- Product Tapas
- Posts
- Amazon to replace 500k workers. OpenAI's platform play continues with Atlas browser launch. Karpathy tells us to chill on AI agent hype.
Amazon to replace 500k workers. OpenAI's platform play continues with Atlas browser launch. Karpathy tells us to chill on AI agent hype.
Plus: 7 AI-native growth levers; How to optimise for AI search; and Get insanely good ChatGPT answers

We track Product so you don't have to. Top Podcasts summarised, the latest AI tools, plus research and news in a 5 min digest.
Hey Product Fans!
Welcome to this week’s 🌮 Product Tapas.
New here? We're your shortcut to staying sharp. Essential stories, practical tools, real insights.
For the best reading experience, check our web or app version and sign up for future editions here.
What’s cooking this week? 🥘
Amazon's automating a third of its workforce (the robots send their regards), Google's turning anyone into a developer (RIP bootcamps), and Anthropic's picking fights with Trump's AI tsar (bold strategy). Meanwhile, Karpathy says agent hype is "slop", and Trump's posting AI videos of himself as a king dropping, err brown ‘slop’, on protestors (because of course he is).
📰 Not Boring → Amazon's automation reckoning, no-code store builders, Claude's Skills economy
⌚️ Productivity Tapas → Multi-AI analysts, autonomous websites, ChatGPT prompt hacks
🍔 Blog Bites → Micro-interactions from tech giants, Ahrefs' AI search laboratory, AI-native growth levers
🎙️ Pod Shots → ElevenLabs' journey from 50 rejections to £200M ARR
Let's go 🚀
📰 Not boring
The Great Automation Reckoning
Amazon plans to replace over 500,000 jobs with robots by 2033, expecting to save $0.30 per delivery by automating its operations
Walmart Partners with OpenAI to create AI-First shopping experiences; key question - is this additive or just dilute your control over customers?
OpenAI launches AI-powered browser ChatGPT Atlas to rival Chrome, featuring a contextual sidecar chatbot and advanced automation capabilities for Pro users. Early review here
So what? The scale here is breathtaking. Amazon's betting that 30 cents per delivery adds up when you're shipping billions of packages - and they're willing to replace a third of their workforce to capture it. Walmart's clearly taking a gamble, but let's see how it pays off. And Atlas? It's not groundbreaking - Comet, Dia et al are already there - but it signals OpenAI's relentless platform play. They're not content being the AI provider; they want to own the interface too.
Building Without Breaking a Sweat
Google releases Google AI Studio; new vibe coding tool that converts simple descriptions into functioning web apps in minutes
Shopify and Lovable team up to enable anyone to create a fully functional online store via chat prompts
So what? The no-code revolution keeps accelerating. You can now literally chat your way into e-commerce: "Make me a store that sells vintage band t-shirts" - and boom, you're in business. The barrier to entry for entrepreneurship keeps lowering.
The Skills Economy
Claude launches Skills; folders that include instructions, scripts, and resources that Claude can load when needed to improve performance on specific tasks allowing users to customise folders and scripts for specific jobs, enhancing productivity across teams
Many think they're a bigger deal than MCP - see more here
Anthropic launched Claude Haiku 4.5; similar levels of coding performance as Sonnet 4.0 (the previous model) but at one-third the cost and more than twice the speed
So what? Claude Skills is so simple yet so effective. Another sign to me that Anthropic gets the user's needs here. Meanwhile, Haiku 4.5 delivers the trifecta: faster, cheaper, better - pick all three. We're past the point of trade-offs.
The State of AI
Air Street Capital releases its State of AI report
Claude and Cursor have quickly become the go-to tools of the year, while many teams are moving away from ChatGPT and Perplexity
LangChain continues to lead as the preferred framework for building AI agents
Enterprise AI deals have exploded — average contract sizes jumped from about $39K in 2023 to roughly $1M in 2025
AI-native startups are scaling faster than traditional SaaS ever did, hitting $5M ARR about 1.5× quicker than 2018's top players
MCP is emerging as the standard protocol for AI agents, now running on around 15K servers worldwide
Case studies on how Unilever is using AI at Scale across Unilever
So what? The velocity of change here is staggering. We're not just seeing incremental improvements - we're watching entire business models get rewritten in real-time.
Odds and Ends
Google Sheets rolls out new AI functionality to assist product teams in tracking real-time user feedback, revolutionising spreadsheet utility for product management
OpenAI partners with Hollywood to crack down on celebrity deepfake videos: After AI-generated videos featuring Bryan Cranston circulated on Sora
Spotify & Netflix Gun for YouTube's Eyeballs
Apple Readies Its First Ever Touch-Screen MacBook
DoorDash will use Waymo's robotaxis for delivery in Phoenix
You still need to think: nice piece on how people still need to direct work, set goals, choose constraints and judge outputs in the new Agent world
Here are all the agents you can use with Linear. And here's how one PM explains how they use agents internally to deploy simple no regrets updates
Spotify adds voice control to its DJ feature
Everything Sam Altman said on podcasts this week
As Anthropic tries to keep pace with OpenAI, it's also taking on the U.S. government as Co-founder Jack Clark's essay "Technological Optimism and Appropriate Fear" got David Sacks' back up (Trump's AI & Crypto Tsar)
Andrej Karpathy tells us to chill on AI agent hype. "Overall, the models they are not there. And I feel like the industry [...] it's making too big of a jump and it's trying to pretend that this is amazing. And it's not—it's slop!"
AWS services recover after daylong outage hits major sites
Another day, another voice cloning model. Fish Audio S1 sounds pretty good though [less snappy on the branding though]
Driverless taxis from Waymo will be on London's roads next year, US firm announces
Annnnd finally…. Donald Trump has shared an AI-generated video of himself as a king flying a plane and dropping brown liquid on protesters, mocking those involved in the 'No Kings' demonstrations
Become the go-to AI expert in 30 days
AI keeps coming up at work, but you still don't get it?
That's exactly why 1M+ professionals working at Google, Meta, and OpenAI read Superhuman AI daily.
Here's what you get:
Daily AI news that matters for your career - Filtered from 1000s of sources so you know what affects your industry.
Step-by-step tutorials you can use immediately - Real prompts and workflows that solve actual business problems.
New AI tools tested and reviewed - We try everything to deliver tools that drive real results.
All in just 3 minutes a day
⌚️ Productivity Tapas: Time-Saving Tools & Workflow Automation
Compound: run multiple AI analysts on a spreadsheet simultaneously
Flint: build “Autonomous Websites” that auto generate, adapt and A/B test pages based on visitor behaviour or external changes whilst staying on-brand
ChatGPT prompt hacks: how to make ChatGPT give you insanely good answers
Done with those? Here’s a more irreverent list
Remember. Product Tapas subscribers get our complete toolkit - 460+ personally tailored, time-saving tools for PMs and founders. Your shortcut to efficiency and what's hot in product management 🔥
Check the link here to access.
🍔 Blog Bites - Essential Reads for Product Teams

UX Design: Five Brilliant Micro-Interactions That Show How Tech Giants Really Understand Users
It wouldn't be a Product Tapas without a built for Mars link…. This time, the focus is on contextual onboarding, intelligent defaults, dynamic personalisation, transparent resource management and advanced interaction patterns. Read the latest from Built For Mars here.
💡 "Google Maps will default to walking or driving, depending on how far away your destination is."
Key Takeaways
Contextual Onboarding:
•Lovable triggers planning onboarding only after multiple negative prompts, showing intervention at the right moment. This prevents overwhelming new users whilst helping struggling ones when they actually need guidance
Intelligent Defaults:
• Google Maps automatically selects transport mode based on distance, reducing cognitive load. Smart assumptions eliminate unnecessary decision-making for obvious choices
Dynamic Personalisation:
• Spotify adjusts playlist recommendations based on time of day to match user context. Real-time adaptation makes suggestions feel more relevant and intuitive
Transparent Resource Management:
• Lovable clearly shows credit balance alongside daily free allowances. Dynamic indicators help users understand when they're using free vs paid credits
Advanced Interaction Patterns:
• Apple Podcasts allows multi-zone scrubbing with finger movement to control playback speed. Sophisticated gestures provide power users with precise control without cluttering the interface
Strategy: How Ahrefs Turned Their Company Into a Living AI Search Laboratory
Tom from Strategy Breakdowns explores how Ahrefs is dominating AI search by making their entire company a public case study for their own optimisation strategies. They're simultaneously proving their methodology works whilst building the intellectual property that becomes their paid product features. Read the full article here.
💡 "By dogfooding their own product and sharing their research publicly, Ahrefs is now the most cited tool for 'AI search optimisation'. Their success becomes both the proof point and the product."
Key Takeaways:
• Self-Experimentation Strategy: Ahrefs made their entire company the test subject for AI search optimisation, systematically experimenting with content clusters, publication relationships, and messaging frameworks
• Public Documentation: Every experiment is documented publicly, creating a credibility loop where they're practitioners showing actual results rather than consultants theorising
• Meta-Marketing Approach: Their research on AI visibility correlations becomes viral blog posts whilst simultaneously driving their own traffic acquisition strategy
• Research-to-Product Pipeline: When analysis revealed different AI platforms favour different content types, this insight became core functionality in their Brand Radar tool
• Feature Development Loop: Platform-specific tracking, domain analysis, and gap analysis features all emerged from their internal experiments and findings
• Proprietary Data Creation: Every company using Brand Radar generates data about AI search patterns, creating increasingly valuable and difficult-to-replicate datasets
• Market Category Creation: They're not just competing in AI search optimisation—they're defining what it means through frontier research papers and case studies
• Content Marketing Integration: Their optimisation experiments become case studies, research becomes thought leadership, and product usage becomes proof points
• Network Effects: Built-in content marketing where the company's own operations serve as ongoing case studies and social proof
• Correlation Discovery: Found that branded web mentions correlate 66.4% with AI visibility and highly linked page mentions correlate 70% with Google AI Overview visibility
• Platform Differentiation: Different AI platforms (ChatGPT vs Perplexity vs Google AI Overviews) show distinct citation preferences and content type favouritism
• Competitive Intelligence: Their head-to-head brand mention tracking across LLMs provides uncomfortably transparent competitive analysis
Strategy: Seven AI-Native Growth Levers That Could 10x Your Revenue
Andrew Davies explores how AI-native companies are achieving explosive growth rates that make even the outlandish "triple, triple, double, double" scaling look pedestrian. He outlines seven practical strategies that enable startups to reach $100M revenue with teams of just 30 people, operating at $500K-$1M ARR per employee. Read the full article here.
💡 "Five years ago, 70% of early-stage venture capital was spent on product development and 30% on go-to-market activities. As the industry becomes more saturated, those ratios have completely flipped."
This shift reflects the new reality where distribution sophistication often matters more than product features in crowded markets.
Key Takeaways:
Distribution Strategy
• Build distribution directly into your product architecture using four types: Borrow (app stores, platforms), Buy (paid ads, affiliates), Build (email lists, content), and Bake-in (usage-driven distribution). Make your differentiator speed to market and distribution sophistication, not just product features.
Proactive Churn Management
• Implement structured exit surveys from launch asking why customers leave and why they originally chose you. Focus on maximizing value from inevitable departures and build systems to maintain relationships with churned customers for future reactivation opportunities.
Born Global Operations
• Avoid artificial market-by-market expansion that wastes opportunities. Companies like Aithor saw 9% revenue uplift just from localizing payment methods and currencies. Solve taxes, payments, and regulatory challenges upfront to unlock previously inaccessible markets.
Dynamic Pricing Models
• Companies rolling out pricing changes every three months outperform others by 103% on average revenue per user. Shift to hybrid models combining traditional and usage-based pricing to handle AI's variable consumption patterns.
Strategic Market Focus
• Constrain your serviceable addressable market for better decision-making speed and smarter resource allocation. When expanding segments (B2B to B2C or vice versa), adapt accordingly: B2C needs simple onboarding and self-service; B2B needs new pricing structures and organisational value propositions. Maintain consistent mission while adapting business models.
🎙️ Pod Shots - Bitesized Podcast Summaries
Remember, we've built an ever-growing library of our top podcast summaries. Whether you need a quick refresher, want to preview an episode, or need to get up to speed fast - we've got you covered.
Check it out here
🚀 From £2M at £9M Valuation to £200M ARR: The Untold Story of ElevenLabs

ElevenLabs | 20 VC
🎥 Watch the full episode here |
📆 Published: September 8th 2025 🕒 |
Estimated Reading Time: 8 mins. Time saved: 112 mins! 🔥 |
“Building in Europe is building on hard mode. But there are some great advantages too."
ElevenLabs went from two Polish founders facing 30-50 investor rejections to one of Europe's fastest-growing AI companies, crossing £200M ARR with just 250 employees. Co-founder Mati Staniszewski shares the raw, unfiltered story of competing head-to-head with Silicon Valley giants—and winning.
Key insights from the full article:
💰 30-50 rejections — the brutal pre-seed that almost killed the company
🎯 Product-market fit overnight — an audiobook author hacked their entire 500-page book
🤝 How A16Z won the deal — Brian Kim was the only investor who actually tested their APIs
😱 The darkest moment — when a customer launched their product two weeks before them
🧠 Research is a head start, not a moat — they're 6-12 months ahead, then execution matters
📈 £1M ARR per employee — exceptional unit economics with dual revenue engines
🏢 No job titles at 250 people — how they maintain speed and prevent hierarchy
🌍 European talent myth — "They don't work as hard as we do," say the Polish team
🔮 Voice will eat software — the contrarian bet most people don't believe
👉 Read the full breakdown — sent separately, check your inbox
That’s a wrap.
As always, the journey doesn't end here!
Please share and let us know what you would like to see more or less of so we can continue to improve your Product Tapas. 🚀👋
Alastair 🍽️.


Reply