🗞️ Opening Line
Happy Monday! Amazon just scrapped its entire Nova AI model family and is restarting from scratch under new leadership. Anthropic launched Claude Sonnet 5 at its lowest price ever. And Nvidia built a major AI security alliance — without inviting OpenAI or Anthropic. Let's get into it.
💡 The Big Update
Amazon Scrapped Its AI Models and Is Starting Over. That's a Bigger Deal Than It Sounds.
Amazon is scrapping its Nova family of AI models and restarting its foundation-model effort under new leadership — a striking admission that its internal AI development has not kept pace with competitors.
Let that sink in. Amazon — the company behind AWS, the world's largest cloud platform, with more compute infrastructure than almost any organization on the planet — just admitted that its AI models aren't good enough. Not "we're iterating." Not "we're pivoting." Starting over.
This matters for a few reasons.
First, it's a rare moment of corporate honesty in an industry that almost never admits failure. Every other major lab has been releasing models and claiming progress even when the benchmarks told a more complicated story. Amazon looked at what it had, compared it to Claude, GPT-5.6, and Gemini, and concluded the gap was too large to close incrementally. That's a significant judgment.
Second, it reshapes the competitive landscape in ways that aren't immediately obvious. AWS is the cloud platform that Anthropic runs on — the two companies have a deep infrastructure partnership. Amazon's decision to restart its foundation model effort could eventually mean it's building models to compete with the same company it's hosting. That tension will be worth watching.
Third, it tells you something about how hard this problem actually is. Amazon has billions of dollars, world-class engineers, and more data than almost anyone. If they couldn't build a competitive frontier model on the first attempt, the barrier to entry in this market is genuinely high. The companies that got there first — Anthropic, OpenAI, Google — have an advantage that's harder to close than it looks.
For businesses evaluating AI tools: this is a good reminder to focus on what the models can actually do today, not on the corporate narratives around them. Amazon's cloud is still excellent. Its models weren't. Those are two separate things.
⚡ Quick Commits
Claude Sonnet 5 Just Launched at Its Lowest Price Ever
Claude Sonnet 5 brings stronger long-run coding, tool use, and debugging at a lower price — $2 per million input tokens and $10 per million output tokens until August 31, 2026, then moves to $3/$15 after. Sonnet has consistently been Anthropic's most useful everyday model — strong enough for complex tasks, fast enough for production use. The intro pricing through August makes this the best time to evaluate it for any workflow you've been running on more expensive models. If you're currently paying $15+ per million output tokens for comparable quality, the math is worth checking. AIApps
Nvidia Built an AI Security Alliance. OpenAI and Anthropic Weren't Invited.
Nvidia launched the Open Secure AI Alliance on July 27, 2026 — an industry coalition to build open-source AI cybersecurity tools, with 30+ founding members including Microsoft, IBM, SpaceX, Adobe, Cloudflare, CrowdStrike, Dell, Hugging Face, Red Hat, Salesforce, and the Linux Foundation. OpenAI and Anthropic were notably absent. The membership list is the story. Infrastructure providers, security companies, chipmakers, and open-source organizations all signed on. The two most prominent AI safety companies did not. Whether that's a snub, a strategic choice, or a philosophical disagreement about open-source security tooling, the absence is pointed. An alliance focused on securing AI systems that doesn't include the makers of the most widely used AI systems raises questions worth asking. BuildFastWithAI
🛠️ The Tool Drop
Claude Sonnet 5
Since it's both the news and the tool this week — if you're not already using Claude Sonnet 5, now is the time to try it. The intro pricing of $2/$10 per million tokens through August 31 makes it the most cost-effective frontier-level model available right now. For everyday business tasks — writing, analysis, research, coding, summarization — it's Anthropic's most capable everyday model yet. Available at claude.ai on all plans, and via API for developers. The price goes up September 1 so August is the window to evaluate it seriously. claude.ai
🎯 The So What
Amazon starting over. Claude Sonnet 5 at its lowest price ever. Nvidia building a security alliance that excludes OpenAI and Anthropic.
Three stories that together say: the AI market is maturing fast and the fault lines are becoming visible. Who can actually build frontier models. Who controls the security infrastructure. Who gets invited to the table when the industry self-organizes.
Amazon's restart is the most honest signal yet that building competitive AI is genuinely hard — and that the companies who got there first have a real advantage. The price war happening simultaneously means that advantage is being competed away at the application layer even as it holds at the model layer.
For your business the window to build AI into how you work keeps narrowing. The tools are cheaper than they've ever been. The organizations figuring out how to use them are compounding an advantage every week. Amazon just reminded us that even unlimited resources don't guarantee you get there.
See you next Monday.
— The Changelog
Not subscribed yet? Get The Changelog free every Monday.
Subscribe here
Know someone who'd find this useful? Forward it along.