The whole thing about Anthropic being sued for overselling Claude Max is a (another) wake-up call for the AI industry. If AI companies want long-term credibility, what they sell can't contradict what they deliver. Overpromising is a short game.

"AI-native" is quickly becoming what "mobile-first" was in 2012. The companies rebuilding their workflows from scratch around AI aren't just more efficient, they're structurally different from competitors trying to bolt AI onto existing processes. The gap is only going to get bigger.

Yesterday @DoorDash dropped an in-app AI agent. Andy Fang said it took a ton of iteration cycles to get it working reliably, which is kind of the whole story. Everyone thinks the hard part is the AI model. The hard part is actaully getting it to behave the same way every single time a real person is actually trying to reorder food or fix a wrong delivery. If your brand has people buying the same stuff over and over, a history of what they have ordered, and some kind of returns or refund flow, you have everything an agent like this needs to run.

For brands and agencies, if you can't tie AI spend to an output metric, you're basically in the same spot as Meta's finance team right now. A lot of usage, fuzzy ROI.

$SPCX went public today at a $1.77 trillion valuation. $75 billion raised, the biggest IPO in history. The old record was Saudi Aramco at $29.4 billion. This one nearly tripled it. Two quotes I wrote down in 2018 stand out today. Munger on AI: "There's more hype in the field than probable reality." Buffett on passing on Amazon: "It's a miracle. I don't invest in miracles." Amazon kept compounding. AI became infrastructure. And today a rocket company with AI stitched into it priced the largest IPO ever. On those calls, they were wrong. But Buffett said something else that aged better: he avoids what he doesn't understand. He knew passing on Amazon cost him billions and stayed disciplined anyway. We watch brands make the same trade every week. Skepticism feels like discipline. Sometimes it's just a decision not to learn. You don't have to invest in miracles. You do have to understand them.

We tested Seedance 2.0 in our video creation workflow and got to a fully edited video with captions, cuts, B-roll, synced audio all in 3 generations or fewer. The approach: lock persona and product visuals as references first, then feed them directly into the model. Character drift gone. Product inconsistency gone. Separate tools gone. Video extension was the biggest game changer. Character, environment, and voice stay consistent across the whole video without starting over, something no tool we've tested has done reliably. About 30% faster generation speed. More output, less iteration, and a much more efficient creative process overall.

Walmart's CGO Seth Dallaire made the case that retail media isn't advertising , it's digital merchandising. Walmart Connect is growing ~40% YoY, Walmart members spend more and churn less, and agentic AI is already serving contextual ad placements. The future of shopping is natural language such as "find me a fragrance-free detergent for sensitive skin." Whoever wins the prompt gets the sale.

Tech workers saying AI tools increase their hours and fatigue is the part of the productivity debate nobody on the platform vendor side will quote. It's also the question every CEO should ask their team this week before signing the next AI seat-license renewal.

Nike didn't pay a dollar for 2010 World Cup sponsorship. Adidas had that locked up. Nike ran "Write the Future" anyway. 7.8M views in a week. Overtook Adidas as the world's #1 football brand. They won by out-creating everyone, not outspending. Same lesson applies to the 2026 World Cup.

For ad creative teams, the gap between AI-generated concept and production-ready asset just got a lot smaller. Now you have resizable objects, rewritable text, adjustable layouts.

It's funny how ChatGPT recently said 'chat is dead' but two months ago they were selling a 100 billion dollar ad business inside that chat. One of those statements is fiction.

Apple kicked off WWDC this week with iOS 27 and a new era of Apple Intelligence. What marketers should be watching: → Deeper Siri integration baked into every app → New design language rolling out across native apps this fall → iPhone Fold confirmed for later this year and a new form factor = new ad canvas

OpenAI just filed for an IPO. The company is valued at $852 billion, raised $122B earlier this year, and now needs to show Wall Street it can actually make money. Their plan: turn @ChatGPTapp into a superapp, shift focus from chatbots to AI agents, and bet that ads become their biggest revenue driver by 2030.

OpenAI is ditching the chatbot model to transform @ChatGPTapp into a superapp with AI agents, coding tools (Codex), and third-party integrations. The pivot is IPO-driven. The goal is to attract enterprise clients, compete with @AnthropicAI, and finally find a path to profitability. The update drops in the coming weeks.

Apple is putting Google's Gemini inside Siri for the heavy reasoning. Two years ago they promised their own models would carry it. Admitting that didn't work, at this scale, takes more guts than most CEOs have.

Most people use AI to answer simple questions. Experts use it to change how they work. It's all about how you extract value. The pros go past asking questions: -Context engineering -Iterating outputs -Analyzing files -Building workflows -Automating tasks -Writing code Pick one thing you do daily and build it.