The Weekly Waypoint, Issue #10
Something unprecedented happened this week. OpenAI, Anthropic, Adobe, Google, and Microsoft all released major agent-focused products within a 72-hour window. This wasn't coordination, it was convergence. The entire industry decided, independently, that agents are the next platform. And they're probably right.
If you're still thinking about AI as a chatbot that answers questions, you're one platform shift behind. The agent era means AI that does things, autonomously, across your tools, on your behalf. This week made that real.
This Week's Big Story: The Great Agent Landing
Here's what dropped, in rapid fire:
OpenAI updated its Agents SDK (April 15), Developers can now build agents that inspect files, run commands, edit code, and handle long-horizon tasks inside controlled sandbox environments. This is the infrastructure layer. If you want to build an AI agent that actually does work in a codebase, this is your starting point.
Anthropic released Claude Opus 4.7 (April 16), The flagship model got a major upgrade focused on advanced software engineering and AI agent tasks. Users report being able to hand off their hardest coding problems to it. It also has a 1M token context window. This is the brain layer, agents need a model smart enough to reason through complex, multi-step problems. Opus 4.7 is that model right now.
Adobe launched Firefly AI Assistant (April 15), This is the application layer. A conversational agent that orchestrates tasks across Photoshop, Premiere, Lightroom, Illustrator, Express, and Frame.io. You don't find the tool and click the button. You describe what you want and the agent does it. "Remove the background, adjust the color balance, and export for social" becomes one prompt instead of twelve clicks.
Google released Gemini Robotics ER 1.6 (April 14) and Gemini for Mac (April 15), Physical agents and desktop agents in the same week. ER 1.6 improves embodied reasoning for robots, the AI that helps machines understand and interact with the physical world. Gemini for Mac puts a native AI assistant on your desktop with a keyboard shortcut. Two very different agent visions from the same company.
Microsoft launched MAI-Image-2-Efficient (April 14), Production-grade image generation at 41% lower cost and 22% faster speed. Not an agent itself, but the kind of infrastructure that makes agent-powered creative workflows viable at scale. Agents need fast, cheap image generation to be useful in production.
Why This Matters: The Stack Is Complete
Look at what just happened across the stack:
- Infrastructure: OpenAI Agents SDK with sandboxes
- Intelligence: Claude Opus 4.7 for deep reasoning
- Application: Adobe Firefly AI Assistant for creative work
- Interface: Gemini for Mac on your desktop
- Physical: Gemini Robotics ER 1.6 in the real world
- Production: MAI-Image-2-Efficient for scale
Every layer of the agent stack got a major upgrade in one week. That's not a coincidence. The industry has been building toward this for two years. The chatbot phase was the demo. The agent phase is the product.
What this means for you: if you're building anything with AI, the question isn't "should I use agents?" anymore. It's "how fast can I adapt what I'm building to be agent-native?"
The Human Cost: Snap Lays Off 1,000
The same week agents arrived, Snap laid off 16% of its workforce, roughly 1,000 employees. CEO Evan Spiegel said the company needs "a new way of working" built around AI.
This is the uncomfortable parallel. The technology that makes agents powerful also makes some roles obsolete. Not all roles. Not most roles. But enough that a company like Snap, not an AI company, a social media company, is restructuring around AI-first operations.
The lesson isn't "AI will replace you." It's "companies that figure out how to use AI agents effectively will outperform those that don't, and that pressure changes headcount." If your company isn't thinking about how agents change workflows, you should be.
The Money: Anthropic at $800 Billion
Anthropic has received investor offers valuing it at $800 billion, more than double its $380 billion valuation from just two months ago. Revenue is reportedly on track for $30 billion.
Accel also raised a $5 billion AI fund, citing returns from Anthropic and Cursor. The money is flowing toward agent infrastructure and AI-native companies at a pace that makes 2023 look cautious.
What this means: the smartest investors in the world are betting that AI agents are the next major platform shift. They're putting real capital behind it. Follow the money if you want to know where this is going.
Practical: Three Things You Should Do This Week
1. Try the OpenAI Agents SDK. Even if you're not a developer, walk through the quickstart. Understanding how agents are built, the tools, the prompts, the guardrails, will change how you think about what AI can do for you.
2. Test Claude Opus 4.7 on your hardest problem. Take the coding task you've been putting off because it's too complex. Give it to Opus 4.7. The results might surprise you.
3. Download Gemini for Mac. It's free, it's native, and it's the first real desktop AI assistant. Use it for a week. Notice what you reach for it for vs. what you still do manually.
The Fun Stuff
- The "agent" naming problem, Every company is calling their product an "agent" now. Microsoft Copilot has "autonomous agents." Adobe has a "creative agent." OpenAI has the "Agents SDK." Your real estate agent is confused.
- Someone asked Claude Opus 4.7 to plan their entire week, It did. Meetings, meal prep, workout schedule, bill payments. Then they realized they'd outsourced their life decisions to a language model. Philosophical crisis ensued.
- Gemini for Mac's keyboard shortcut, The shortcut is apparently customizable but defaults to something that conflicts with half the apps on your computer. Classic Google UX.
What I'm Watching Next Week
OpenAI's GPT-5.4-Cyber launched this week for vetted security professionals, the first model specifically trained for offensive and defensive cybersecurity. Expect the security industry to have strong opinions. Also watching for more companies announcing AI-driven restructurings after Snap's move. The dam may be breaking on tech layoffs justified by AI efficiency.
Deep Dive This Week
Pro members go deeper: "The Agentic AI Revolution: Building Systems That Think, Decide, and Act", the complete guide to understanding and building with AI agents. We cover the architecture, the platforms, the security implications, the business models, and the practical steps to go from chatbot user to agent builder. The newsletter gave you the headline. The deep dive gives you the playbook.
The chatbot era taught us to ask better questions. The agent era will teach us to delegate better tasks. Start practicing.
- James
P.S., Have you built anything with AI agents yet? Reply and tell me what you've tried. I want to feature real examples in future issues.