The Captain's Log

AI Weekly – 28 July 25

This week generative AI got practical – and occasionally weird. Netflix broadcast its first AI-generated video scene, Midjourney unlocked perfectly looping clips, and Walmart put “super agents” to work for 240 million shoppers. WeTransfer backed down after an artist backlash, Trump launched an “un-woke” AI plan, and Vogue ’s flawless new cover star turned out to be 100 % synthetic. Buckle up.

In this week’s briefing

1. AI in Content & Creativity

2. AI in Marketing & Sales

3. AI in Business & Adoption

4. Tools, Models & Platforms

5. AI Agents & Automation

6. Policy, Power & Ethics

7. AI in Society, Health & Education

8. And Finally…

1. AI in Content & Creativity

Netflix collapses a tower – with GenAI video: El Eternauta features Netflix’s first on-screen AI-generated VFX, delivered ten-times faster than traditional work, says Co-CEO Ted Sarandos. Source

Midjourney adds start/end-frame control – hello seamless loops: Three weeks after launching video, Midjourney now lets you specify the exact first and last frames so clips loop perfectly (or hit a precise intro/outro). They're also making video generation available vis their Discord channel. Source

Pika Labs launches the first AI-only social video app (iOS): Pika claims it's new Social AI Video lets users replace faces, gestures and props by text prompt; early adopters rave about “Pikaswaps”, but it's still in beta and requires an invitation to use. Source

2. AI in Marketing & Sales

Walmart pilots four e-commerce “super agents”: Customer-facing Sparky plus agents for staff, suppliers and devs push Walmart toward 50 % online sales. Source

PwC : AI-fuelled ads will push media to $3.5 trn by 2029: The Global Entertainment & Media Outlook sees digital ads hitting 80 % of revenue, thanks to generative-ad personalisation. Source

OpenAI pilots a checkout button inside ChatGPT: Type, click, buy: selected US users can now complete purchases without leaving the chat window. Source

3. AI in Business & Adoption

Microsoft memo: “Invest in AI skills” after 9 000 layoffs: In a note to staff, Microsoft CEO Satya Nadella acknowledged the job cuts but said Microsoft’s next growth spurt hinges on security, quality and AI transformation. Employees were urged to upskill in Copilot and other in-house AI tools to stay relevant as the company doubles down on artificial-intelligence products. Source

Uk government signs OpenAI MoU to modernise public services: Pilots cover NHS triage bots and local-council form helpers; the aim is to “unlock billions” in productivity. Source

RELX says GenAI tools are lifting profits: demand for its generative AI tools from lawyers and scientists boosted half-year profit by 9%. Source

4. Tools, Models & Platforms

Google & OpenAI AI models win gold at the world's most prestigious Maths Olympiad: Frontier models solved five of six IMO problems, hitting the human gold-medal threshold and shortening the expected timeline to reach this level according to several leading AI research assistants. Source

Perplexity in talks to pre-install its Comet AI browser: The wannabe Google-search challenger wants Comet shipped as the default on new phones, taking aim at mobile Chrome. Source

Alibaba Group ’s Qwen3-Coder matches Anthropic 's Claude Sonnet 4 on tool-use: A leaner model aces complex coding chains while running on less compute. Source

Amazon buys Bee – a £40 bracelet that records your life – The always-listening wearable transcribes conversations, sets reminders and stores “memories”. Amazon says users keep full control; privacy lawyers are unconvinced. Source

5. AI Agents & Automation

OpenAI Agent Mode now live for UK, EU & friends: The phased roll-out is complete; Plus, Pro and Team users everywhere can delegate multi-tool tasks. User feedback is mixed, but no-one is doubting it will improve and heralds the mainstreaming of AI agents. Source

Lovable ’s “agent mode”, now in beta, cuts build errors by 90 %: Multi-step coding agent reads, plans and takes actions autonomously Source

6. Policy, Power & Ethics

Trump's newly unveiled AI Action Plan aims to slash red tape and “partisan bias”: Draft executive order contains more than 90 policy actions and would force agencies to justify any rule hampering AI and audit models for “politicised outputs”. Source

Sam Altman, warns banks: voice prints are toast: Deep-fake voice fraud is “imminent”, the OpenAI CEO told the Kansas City Fed, urging urgent counter-measures. Source

WeTransfer U-turns after artists revolt over AI-training clause: Facing a backlash, WeTransfer scrapped language that let it feed user files to training models. Source

7. AI in Society, Health & Education

Ash: a $93 m AI therapy companion: Ash by Slingshot AI ’s free app, trained on clinical data and overseen by psychologists, offers 24/7 listening and CBT-style prompts. Source

OpenAI 's ChatGPT becomes a lifeline for neurodivergent professionals: Structured, non-judgemental feedback helps autistic and ADHD users navigate work and relationships, though over-reliance is flagged. Source

8. And Finally…

“Does this look real?” – AI Vogue model sparks beauty-standard panic: A hyper-real Vogue cover star turned out to be 100 % AI; social feeds lit up with praise, outrage and existential dread. Source