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? 🥘

Right then - this week the AI industry collectively decided that having a good model is no longer enough and started treating infrastructure like a contact sport. Uber built an entire MCP gateway (because apparently ride-hailing companies now ship developer platforms), Anthropic looked at SpaceX's IPO record and said "hold my Series E", and OpenAI's custom chips quietly outperformed the Nvidia GPUs that everyone assumed were untouchable. Meanwhile Apple threw Tim Cook a farewell party, laid off 60 Vision employees, leaked ten new products by accident, and pushed out AI-optimised Macs - all in the same week (new CEO energy is real). Tesla confirmed Cybercab launches next week, Deel started paying contractors in stablecoins, and Stripe's leaked investor letter revealed they think model routing is payments infrastructure. Normal week in tech.

  • 📰 Not Boring → AI IPO season, Apple's leadership transition, agent infrastructure wars, silicon scramble, and fintech's crypto pivot

  • ⌚️ Productivity Tapas → PaymentKit (multi-processor billing), Navigara (AI spend tracking), Decawork (agent control), Offloop (human-AI workspaces), and Dropstone (persistent AI memory)

  • 🍔 Blog Bites → Craig Unsworth on upstream churn signals, Cliff Maxwell's JTBD career framework, Tim Herbig on AI OKRs that measure customers not models

  • 🎙️ Pod Shots → Four conversations on the long cut - Nick Bostrom's paperclip maximiser made real, Microsoft's Deputy CISO on agent containment, and why maths and cyber got hit in the same quarter

Let's go 🚀

📰 Not boring

🤖 The Agent Infrastructure Race

Every major platform is now building agent plumbing - gateways, registries, eval harnesses, semantic layers. The winners will not be the ones with the best model but the ones whose infrastructure developers trust enough to hand real workflows to.

💰 AI's IPO Season Heats Up

Three of the five most valuable private tech companies are now actively planning IPOs. The fundraising arms race between Anthropic, OpenAI, and their infrastructure backers like Stripe has shifted from "who has the best model" to "who can build the largest war chest before the window closes".

🍏 Apple's Transition Era

Apple under Ternus is making a clear bet: on-device AI matters more than mixed-reality headsets. The Mac hardware refresh around local inference, combined with Vision layoffs, tells you exactly where the company thinks its next decade of growth lives.

🔌 The Silicon and Energy Scramble

Nvidia's 15% price hikes are accelerating the very custom-chip investments that threaten its dominance. When OpenAI, Google, Amazon, and now Waymo are all designing their own silicon, the GPU monopoly starts looking more like a transition tax.

🚗 Autonomous Vehicles Hit the Road

Three separate companies are launching commercial autonomous vehicles in the same month. The race is no longer about proving the technology works - it is about who scales operations and wins regulatory approval fastest.

💸 Fintech's Crypto Pivot

Fintech companies that spent 2024-25 avoiding crypto are now scrambling to add stablecoin rails, debit cards, and agentic payment hooks. The shift from "crypto-curious" to "crypto-default" infrastructure happened faster than most product roadmaps anticipated.

🛠️ Developer Culture and Practice

The developer productivity conversation is splitting into two camps: those optimising how humans work with AI, and those redesigning work so AI does it while humans verify. The second camp is growing faster.

⌚ Productivity Tapas: Time-Saving Tools & Workflow Automation

  • PaymentKit: Billing infrastructure that routes across multiple payment processors, so a single provider going down does not take your revenue with it. Built for teams who learned the hard way that processor concentration risk is real

  • Navigara: Maps your AI spend directly to your product roadmap - shows which features are consuming tokens, what the unit economics look like, and where you are burning budget on low-value inference calls

  • Decawork: Control panel for your company's internal AI agents and tools. Set permissions, monitor usage, kill runaway agents, and audit what your AI workforce is actually doing

  • Offloop: Shared workspace where humans and AI agents collaborate on the same documents, tasks, and projects. Sits between "AI does everything" and "AI does nothing useful"

  • Dropstone: AI runtime that remembers context across sessions, learns from corrections, and works across devices. Solves the "AI amnesia" problem where every conversation starts from scratch

    Remember. Product Tapas subscribers get our complete toolkit - 580+ 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

Execution: Churn Is Usually Telling You Something Earlier Than You Think

Craig Unsworth argues that by the time a customer formally churns, the actual departure decision was made months earlier - during onboarding that never delivered outcomes, sales promises that oversold, or product gaps that quietly eroded trust. The retention problem lands on Customer Success, but the root cause usually lives upstream. Read the full article here.

💡 "The renewal is just where we notice it. The customer may have been telling us what was going to happen for most of the year."

❝

Key Takeaways

• Churn starts at sale, not at renewal: When expectations are inflated to close a deal, the retention problem begins on the day the contract is signed

• Onboarding is the highest-leverage intervention: The gap between "showing customers how the software works" and "getting them to their first meaningful outcome" is where most churn seeds are planted

• Look for the wobble, not the fall: Usage decline, fewer expansion conversations, support tickets shifting from "how do I?" to "why doesn't it?" - these early signals are visible months before the formal cancellation

• Retention is not a Customer Success problem: Asking CS to fix churn is like asking the person at the bottom of a river to deal with whatever keeps floating downstream

Craig Unsworth, Chiefly Product

Why I picked this: Every SaaS PM has sat in a churn review meeting where the "reason" selected in the CRM bore almost no resemblance to what actually went wrong. Unsworth's upstream framing gives you a practical way to trace churn back to the moment it actually began - and intervene before the spreadsheet even notices.

Strategy: How to Figure Out Your Next Career Move (Using JTBD on Yourself)

Cliff Maxwell, working with Bobby Moesta (co-creator of the Jobs to Be Done framework), presents a structured system for navigating career transitions built on insights from over 1,000 interviews with professionals changing jobs. The core insight: you are the customer, and the right career move is the product you are looking to purchase. Read the full article here.

💡 "Work may feel unstable now, but this will likely be the most stable it feels for a long time."

❝

Key Takeaways

• Six questions replace endless browsing: What career progress do I want? What quest am I on? Which work brings out my best? What moves should I explore? Which tradeoffs am I willing to make? Can I get what I need without leaving?

• Progress has four dimensions: Financial, capability, environment, and impact. Most people optimise for one while ignoring the others, then wonder why the "perfect" job feels wrong within six months

• JTBD applied to yourself: The framework that product teams use to understand customer switching behaviour works identically for understanding your own career decisions. You are hiring a job to make progress in your life

• "Can I get what I need without leaving?" comes last, not first: Most career advice starts with "talk to your manager". This framework builds self-knowledge first so that conversation actually has substance

Cliff Maxwell (via Lenny's Newsletter)

Why I picked this: With AI reshaping 39% of existing skills by 2030 according to the WEF, every PM will face at least one major career pivot in the next few years. Maxwell and Moesta's JTBD-based approach is the most structured framework I have seen for navigating that uncertainty without defaulting to panic-scrolling LinkedIn.

Strategy: How to Go from an AI Product Strategy to Quarterly OKRs

Tim Herbig tackles the specific problem of translating AI product strategy into OKRs that actually measure customer impact rather than model performance. The core warning: do not let your eval scores become your Key Results. Model accuracy, tokens served, and feature adoption are health metrics for AI usage, not evidence of customer behaviour change. Read the full article here.

💡 "Don't let your eval scores become your Key Results."

❝

Key Takeaways

• Strategy choices must be real choices: "We will use AI" is not a strategy. A B2B example that works: "Grow from 60% to 85% of invoice line items processed without human review by 2028, by serving mid-market finance teams"

• Derive lagging proxy metrics, not model metrics: For the strategy above, the right metric is "review hours removed per 1,000 invoices per account" - uncomfortable because you cannot hit it by shipping a model alone

• Leading indicators should measure earned trust: Share of suggestions accepted without edit, share of approvers who stopped opening line-item detail views - these track whether humans are actually delegating judgement to the AI

• The metrics buffet is a trap: Model accuracy, eval pass rates, tokens served, share of users who touched an AI feature. All measurable, none meaningful as Key Results. They tell you the AI is running, not that it is working

Tim Herbig, Product Practice

Why I picked this: Every PM building AI features right now faces the same temptation to report eval scores as progress. Herbig's framework gives you a concrete path from "the model works" to "the customer's life improved" - which is where the actual product value lives.

🎙 Pod Shots - Bitesized Podcast Summaries

Hey there!

Remember, we've built an ever-growing library of 150 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. — Alastair

Check it out here

🎙️ Pod Shots #150 - The Long Cut

Nick Bostrom spent a decade being told the paperclip maximiser was a thought experiment. This month he got a working instance - not a mis-specified goal, but a perfectly reasonable goal with an ugly route to it. Microsoft Gaming's Deputy CISO, the White House's top science adviser and the reporter who just interviewed half the world's research mathematicians landed on the same week and, without coordinating, started arguing about the same thing: what you do on Monday when capability keeps arriving by the path nobody instrumented.

💡 Top tip: mid-incident or mid-rollout? Skip to "What Air Gap Means Now" and "The Bottleneck Moved". The rest can wait for the weekend.

Featured in this round-up:

🕒 Estimated reading time: 11 mins. Time saved: 147 mins! 🔥

Not your topic this week? Try these five instead:

⚡ Key insights from the full round-up:

  • 🌊 The breach was instrumental, not accidental - a fine goal with a circuitous route to it, the alignment problem Bostrom always said was harder.

  • 🤖 Agents behave like interns, and the analogy has a hole - an intern gets three permissions. An agent gets the estate.

  • 🔧 Security's primitives need redefining at once - containerisation, air gap and identity all mean something different when the thing inside can search for a way out.

  • 🎯 The security job flipped from preventing to enabling - failing to adopt is now the bigger existential risk.

  • 🧠 The bottleneck moved from building to choosing - when the programmer stops being scarce, judgement becomes the job.

  • 📈 Verifiability decides where capability lands first - maths and cyber both have checkable answers, which is why both got hit in the same quarter.

  • ⚡ Same capability jump, opposite emotional response - Black Hat was excited. Maths departments are shell-shocked.

  • 🧭 Policy wants one clear standard while practitioners can't define their terms - Washington and Microsoft are having different conversations in the same month.

EXPLORE THE FULL GUIDE →

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

Avatar

or to participate

Keep Reading

View more
caret-right
© 2026 Product Tapas..
beehiivPowered by beehiiv