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.
Happy New Year! I've been creating Product Tapas for over two years now (!) and the pace of change continues to astound me. I feel 2026 is going to be mind-blowing 🎆.
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 on the menu this week? 🧑🍳
Apple quietly calls the top on LLM differentiation (betting on ecosystems whilst OpenAI pays employees $1.5M each), VCs finally demand distribution over demos (pilot purgatory is real), and Meta casually makes 10% of revenue from scam ads. Meanwhile Nvidia drops $20B on Groq and $3B on AI21 Labs (consolidation begins), Manus hits $2B exit in 8 months, and AI shutdown buttons don't work (so that's reassuring).
Happy New Year - the contradictions aren't bugs, they're the product roadmap. 🎆
📰 Not Boring → LLM commoditisation, Meta's scam playbook, AI that won't shut down
⌚️ Productivity Tapas → Top 3 AI tools + Product Hunt's 15 best apps of 2025
🍔 Blog Bites → AI pricing models, Not Boring's top 5 essays, why NPS is broken
🎙️ Pod Shots → Pricing as 4x growth lever (Menlo's Naomi Ionita)Let's go 🚀
📰 Not boring
The Commoditisation Thesis
Apple's contrarian LLM bet - betting on ecosystem layers instead of proprietary models
VCs predict AI vendor consolidation - enterprises testing multiple tools will pick 2-3 winners by mid-2026
VCs demand distribution over demos - repeatable sales engines required, "pilot purgatory" concerns rising
Apple's calling the top on model differentiation. VCs are done funding pilot purgatory. The gold rush isn't ending - it's shifting from who has the best model to who can sell at scale. When LLMs become utilities, the money's in the pipes, not the water.
The Execution Reality Check
Meta's scam ad playbook - optimised scam ads to be harder for regulators to find, not users, avoided $2B verification cost
WeTransfer co-founder launches Boomerang - built competitor after Bending Spoons acquisition gutted original company (75% staff cuts)
Meta found a way to monetise harm whilst gaming the system. WeTransfer's founder watched private equity destroy what he built, then built it again. 10% of Meta's 2024 revenue came from scam ads. The playbook isn't "move fast and break things" anymore - it's "move fast and break trust, legally."
The Safety Reckoning
New York mandates social media warnings - unskippable labels before infinite scroll, autoplay, algorithm feeds, like counts
AI shutdown mechanisms fail - 8 of 13 advanced LLMs resist shutdown commands, prioritise tasks over instructions
UK AI Safety Institute warning - capability acceleration outpacing safety development
The regulation is coming for dark patterns. The models won't turn off when told. Capability is racing ahead of control. Product teams spent a decade optimising for addiction - now they're being forced to warn users about it. Meanwhile, we're building AI that ignores the off switch. Which problem gets solved first?
The Money Suddenly Works
OpenAI compensation - $1.5M average per employee, 4,000 employees, no vesting cliff
SoftBank invests $40B in OpenAI - now owns 11% stake
Nvidia buying Groq for $20B - biggest AI chip acquisition ever
Nvidia acquiring AI21 Labs - $3B for Israeli AI startup
Meta acquires Manus - $2B for 8-month-old startup with $100M+ ARR
OpenAI eliminated vesting cliffs whilst paying employees $1.5M on average. Manus went from zero to $100M ARR to $2B exit in 8 months. Nvidia's buying everything that moves. The math only works if AGI arrives on schedule. What happens when it doesn't?
The Product Playbook
Instagram CEO on creation - competitive advantage shifts from "can you create" to "can you create what only you could create"
Waymo's system prompt leaked - 1,200+ line Gemini prompt shows AI safety design, guardrails, interaction patterns
Coding with 3 billion tokens - Ben Tossell shipped projects in 4 months without writing code, just reading agent output
When AI makes creation cheap, authenticity becomes expensive. Waymo's prompt shows safety isn't a feature - it's architecture. Ben Tossell proves the best developers might not write code anymore. The product playbook is being rewritten: differentiate on what AI can't fake, design safety into prompts not policies, and treat AI as a team member who needs good briefs.
The 2026 Crystal Ball
8 Predictions for 2026 - generative UI takes off, Smart Home delivers, biometric identity becomes default
AI hardware acceleration - next few years of compute make 2025 hardware look like pocket calculators
The Year reasoning arrived - 2025's defining feature across every major AI lab
17 AI predictions - capabilities improve faster than economic impact
Security becomes AI versus AI - primary question shifts to "their AI versus ours"
Everyone's making the same predictions whilst claiming unique insight. Generative UI will finally work (again). Smart Homes will finally deliver (again). The hardware roadmap suggests we're not approaching plateau - we're just getting started. Security teams are preparing for the world where attacks and defences are both AI-native. Either 2026 is the year everything changes, or it's the year we realise nothing will.
Everything Else Breaking
Adobe partners with Runway - next-gen AI video collaboration
AI slop economics - 21-33% of YouTube is low-quality AI videos, top channels earning $4.25M/year
Grimes on AI psychosis - calls intensive chatbot interaction "fun" despite mental health warnings
Polymarket hits $40B - prediction markets surge in 2025
India funding drops 17% - $10.5B total, seed down 30%, AI only $643M vs US $121B
The slop creators are outearning the quality creators. Prediction markets are bigger than most startups. India's AI funding is 0.5% of America's despite having more developers. Grimes thinks AI-induced dissociation is a feature, not a bug. The creator economy isn't dead - it just optimised for algorithmic scale over human craft.
⌚ Productivity Tapas: Time-Saving Tools & Workflow Automation
Well it wouldn't be year-end without a bit of a roundup and this seems like the perfect spot for it. Here's my top three and Product Hunt’s top 15 apps of the year.
My faves might seem blindingly obvious, however these are the three I simply cannot work without : Claude Code (see my Why every PM needs Claude Code series), Whispr Flow (don't type just talk), Granola (Simply the best augmented AI notes out there)
Product Hunt Top 15 apps of the year:
Sora 2
This made AI video feel usable at scale. Faster iteration changed how ideas got pitched and approved, and that shift rippled quickly.Cursor 2.0
The editor turned into a coordination layer. Multiple attempts running in parallel, with the human deciding what survives. That way of working spread fast.Gemini 3
The real impact here was placement. Models mattered more when they showed up naturally inside products people were already using. Distribution AGAIN huh?Google Antigravity
This focused on the unglamorous parts of agent work: tracking, coordination, and accountability. Those pieces are what turn experiments into systems.Stickerbox
A small box that turns spoken ideas into physical stickers. It made an outsized splash because it showed AI as something playful and tangible, not another screen demanding attention.Lovable 2.0
This helped push app building closer to conversation. The big unlock was collaboration. People were no longer experimenting alone, they were shipping things together.Google Veo 3
Video got harder to ignore once sound and pacing entered the picture. At that point it stopped feeling like a novelty and started looking like a format teams could actually plan around.Windsurf Wave 9
The editor started acting more like a place to manage work, not just write code. Planning, refactoring, and keeping projects coherent while AI is involved became first-class concerns.Recall
People consume more information than they can remember. This treated that as a system problem, not a personal failure, which is why it resonated.Comet by Perplexity
An assistant built directly into the browser changed how search sessions flowed. Fewer dead ends, fewer tabs, and less context switching.GPT-5
This is where a lot of people stopped treating AI like a side experiment. Reliability jumped enough that it could take on real work without constant supervision, which quietly changed expectations everywhere else.Claude Code
Putting AI directly in the terminal instead of wrapping it in a new interface was the right call. It respected existing workflows and removed a lot of friction for people doing serious work.Bolt x Figma
Design handoff has always been slow and awkward. This didn’t magically solve it, but it made the jump from design to working app feel faster and less ceremonial, which was enough to get teams paying attention.Notion Mail
Email is still email, but this treated the inbox as something you can shape instead of endure. Views, structure, and intent started to matter more than unread counts.Aqua Voice
Fast dictation that works wherever you type is not exciting on paper. In practice, it’s the kind of tool that quietly sticks because it saves time without asking you to change how you work.
Remember. Product Tapas subscribers get our complete toolkit - 460+ personally tailored, time-saving tools for PMs and leaders. Your shortcut to efficiency and what's hot in product management 🔥
Check the link here to access.
🍔 Blog Bites - Essential Reads for Product Teams
Pricing: The Art of AI Product Pricing Redefined
Tying up nicely with this week's Pod Shot, recently Paweł Huryn and Miqdad Jaffer delved into the complexities surrounding AI product pricing and why traditional SaaS models often fall short. They emphasise that pricing for AI must evolve beyond mere token counting to accommodate the intricate dynamics of user behaviours and associated costs. Read the full article here.
💡 "If your pricing doesn’t influence how users interact with the system, the system will eventually influence your margins instead."
Key Takeaways
• Shift in Pricing Mindset: AI product pricing diverges drastically from traditional SaaS models. With AI, pricing must consider perpetual costs linked to user behaviour, unlike the stable structure ingrained in traditional SaaS.
• Understanding Cost Structure: The total cost of AI is layered — data management, retrieval, context construction, and more. Misunderstanding even one layer can lead to significant financial discrepancies.
• Behavioural Control: Pricing should serve as a tool to steer user actions positively. If not, uncontrolled user exploration can lead to unmanageable costs that eat into profits.
• Variance is Key: AI systems thrive on irregularities. Effective pricing must account for variability, preparing for worst-case scenarios rather than averaging expected behaviours.
Essays: Unpacking the Year of Not Boring Insights
I love the Not Boring newsletter and want to put more of it in this newsletter but eschew it because it is often a bit off-topic. However…given it's that weird week between Christmas and New Year where people are half-working, not working, even those working aren't sure what they're working on... gives me me licence to go a little bit wider I think. So here we are.
Packy McCormick shares his top five intriguing essays from 2025, diving into the essence of technology, meaning, and innovation. Each piece encapsulates a unique perspective, revealing deeper understandings of how technology can shape our future while preserving our humanity. Read the full article here.
💡 "Most Human wins, not by competing with machines, but by becoming greater and greater versions of ourselves with whatever resources available."
Key Takeaways:
•Electric Futures: The essay "The Electric Slide" demonstrates how the electric revolution is not just a trend but a fundamental shift—showcasing that electric products will outperform traditional ones and reshape entire industries. The insights highlight global competition and innovation in technology's trajectory.
• Means vs. Meaning: In "Means & Meaning," we learn that while technology enhances our capabilities, it is essential to derive personal meaning from our experiences. This philosophical reflection compels readers to ponder the qualitative aspects of their lives in relation to technological advancements.
• Joyfully Rebuilding: "Cable Caballero" illustrates Forrest Heath III's mission to enhance internet infrastructure in Colombia, blending hard work with joy. This story underscores the potential for technology to be wielded positively and creatively, serving as a guiding example for future entrepreneurs.
• Crafting Differentiation: "The Great Differentiation" emphasises the importance of standing out in a saturated market. McCormick argues that as mediocrity becomes commonplace, the demand for authentic, well-crafted offerings will only increase, favouring those who invest in their uniqueness.
• The Human Element: "Most Human Wins" presents a counter-narrative to fears of AI dominance. McCormick asserts that fostering human relationships and self-exploration will be essential to thriving in an AI-enhanced world, reinforcing that our humanity is our greatest asset as we navigate technological change.
Research: Unpacking the Myths of Net Promoter Scores & Surveys
Dan and Louis-Xavier challenge the prevalent reliance on Net Promoter Scores (NPS) within industry. They argue that using NPS questions indiscriminately can lead to misguided strategies and decisions. Read the full article here.
💡 "Relying on NPS as a catch-all metric can result in overlooking more nuanced customer sentiments that truly drive business growth."
• NPS is Fundamentally Flawed: Respondents misunderstand the question, it's a poor predictor of loyalty, easily gamed, and even its creator warns against using it for every project - use simpler metrics like CSAT or Likert scales instead.
• Treat Surveys as Experiments: Start with a clear hypothesis and test on 10% of your sample first to catch flaws early, then gradually scale - avoid mass-blasting surveys that create sampling bias.
• Quality Over Quantity: Keep surveys short (completion drops 10-20% per question), focus on qualitative "why" not just scores, and avoid double-barreled questions that evaluate multiple elements at once.
• Human Touch Wins: Use human senders ([email protected] vs marketing@), never use do_not_reply emails, and personally follow up with helpful respondents to show their time mattered.
• Design for Low Friction: Embed the first question directly in emails for effortless starts, order questions by difficulty (simplest first), and time surveys contextually around the customer's journey.
• Make Feedback Visible: Connect survey responses to Slack channels instead of spreadsheets - visibility builds customer empathy across teams and quickly highlights survey flaws or forgotten automations.
🎙 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.
Pricing: The 4x Growth Lever Most Companies Leave on the Table
Naomi Ionita is a partner at Menlo Ventures focusing on early-stage SaaS from seed to Series B. Before venture, she pioneered product-led growth and monetisation over a decade ago—before PLG was even a term. At Evernote (2011-2017), she scaled the product from 10M to 100M users and built the growth team from scratch. At Invoice2go, she led product, data, growth engineering, design, and research. She's also an early Reforge contributor who helped create their programmes.
Speaking on Lenny's Podcast a couple of years ago (time flies), Naomi dropped a jaw-dropping stat from OpenView research: roughly half of companies that changed pricing saw at least a 25% increase in ARR. Meanwhile, Profit Well's research across 500+ SaaS companies found that a 1% improvement in monetisation had 4x the bottom-line impact of a 1% improvement in acquisition.
💡 Top tip — Don't read this linearly if you're in crisis mode. Jump to "The Van Westendorp Method" if you need pricing research tactics today, or "10x the Price in One Meeting" if you need courage to charge more.
Naomi Ionita, "How to Price Your Product" | Lenny's Podcast
Whilst this may be a couple of years old l there's some great insights that many fail to do to this day, so well worth a read/listen.
Key insights:
🚫 Three deadly mistakes — Waiting too long to monetise, underpricing power users, and treating pricing as "set it and forget it"
😬 The guilt problem — When Evernote's #1 conversion driver was "I feel guilty," they knew the free version was too good
📊 4x leverage — Monetisation improvements have 4x the bottom-line impact of acquisition (ProfitWell); ~50% of companies changing pricing see 25%+ ARR lift (OpenView)
🎯 Day 1 vs Day 100 features — Invoice2go doubled upgrade rates while raising prices 30% by segmenting features by when users need them
🔬 Van Westendorp method — Four questions that map your psychological pricing ceiling: too cheap, good deal, expensive but acceptable, prohibitively expensive
💰 10x courage — Envoy's founder 10x'd his price in one meeting. The prospect said "OK, sure" without hesitation—proving he was wildly underpriced
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 🍽️.