Tipping the AI Scales

Hello, and happy Sunday! Was this newsletter forwarded to you? Sign up to get it in your inbox.

Release notes

OpenAI throws its weight around

This week, OpenAI slashed costs on o3 by 80 percent after announcing o3-Pro—making premium-level performance dramatically more accessible. Anthropic, meanwhile, opted to keep its per-token-pricing the same across its newest Claude 4 models and previous models. OpenAI is clearly leveraging its funding advantage—it’s sitting on $40 billion versus Anthropic’s $3.5 billion war chest, meaning that OpenAI has 10 times more capital to throw at winning.

But it’s an indicator of a larger strategy that’s about way more than just token pricing. Late last month, the company announced "Stargate UAE," which would see the construction of a massive AI computing cluster in Abu Dhabi and free access to ChatGPT Plus for all of the United Arab Emirates’ nearly 10.5 million residents. In the United States, OpenAI struck a deal with California State University (CSU) in February to give ChatGPT to 460,000 students across 23 campuses—effectively delivering a small city’s worth of users to the platform, and making CSU the first AI-powered university system in the country.

For OpenAI’s customers—in this case, a nation-state and a sprawling institution of higher learning—it’s a way to future-proof themselves, and remain competitive in a world where AI is already changing how students learn and how researchers work. For OpenAI, it's a strategic land grab: Capture the next generation of knowledge workers early, and solidify its already dominant position in the public imagination. OpenAI’s rivals are surely taking note. The time to get big and find out how to be indispensable is now, because AI is rapidly evolving from being a consumer product into what looks more like a public utility. And that’s only something that can be delivered at scale.

Meta makes an enormous bet on data

And then there’s Meta’s reported $15 billion investment in Scale AI for a 49 percent stake in the company. AI models need data to train on, and Scale AI is one of the largest providers of that data. OpenAI, Google, and Microsoft all depend on Scale’s high-quality data (which is labelled by humans) to train their models. Meta’s investment in Scale shows that even with AI, there’s still no substitute for high-quality human labeling.

You can see the synergy: Scale AI brings the infrastructure for high-quality, human-labeled training data; Meta brings user behavior data from over 4 billion people, which could be used to help refine and personalize AI systems at scale. If Meta can learn how Scale produces effective training data with human labelers, it could apply those insights across its massive user base—building models that are more attuned to human behavior, which is exactly what a company rooted in social networks and communication needs.

When it comes to building AI that is human-centric, and benefits the most people possible, though, is this the kind of scale that’s needed? That’s a harder question to answer.—Alex Duffy

Knowledge base

🎧 🖥 "How Two Engineers Ship Like a Team of 15 With AI Agents" by Rhea Purohit/AI & I: Cora engineers Kieran Klaassen and Nityesh Agarwal shipped six features, five bug fixes, and three infrastructure updates in a single week using AI agents. Their secret? A “prompt that writes prompts” workflow. Read this if you want to 10x your engineering capabilities without hiring a single person. 🎧 🖥 Watch on X or YouTube, or listen on Spotify or Apple Podcasts.

"How to Keep Your Writing Weird in the Age of AI" by Katie Parrott/Working Overtime: AI is the ultimate grammar nerd—always trying to smooth your quirky sentences into professional blandness. Katie Parrott discovered this while falling down a biblical rabbit hole. Read this if you want practical strategies for collaborating with AI while preserving your unique voice.

"ChatGPT Taught Me Special Relativity—And Changed How I Learn"by Nityesh Agarwal/Source Code: Ever wished you had a genius tutor who'd patiently explain mind-bending concepts at 2 a.m. without judging your dumb questions? Us too. Nityesh spent two weekends firing over 100 questions at ChatGPT until he genuinely understood Einstein's special relativity—despite having no physics background. Read this if you want to unlock ChatGPT's hidden potential as your personal tutor.

"How to Figure Out What People Want" by Dan Shipper/Chain of Thought: “Make something people want” is the startup world's equivalent of telling an investor to “beat the market.” Thanks for nothing! Read this column from Dan Shipper if you want to stop asking strangers what they want and start being the kind of person who can generate desires in others.

"Apple’s AI Vision: Quiet Upgrades, Not Quantum Leaps” by the Every team: At WWDC 2025, Apple showed off its vision for AI integration—and it's remarkably restrained. Instead of building the biggest, most powerful models, it’s cleverly weaving capable (if unsexy) AI into its ecosystem, deepening ChatGPT integration, and bringing back skeuomorphism with a 2025 twist. Read this if you want to understand whether Apple's AI strategy will help it win the long game or leave it in the dust.

Join Every

We're hiring a social media strategist. Be the voice of Every on social media. You'll maintain and evolve Every's established voice on X and LinkedIn, building on our existing playbook while experimenting with new formats to grow our audience and engagement. You're fluent with AI tools, using them to scale content creation and streamline workflows. You're deeply attuned to the online tech conversation, actively participating in discussions about AI and technology trends, with a keen sense for both marketing a subscription bundle and extracting the most compelling insights from our stories. You’re a master of metrics, with the ability to identify and double down on what’s working. The ideal candidate has excellent writing skills, fluency in social analytics, and familiarity with tools (including, but not limited to, Figma, Descript, and Buffer). You should move fast, be scrappy, and—most of all—love Every. If you're interested, email Kate Lee at [email protected] with your LinkedIn profile, examples of tweets (written for either a personal and/or corporate account), and why you'd like to join us.

Alex Duffy is the head of AI training at Every Consulting and a staff writer. You can follow him on X at @alxai_ and on LinkedIn, and Every on X at @every and on LinkedIn.

We build AI tools for readers like you. Automate repeat writing with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.