Hi Reader
If you’re new here, welcome. Each Sunday this lands in your inbox with a selection of things I’ve noticed and bookmarked during the week – stories that say something useful about how work, technology and business are playing out. No hot takes, no hype cycle chasing. Just material that’s worth your time if you build, run or think seriously about businesses.
Enjoy!
AI growth creates ‘new-collar’ jobs
Amid the familiar claims that AI is hollowing out entry-level work, LinkedIn’s latest labour market data tells a more nuanced story. Drawing on platform-wide hiring and skills data, the company says that AI is reshaping roles rather than wiping them out, driving demand for what it calls “new-collar” jobs that combine technical capability with distinctly human skills. Hiring overall remains sluggish and confidence is low, but growth in AI-related roles, in-house leadership positions and creator-led work suggests the labour market is fragmenting rather than collapsing. The real challenge, the data implies, is not job destruction but preparedness – with many workers feeling under-skilled for where demand is moving. Read the report here.
The accounting mistakes startups keep making
Accounting is one of those things founders often park until later – usually just before a funding round, year end or HMRC deadline. A useful SeedLegals article walks through the small, common mistakes that compound into real problems: messy cap tables, missed filings, misclassified spend, VAT confusion and payroll oversights. It’s a great (and short!) read for anyone who’s moving fast and wants to avoid expensive clean-up work later. Check it out here.
Waymo shows why scaling is harder than building
I’m genuinely excited for the arrival of driverless taxis in my city, which the Guardian reports as scheduled for late 2026 in London. And on a business level, Waymo is a reminder of how technological shifts play out. The company is approaching a reported $110bn valuation on the back of tangible progress – millions of paid rides, meaningful revenue and safety data that looks genuinely better than the human alternative. But the harder part now isn’t engineering, it’s integration: public trust, regulation, urban fit and who controls the data that proves the system works. It’s a reminder that breakthrough products don’t win on capability alone – they succeed or fail on how well they coexist with existing systems. The biggest opportunities – and risks – sit in that messy layer between what technology can do and what society will tolerate. TechCrunch offers a deeper dive on the story.
What does ‘middle class’ even mean now?
Two stories surfaced this week – neatly pulled together in the always excellent Sunday Strategy newsletter – that reveal how slippery the idea of the middle class has become. On one hand, Yale University is moving to tuition-free education for families earning up to $200,000, a level of income that would once have signalled clear financial comfort but now barely stretches across housing, childcare and fees in many cities. On the other, new analysis from Pew Research Center shows the middle class shrinking in size and influence, with incomes rising far more slowly than those at the top and a growing sense of economic drift in the middle. Read together, they raise an uncomfortable question for founders, employers and parents alike: when six-figure households qualify for aid, and the middle no longer feels stable, what does “doing fine” look like?
Is Jony Ive making… a pen?
A new leak suggests that the first hardware collaboration between Jony Ive and OpenAI could be something as unassuming as a pen. If true, that minimalism feels consistent with what Ive hinted at in recent interviews, including the one I highlighted a couple of weeks ago: less focus on screens and features, more attention to how technology fits into everyday behaviour. Sam Altman has already said the device will be simpler than a smartphone, and a pen-like form factor would put AI into something people already use, rather than asking them to adopt a new kind of device. I guess the thinking is that products scale faster when they fit existing behaviour instead of trying to change it. Mashable has the story.
Dealmaking on tougher terms
Global M&A is moving again, but it isn’t snapping back to old habits. A write-up from the London M&A Forum shows how deals are now being shaped inside a more complex system – heavier regulation, louder activism, geopolitical risk and greater board scrutiny. The consistent message across regulators, investors and advisers was pragmatic rather than pessimistic: preparation now starts earlier, regulatory strategy is built in from day one and value is increasingly shaped by early decisions on structure and governance. For a clear picture of how deals are being done in 2026, check out the round-up here.
Why physical channels still cut through in an AI-heavy world
As digital marketing becomes more automated, strategist Tom Roach looks at why direct mail is still being used by brands that are otherwise heavily digital-first. His article focuses on how physical mail performs on basic measures such as attention time, recall and trust, drawing on industry studies and recent campaign data. The argument is straightforward: items that arrive through the post are handled, kept and noticed for longer than most on-screen messages. For businesses thinking about customer acquisition, brand recall or credibility, it’s a clear look at where physical channels still earn attention. Note that it’s published by MarketReach – part of Royal Mail – so it may not be totally objective, but still well worth a read here.
We are more creative than AI – but only just
A new study reported by ScienceDaily compares the creative output of more than 100,000 people with that of leading generative AI models using the same creativity benchmarks. The researchers found that AI systems now outperform the average human participant on idea-generation tasks such as divergent thinking. However, the highest-scoring human participants (the top 10%) still produced more original and higher-quality work, particularly in tasks involving storytelling and richer forms of expression. The study also notes that AI performance varies significantly depending on prompt design and constraints, underlining how much human input still shapes the output. Find out more here.
AI prompt of the week: challenging your audience assumptions
Most marketing strategies are built on assumptions about target audiences – stereotypes borrowed from industry playbooks that feel true but might be outdated. By forcing AI to articulate its assumptions and identify contradictory evidence, you get a more nuanced view of who you're trying to reach and where your strategy might be missing the mark.
You are a marketing and content specialist creating content to maximise engagement with my target audience. Based on the profile below, list the 5 key assumptions you're making about this audience, explain why you've made each assumption and identify evidence that might contradict each one. This helps me understand where I might be relying on stereotypes vs. reality.
Audience profile:
· Age range: [e.g. 30 to 40 years old]
· Gender: [e.g. Male]
· Location: [e.g. London, United Kingdom]
· Industry/sector: [e.g. Media]
· Job title: [e.g. Digital Marketing Manager]
What business acquirers scrutinise
Buyers don’t value businesses on a single headline number. They look for a balanced picture of how growth is generated, how durable it is and where risk sits. It’s a useful checklist for sanity-checking how your business would look under diligence, long before a deal is on the table.
Drop me a line
You can reply directly if you want to say hello or point me at a piece I’ve missed. Sometimes those conversations inform future editions; sometimes they’re just supremely useful to have.
Cheers!
Adam