier-level reasoning.

Step 5: Deploy and share - Hit Deploy. Replit hosts the app and manages the infrastructure. Connect a custom domain from the Deployments tab and go live.

Core subscribers can now create 30x more than before on Replit using Free Mode, including up to 30 hours per month of chat, for just $20 per month. You can try Replit Free Mode here.

🎓 Google turns Gemini into an AI study hub

Google is making Gemini and Search much more focused on education. Its latest tools cover the entire study workflow, from understanding difficult concepts to researching topics, creating revision materials, and preparing for exams.

The Shift:

1. Gemini Gets a Student Hub - A new dedicated hub combines study notebooks, flashcards, and practice quizzes in one place. Gemini can also identify deadlines from a syllabus and add them to Google Calendar.

2. Search Becomes an AI Tutor - Google Search can generate custom quizzes and interactive learning experiences for specific questions. Students can even upload notes, PDFs, and slides to create condensed study documents.

3. Interactive 3D Learning - Gemini can now generate functional 3D simulations, tables, and interactive visuals around a prompt. Students could ask to see DNA in 3D, then rotate and explore the generated structure.

4. Research You Can Talk To - Deep Research is coming to Gemini Live, letting students start complex research and leave it running in the background. Once finished, they can discuss the report and ask follow-up questions by voice.

Google is moving beyond giving students AI-generated answers. Gemini and Search are becoming an integrated learning system that can research, explain, visualize, test, organize, and coach students through their studies.

Together with Wispr Flow

10x the context. Half the time.

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

🔨AI Tools for the Shift

🎨 Hermai Brand API – Automatically white-labels B2B SaaS products with each customer's branding, helping teams deliver personalized experiences without manually rebuilding themes.

🧪 Checksum AI – Acts as a testing companion for AI coding agents, helping catch issues and validate generated code before it ships.

💬 Peach Co-Pilot – Works as an AI sidekick inside WhatsApp to help busy professionals manage everyday tasks and conversations.

🖥️ Berd – Provides a playful desktop environment for building and experimenting with AI agents, with an open-source approach.

🎨 MiniMax Design – Lets you assemble your own team of specialized AI agents for open-ended creative and design projects.

🚀Quick Shifts

💳 Stripe is buying AI gateway OpenRouter. Stripe has agreed to acquire OpenRouter, which gives developers access to hundreds of AI models through one interface, reportedly valuing the startup at $7.5 billion. 

💻 Meta AI is becoming a desktop productivity assistant. Meta’s new Mac app can analyze shared windows, answer questions, create content, and provide system-wide dictation, while new integrations let it analyze Meta accounts, ad campaigns, and Google Workspace data. 

🔗 Claude can now do more inside Gmail and Google Drive. Claude’s Workspace connectors can send emails and manage Drive files, while Cowork is now available across web and mobile for paid users, letting work continue across devices. 

🔐 OpenAI Introduces Privacy-First Safety Monitoring - Private Safety Processing can detect misuse across multiple conversations without retaining customer data or requiring human review, positioning it as an enterprise-friendly alternative to Anthropic’s 30-day retention policy.

📉 AI adoption is growing, but public trust is getting worse. 52% of Americans now say they are more concerned than excited about AI, as job fears, unwanted AI features, data centers, and unfulfilled promises fuel growing backlash. 

💃Prompt of the day

Create a Chaotic Paparazzi-Style Celebrity Photo

Image Prompt: Create a candid, ultra-realistic paparazzi photograph of [INSERT PERSON] unexpectedly spotted at [INSERT LOCATION], captured while they hurriedly try to avoid the camera.

Context

• Subject: [Insert person]

• Location: [Insert location]

• Situation: [Insert what they were caught doing]

• Mood: Chaotic, spontaneous, tabloid-like

• Camera style: Handheld paparazzi photography

Instructions

Show [INSERT PERSON] moving quickly through the scene while glancing toward or away from the camera with a startled, uncomfortable, or caught-off-guard expression. Include [INSERT PROPS OR OBJECTS] that make the situation visually interesting and unexpected. Capture them mid-stride with natural clothing movement, swinging objects, imperfect posture, and believable motion. Use messy handheld framing, slight camera shake, motion blur, harsh direct flash, blown highlights, visible flash glare, digital noise, and inconsistent exposure. Keep the surroundings recognizable but softly blurred by movement and shallow depth of field, with background people or objects appearing incidental rather than staged.

🤳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 👇