TechLook
Every major tech company reported this week. Every single one beat expectations. That doesn't happen often — and when it does, it's worth asking what it actually means for the people building on top of these platforms.
The short answer: the AI revenue machine is no longer theoretical. It's showing up in the numbers.
Big Tech Just Confirmed the AI Economy Is Real
The five largest technology companies reported earnings this week, and the beat was near-total. Alphabet posted $109 billion in revenue — up 22% year-over-year, its fastest growth since 2022. Google Cloud hit $20 billion with a $460 billion backlog, growing 63%. Amazon posted $181 billion with AWS accelerating to 28% growth — its fastest in 15 quarters. Amazon's own AI chip business is now a $20 billion run-rate product. Meta posted $56 billion, up 33%, and raised its 2026 capex guidance to $145 billion. Microsoft crossed 20 million paid Copilot seats and hit $37 billion in annualized AI revenue, up 123% year-over-year.
Here's everything you need to know:
- Alphabet: $109B revenue (+22% YoY), Google Cloud at $20B with $460B backlog, growing 63%
- Amazon: $181B revenue, AWS up 28% to $37.6B, AI chip business at $20B run-rate
- Meta: $56B revenue (+33%), raised 2026 capex to $145B citing data center component pricing
- Microsoft: $82B revenue (+18%), 20M Copilot seats, $37B annualized AI revenue
- Amazon's Q1 capex hit $44B — up from $25B a year ago
- Alphabet now has 350 million paid subscriptions across its services
Amazon's free cash flow tells the other side of the story. TTM free cash flow fell to $1.2 billion from $25.9 billion a year earlier, driven by $59.3 billion in property and equipment purchases. The AI demand is real. So is the infrastructure bill.
The implication for builders is straightforward: enterprises are committing real budget to AI-powered products and infrastructure. If you're selling into this market — whether as a SaaS, a developer tool, or an agency — the支票 is open. But the companies writing those checks are the ones who own the foundational layers, which means the pressure on margins and differentiation will keep building.
What happens if AI demand cools before the capex cycle matures? That's the question nobody in earnings calls is being forced to answer yet.
Claude's New Connectors: AI Just Joined Your Creative Stack
Anthropic released nine connectors on the Model Context Protocol this week, integrating Claude directly into professional creative software. The list is broad: Adobe Creative Cloud (50+ tools across Photoshop, Premiere, Illustrator, Lightroom), Blender, Autodesk Fusion, SketchUp, Ableton Live + Push, Splice, Affinity, and Resolume. Claude can now read live project context inside these apps and execute tasks on behalf of the user — eliminating the copy-paste window between assistant and professional workflow.
This is the widest single-day creative integration for any AI assistant to date, spanning design, 3D, video, music, and live performance in one launch.
Here's everything you need to know:
- Connectors built on the Model Context Protocol — same standard Anthropic has been pushing for industry adoption
- Adobe integration covers 50+ tools including Photoshop, Premiere, Illustrator, Lightroom, Express
- Blender, Autodesk Fusion, and SketchUp get natural-language 3D modeling interfaces
- Ableton Live integration is grounded in official documentation for music production
- Splice lets Claude search royalty-free sample catalogs natively
- Resolume enables real-time natural-language control for VJ performances
The Model Context Protocol is becoming the de facto integration layer. If you're building AI-native products that need to fit into professional workflows — design, video, music, engineering — understanding MCP and building for it isn't optional. It's table stakes.
The broader signal: AI is graduating from chat interfaces into the tools professionals already use. The user doesn't go to the AI. The AI comes to the user's work.
Prompt Engineering Is Dead. Memory Engineering Is What Matters Now.
OpenAI co-founder Greg Brockman said it plainly this week: the skill worth having is no longer writing better prompts. It's engineering what the model carries forward.
The shift is being driven by the new generation of "intuition" models — GPT-5.5 and Claude Opus 4.6 can now take high-level goals, read context, use browsers, build spreadsheets, and complete messy workflows without step-by-step instruction. The model figures out the steps. You just point it at outcomes.
Tools like OpenClaw are already shipping memory and context file injection before every generation — engineering the model's recall rather than crafting its instructions.
Here's everything you need to know:
- Greg Brockman says GPT-5.5 can complete messy workflows from high-level goals without step-by-step prompts
- "Intuition" models (GPT-5.5, Opus 4.6) signal that prompt engineering is losing relevance as a moat
- OpenClaw already injects memory and context files before every generation
- The new competitive advantage: what you teach the model to remember, not what you tell it to do
- Cursor and Airbnb face U.S. House probes over Chinese AI use in their tools — an early signal that agent provenance is becoming a regulatory concern
The practical implication: if you're hiring for AI productivity or building AI-native products, stop testing people's prompt skills. Test their ability to build systems that manage context, maintain memory, and handle long-running agentic workflows. That's the real craft now.
Oracle Ships Agent Skills — Enterprise Software Finally Gets Agent-Ready
Oracle's NetSuite launched SuiteCloud Agent Skills this week: official instruction packs that teach AI coding agents how to customize NetSuite safely and correctly. This is the move enterprise software vendors have been avoiding — until now.
The pattern is clear. The foundational models got good enough that agents can do real implementation work. The question was whether enterprise software vendors would help them or block them. Oracle chose help.
Here's everything you need to know:
- SuiteCloud Agent Skills are official instruction packs for AI coding agents working with NetSuite
- Agents learn how to customize NetSuite safely and correctly from vendor-provided playbooks
- Greg Isenberg notes Salesforce's Headless API signals the same shift: SaaS is becoming API infrastructure, AI agents are the product layer
- Vertical operators can now build agents on top of Salesforce, HubSpot, Workday — replacing consultants, finishing workflows, charging by outcomes
- The implication: the next wave of SaaS isn't a prettier dashboard. It's an agent that replaces the implementation partner.
For founders and builders, this is a green light. If you're targeting enterprise workflows — finance, HR, operations, supply chain — the integration path just got smoother. Build your agent to understand these systems, and you can compete directly against the implementation consultants who have been harvesting six-figure fees for years.
⚡ Quick Hits
- Google: Gemini can now create Docs, Sheets, Slides, Word docs, Excel, PDF, CSV, LaTeX, and Markdown directly in chat — file generation is now native.
- Apple: iOS 27 will add Extend (expands photo backgrounds with generative AI), Enhance (auto lighting fixes), and Reframe (perspective shifting for spatial photos). WWDC June 8.
- Parag Agrawal: The former Twitter CEO's startup Parallel Web Systems raised $100M at a $2B valuation — nearly tripling from $740M five months ago. The company builds search and research APIs for AI agents.
- Meta: Opened its ad ecosystem to third-party AI tools including ChatGPT and Claude. Advertisers can now connect AI assistants directly to Meta ad accounts for faster campaign setup and reporting.
- Google: Signed a Pentagon AI contract, joining OpenAI and xAI in supplying models for government and classified work.
Techlook — AI & tech signal for founders and builders.