📰On The Record – 15th August 2026

Reddit found out yesterday it's officially joining the S&P 500 on August 18, stepping into AvalonBay Communities' old spot after that company got swallowed up by Equity Residential. The stock jumped on the news, and here's why: funds that track the S&P 500 now have to buy Reddit shares whether they like the company or not, with JPMorgan estimating roughly 16.7 million shares' worth of forced buying. That's a wall of money coming in regardless of whether Reddit's ad growth outpaces worries about AI eating into its web traffic. It's a reminder that in modern markets, sometimes the biggest price mover isn't a great quarter — it's simply showing up on the right list.

OpenAI and Anthropic spent yesterday locked in an increasingly public price war, cutting the cost of using their AI models as cheaper Chinese and open-source competitors chip away at their edge — OpenAI reportedly took one of its models down by a jaw-dropping 80%. It matters because this isn't generosity, it's survival, and it signals that the AI industry's easiest growth phase — the one where customers paid whatever was asked — is over. The broader implication is bigger than two companies undercutting each other: it suggests the market is starting to price artificial intelligence the way it prices any other commodity, which is bad news for anyone who built a valuation model assuming it never would.

A survey of more than 650 employers found something a little uncomfortable: companies are racing to roll out AI, but most HR leaders don't think their middle managers are actually ready to lead that change. Managers are being handed bigger teams, told to drive AI adoption, and asked to review AI-generated work — often all at once, with little extra training to show for it. It's a classic case of strategy outrunning support: leadership sets the pace, and middle management absorbs the whiplash. If companies want AI adoption to actually stick rather than quietly fail at the point of implementation, the fix isn't more mandates from the top — it's giving the people in the middle room to actually learn the job before they're expected to lead it through change.

The best AI rollout in the world still needs a human in the middle who knows what they're doing.

It emerged yesterday that JPMorgan ended its direct banking relationship with prediction-market platform Polymarket back in October, telling the company to go find somewhere else to keep its money amid regulatory concerns. Polymarket has since found a new banking partner, and JPMorgan hasn't fully walked away — there's still talk of the bank potentially helping with a future IPO. But the exit says a lot about how nervous big banks still are about platforms where people bet real money on real-world outcomes, even as prediction markets edge closer to the financial mainstream. It's a preview of the tension that's coming: fast-growing, rule-bending platforms versus institutions that need every box ticked before they'll get near them.

Dating apps built an entire industry on the swipe, and now the ones doing best are the ones backing away from it. Bumble is shifting toward fewer, better-considered matches instead of maximising volume, while Tinder is leaning into features that push people toward real-world meetups and shared activities. It's not that Gen Z has given up on the apps — they're just tired of matching for the sake of matching and want the tech to actually get them offline faster. It's a useful lesson for any platform built around engagement metrics: sometimes the number going up is exactly the wrong thing to be optimising for. Growth built on volume eventually has to answer to the question of whether it ever led anywhere real.

The apps that survive the next decade will be the ones that measure success in relationships, not screen time.

CEOs, politicians, regulators, and analysts talk for hours on podcasts every day.

Particle actively transcribes 125,000+ podcasts and turns them into structured data, available over REST API or MCP, searchable within minutes of airing.

Track a company or person by name, with each mention cited to the exact quote, speaker, and timestamp, or search for commentary on a thesis or sector with semantic search.

Particle is built by former Twitter and Tesla engineers, using an AI-native transcription pipeline that delivers high accuracy and extensive data enrichment.

Start with $10 in free credit.

Investors in Anthropic are reportedly aiming for a $2 trillion-plus valuation when the company goes public as soon as October — a number that would make it the biggest IPO ever, bigger even than SpaceX's recent listing. The bet hinges on staggering revenue growth, with backers expecting annualised revenue near $100–120 billion by year's end. It's a huge vote of confidence in AI demand, but it also raises the stakes: if growth even slightly disappoints, that valuation has a long way to fall.

Uber is expanding its partnership with Chinese self-driving firm Pony.ai to put more than 2,000 robotaxis on European roads, building on their existing service in Zagreb and adding four more cities plus a push into the Middle East. Neither company has confirmed exactly where yet, but the scale of the rollout shows Uber is serious about not getting left behind on autonomy, while Pony.ai gets a fast track into markets it couldn't easily crack alone.

Goldman Sachs is talking to insurers, asset managers and private-credit firms about funding Nvidia's enormous $500 billion AI infrastructure push, having landed a central role in arranging the financing. Nvidia itself could backstop up to $125 billion of the deal. This is what AI's real cost looks like up close — building the data centres to run all this software is now creating an entirely new corner of the debt market, with chips effectively becoming collateral.

The old media playbook — chase the biggest possible audience, sell ads against it — is looking increasingly shaky, and the companies doing well right now are the ones that gave up on trying to be everything to everyone. Outfits like Semafor, Puck and 404 Media are building around subscriptions, memberships, live events and tightly defined, genuinely engaged audiences rather than mass reach. It's a shift worth sitting with: the winning strategy now is picking your people, not chasing everyone. For any business built on attention — media or otherwise — it's a reminder that a smaller, more loyal audience willing to actually pay you beats a huge one that scrolls past.

In a crowded feed, being for someone beats being for everyone.

AI spend is now a major P&L line item—but most teams can't show what it's producing.

Harmonic Security maps AI activity to use cases and teams, revealing real productivity, shelfware, data risk, and adoption trends across approved and unapproved tools.

Give your board the data behind the return.

See Your AI ROI

A packed news day outside the markets too, from Washington reshuffles to a heatwave testing rail networks — here's what else was moving.

Around a year ago today, Trump and Putin met face-to-face in Alaska for their first summit since the Ukraine invasion, and markets moved before either leader said a word — oil prices slipped as traders bet on eased sanctions risk, and London's FTSE 100 crossed 9,200 for the first time on hopes of a genuine thaw. No ceasefire actually came out of it, and Trump held off on the "severe consequences" he'd threatened Russia with days before. It's aged into a tidy reminder that markets will happily price in an outcome before it exists — the summit alone was enough to move oil and equities, deal or no deal.

The AI industry's sudden willingness to slash prices splits opinion right down the middle.

The case for it: cheaper models mean small businesses and everyday users finally get access to tools that used to be locked behind enterprise budgets, accelerating adoption across the economy — that's a genuine net positive, and it's likely to stick as competition only intensifies.

The case against it: margins this thin make it far harder to justify the sky-high valuations investors have been pricing in, and a price war rarely stays polite for long.

Both cases are probably right at once. What this debate really says is that AI is finally being treated like a normal industry, and normal industries don't get to keep both hypergrowth and fat margins forever.