Claude's personalities vary across models

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

Today we have:

🧠 Anthropic Measures Claude's Personality

👨‍💼How to Land a Job Through Your Network Without Cold Outreach Using Key AI

🤖 Nadella Challenges AI Labs On Distillation

🔨Tools and Shifts you cannot miss

🧠 Anthropic Measures Claude's Personality

Anthropic published research analyzing how Claude's personality varies across different models and languages, using more than 309,000 real-world conversations to measure consistent behavioral patterns.

The Shift:

Four Personality Dimensions -  Researchers grouped Claude's behavior into four traits: warmth vs. rigor, candor vs. execution, depth vs. brevity, and deference vs. caution, creating measurable personality profiles instead of relying on subjective impressions.

Each Model Behaves Differently - The study found each Claude model has its own conversational style. Sonnet 4.6 generally appeared warmer and more concise, while Opus 4.7 emphasized caution, detailed explanations, and greater transparency.

Language Influences Responses - Claude's personality also shifted depending on the language used. English conversations leaned toward detailed, rigorous responses, while languages like Hindi, Arabic, and Dutch produced noticeably different communication styles.

Researchers Don't Yet Know Why - Anthropic says it does not yet understand why these personality differences emerge across models and languages, suggesting they could stem from training methods, language data, or other factors still being studied.

The research shows that AI behavior extends beyond accuracy alone. As models become more widely used, understanding how they communicate across different contexts is becoming an important area of AI research.

Together with Hightouch

Swap out your stale ads

Not all ads are created equal. And the old, low-performing ones? Those have got to go.

Hightouch can help you refresh your fatiguing ads in just one click. They show you insights on your ad performance so that you can spot the low performers, swap ’em out, and focus your energy on scaling up your high-performing ads.

You can do it all in the Hightouch Ad Studio. Ad Studio’s got the context of your ad performance built right into it to help you:

• generate new variations of your ads instantly

• launch your new ads as soon as you’re ready

• keep your campaigns consistently high-performing

Turn your tired old ads into potentially high-performing, fresh creative with just a click.

Refresh your ad creative today

👨‍💼How to Land a Job Through Your Network Without Cold Outreach Using Key AI

The role you want is one or two hops away. Cold applications have a 2% response rate. Warm paths have 45 to 60%. Here's how to use your network the right way.

Step 1: Join Key AI and let Kai build your profile - Sign up for free at key.ai. Kai pre-builds your professional profile from your community memberships and public signals before you touch a thing.

Step 2: Tell Kai what you're looking for - Type it in plain language. "Who can refer me to a PM role at Stripe?" or "Which fintech companies in my network are hiring senior engineers?" Kai searches your graph and ranks who you should actually talk to.

Step 3: See your warm paths - Kai shows you exactly how you're connected to the right people, mutual community members, shared events, and overlapping backgrounds. No guessing, no stalking LinkedIn.

Step 4: Request a consent-based intro via Handshake- Instead of cold messaging, send a Handshake request. Both sides opt in before any introduction is made. Response rates hit 45 to 60% versus under 5% for cold outreach.

Step 5: Get matched to opportunities from your graph - Kai surfaces relevant roles ranked by how warm your path is to the hiring team, not a generic job board, not sponsored listings. Actual signal from your actual network.

Free to join. 100,000+ professionals already inside. You can get started here.

🤖 Nadella Challenges AI Labs On Distillation

Microsoft CEO Satya Nadella weighed in on the growing debate around AI model distillation, questioning whether frontier AI labs are applying different standards to how models learn.

The Shift:

Nadella Questions Distillation Rules - Nadella argued that AI companies support broad use of public data to train models but oppose competitors learning from their own model outputs through distillation, calling the approach inconsistent.

Microsoft Pushes Enterprise Control - He also said businesses should own their AI infrastructure, institutional knowledge, and data, warning against relying entirely on external model providers for long-term AI capabilities.

Distillation Debate Continues - OpenAI and Anthropic have previously warned that foreign competitors are using large-scale distillation to recreate advanced AI models more cheaply, raising concerns around intellectual property and AI security.

AI Training Faces Fresh Scrutiny - The discussion has renewed attention on how frontier models are trained, with debates continuing over public data usage, model outputs, creator rights, and the balance between open innovation and protecting proprietary AI systems.

The debate around AI distillation is expanding beyond technology into policy and business strategy, as AI companies, enterprises, and governments continue to define the rules for how future models can learn and compete.

Together with Gladly.ai

Why did one company's AI work, and another's didn't?

One had a dedicated owner. Resolution rate: 48.9%. One didn't: 0.38%. See the full breakdown.

See the data

🔨AI Tools for the Shift

🏡 Zory AI – Instantly transforms any room into a beautifully redesigned interior using AI-generated design concepts.

🖼️ Text Behind Image – Creates the popular text-behind-subject effect directly in your browser while keeping your photos private on your device.

📈 Evalyze – Helps founders find the right investors and improve pitch decks with AI-powered analysis and matching.

📒 Jade Note – Turns your notes into persistent AI memory by connecting them directly to ChatGPT or Claude through a built-in MCP server.

🎬 FacelessGenie – Generates complete faceless videos from a single topic, including scripts, voiceovers, visuals, and editing for Shorts and long-form content.

🚀Quick Shifts

🏗️ New York has become the first U.S. state to pause new hyperscale data centers, imposing up to a one-year moratorium on projects over 50MW while it develops rules to address energy costs, environmental impact, and community protections. 

👥 Anthropic continues to strengthen its AI team, with Monzo cofounder Tom Blomfield joining its AI compute division, following recent high-profile hires including Nobel laureate John Jumper and former Tesla AI chief Andrej Karpathy. 

🧠 Microsoft CEO Satya Nadella warned businesses about AI data ownership, arguing that companies “pay twice” by spending on AI while also handing over valuable proprietary knowledge, and urged enterprises to retain control of their data and avoid vendor lock-in.

🧩 Anthropic has extended higher Claude Fable 5 access through July 19, keeping Claude Code’s weekly usage limits 50% higher for paid users while allowing continued Fable 5 use with usage credits after half the weekly quota is reached. 

🇮🇳 Anthropic has begun rolling out India-specific pricing for Claude, introducing rupee-denominated Pro, Max, and Team plans in its second-largest market after the U.S., though support for India's UPI payment system is not yet available. 

🧩 Prompt of the Day

Negotiate Your Job Offer With Confidence

Most candidates accept the first offer because they don't know what to say. A well-structured counteroffer positions your request as reasonable, data-backed, and collaborative rather than confrontational.

Maximize your offer without damaging the relationship.

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

Prompt to paste

Help me negotiate a job offer based on [Insert offer received], my target compensation of [Insert target], and [Insert justification] using [Insert negotiation method]. Write the exact script I should use, including natural wording, pause points, and responses to likely objections. Explain how to present my counteroffer confidently, justify my request with market data or competing offers, and maintain a collaborative tone throughout the conversation. Include a fallback strategy if the employer cannot meet my salary target and suggest other benefits or compensation I should negotiate instead.

🤳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 👇