We're halfway through the year, which always sneaks up on me. You blink and the first six months are gone, and this week gave me a natural spot to stop and take stock, because a lot of the news was itself about taking stock. Where the market landed at the half. What AI is quietly doing to the job market. Whether the race between these models is even measurable anymore.
So here's the ten, same as always. What actually happened, the numbers and names first, then a little of what I make of it. My hour to slow down in a week that didn't. Let's get into it.
AI: The Race and the Bill
1. Musk unveiled Grok 4.5, and claimed it rivals the best models out there
In late June, Elon Musk said his newest model, Grok 4.5, had entered private testing at SpaceX and Tesla, running internal workflows and coding-heavy jobs. He didn't hedge the claim either. He said it's as good as, if not better than, Anthropic's Claude Opus. He also said the plan is to ship a brand-new foundation model, trained from scratch, every single month through the end of the year, and that he'd moved dozens of engineers off the Starlink and Starship teams to pull it off. Here's the honest asterisk, and at the time it was a big one: no public access, no independent benchmark, and the engineers running those glowing tests work at Musk's own companies. It did go public about a week after he said all this.
I've learned to take any AI performance claim with a grain of salt, and to take Musk's with the whole shaker, while still paying attention, because he does ship. But the part that actually stops me isn't the bragging, which I've learned to price in like weather. It's the pace, the fact that a company can now stand up a brand-new foundation model from nothing every thirty days, on hardware it built itself, with engineers pulled off rockets to do it. A new model from scratch every month. Not long ago these leaps were measured in years. Now the top labs are trading the lead in a matter of weeks, and this week it was Musk's turn to plant a flag. Whether the flag means anything, we'll know when somebody outside his companies gets to poke at it.
2. A record half-trillion dollars poured into startups, and almost all of it chased AI
On July 2, Crunchbase reported that global venture funding hit a record $510 billion in the first half of 2026, more than all of last year combined. Over 70 percent of the money in the second quarter went to AI companies, up from about half a year earlier. Two names, OpenAI and Anthropic, together pulled in something like $217 billion, nearly half of everything. Anthropic alone raised $65 billion in a single round in late May, at a valuation of $965 billion, which put it past OpenAI and made it the most valuable private company in the world. It has since filed confidentially to go public.
This is the number that reframes all the others. When you read that AI is a big deal, this is what big deal actually means in dollars, an extraordinary share of the entire world's investment capital funneling into a small handful of companies. I find that thrilling and a little unnerving in the same breath. Thrilling because that kind of money builds real things fast. Unnerving because when this much rides on so few players, a stumble by any one of them stops being their problem and starts being everyone's. Concentration cuts both ways, and we're concentrated.
3. The first real sign of what AI is doing to jobs, and it's the young who feel it
Back in June, not this week, but it belongs on the list. A joint report from the World Economic Forum and PwC pulled the picture together. The number everyone is quoting isn't actually theirs, it's from a Stanford study they cite, by Brynjolfsson and colleagues: US employment in the entry-level jobs most exposed to AI has fallen about 16 percent since late 2022. In a survey of more than 9,000 entry-level workers across 48 countries, 68 percent said AI made them more productive, 45 percent said they now work longer hours, and 28 percent said half or fewer of their current skills will still be relevant in three years. PwC has a word for what's happening: seniorization. The entry-level work didn't vanish. It got folded upward into jobs that now require experience nobody can get.
This is the AI-and-jobs story told honestly, and it's not the one we were braced for. The fear was mass layoffs, robots marching people out the door. What's actually happening is subtler and maybe sadder: not people getting fired, but fewer first jobs ever opening up, and the ones that remain asking for a resume you can't build without them. The bottom rung of the ladder is being quietly sawed off, which hits the young, who need that rung to climb. So what do you tell a twenty-two-year-old about that? I don't have a tidy answer. But if you've got a kid heading into the workforce, this is the trend I'd be paying attention to, not the sci-fi one.
Medicine and the Mind
4. The FDA approved a precision cell therapy that reengineers a cancer transplant
On June 30, the FDA approved Tregzi, from a company called Orca Bio, the first therapy of its kind, for adults with blood cancers getting a stem-cell transplant from a matched donor. The idea is precision. It uses a purified set of the donor's regulatory immune cells to cut down on one of the most feared transplant complications, graft-versus-host disease, where the new immune system turns on the patient's own body. In the phase 3 trial, 187 patients, 78 percent of the people who got Tregzi were alive and free of chronic graft-versus-host disease a year later, against 38 percent with a standard transplant. Overall survival was 94 percent against 83 percent.
The story of cancer treatment keeps moving in one direction, from the blunt instrument toward the scalpel. A traditional transplant is a sledgehammer, effective and brutal. Engineering the exact mix of cells you put back in, to keep the good effect and dial down the friendly fire, is the scalpel version. The real-world catch is the price and the plumbing. Orca set the list price at $428,000, and it's launching at a handful of transplant centers, aiming for about 25 by the end of the year. But the trajectory is the hopeful part. We keep learning to aim.
5. Scientists identified a brand-new way brain cells die in dementia
Another one from late June. On June 25, researchers at King's College London, working with the UK Dementia Research Institute, published a study in Nature Communications describing a previously unknown mechanism of brain-cell death, which they named karyoptosis. In it, toxic protein buildup causes a neuron's nucleus, the part that holds its DNA, to warp, shrivel, and then come apart. Analyzing about 3,000 brain cells from 28 patients who had either frontotemporal dementia or late-stage Alzheimer's, they found signs of the process in 35 percent of frontal cortex cells from the Alzheimer's brains, against 15 percent in healthy older ones. They also found a specific handle to grab: the interaction between an enzyme called p38 MAP kinase and a protein called LaminB1. Blocking it cut the karyoptosis markers in rat neurons in a dish.
I'll let myself take this one a little personally, because dementia isn't an abstraction in my world. I run memory care homes. I know its names and its faces, and I've watched what it does up close for years. And for all of that, the plain truth has always been that we understand the disease far less than families assume. So a study that maps out a new way the cells actually die, and points at something to block, is the kind of upstream science that eventually turns into a treatment that slows the thing down instead of just managing it. This is early, tissue and rat neurons, no human therapy yet. But you don't get to the treatment without first understanding the mechanism, and this week we understand one more piece of it than we did.
Money at the Half
6. The June jobs report came in weak, with people leaving the workforce
On July 2, the Bureau of Labor Statistics reported that employers added just 57,000 jobs in June, about half the 115,000 economists were looking for, and down from a revised 129,000 in May. Unemployment ticked down to 4.2 percent, which reads like good news and isn't. The rate fell because roughly 720,000 people left the labor force entirely, dropping participation to 61.5 percent, the lowest since March 2021. Leisure and hospitality lost 61,000 jobs. And the prior two months got revised down by a combined 74,000.
Here's what gets me about this one. The single number everybody quotes went the good direction, and it went there for the worst possible reason. Unemployment didn't fall because people found work. It fell because three quarters of a million of them stopped being counted as looking for it. That's not a recovery, it's a smaller denominator. Where did those people go? Nobody in the release can tell you, and that's the part that should bother us more than it does. It's why I've gotten allergic to anyone who reads me one number and tells me the economy is simply good or simply bad. A rate that improves because folks gave up is a different, quieter kind of trouble than layoffs. It doesn't make loud headlines. It just slowly drains the thing.
7. The stock market closed a wild first half up almost ten percent
On June 30, the S&P 500 finished the first half of the year up about 9.6 percent, which sounds calm until you look at how it got there. The index dropped about 8 percent from late February through the end of March, its worst month since May 2025, on an oil and inflation shock out of the Middle East that pushed Brent crude above $113 a barrel. Then it turned and ran all the way to record highs on the back of spending on AI infrastructure. Nine of the twelve biggest contributors to the gain were semiconductor or memory-chip stocks. Micron more than tripled.
So the retirement accounts look good at the halfway mark, and that's worth being glad about. But I'd read the fine print on my own before celebrating too hard, because a market that's up almost ten percent while leaning that heavily on a few chip names is a market with a short list of load-bearing beams. When the whole building rests on the same handful of columns, it's strong right up until one of them wobbles. I'm not predicting a fall. I'm just noting that the calm number and the concentrated engine underneath it are two different facts, and only one of them shows up on the statement.
Wonder and Play
8. The most ambitious sky survey ever attempted switched on
On June 30, the Vera C. Rubin Observatory in Chile began its main mission, a ten-year project to photograph the entire southern sky over and over, which the people running it are calling the greatest cosmic movie ever made. It uses the largest digital camera in the world, 3,200 megapixels, and takes a fresh image roughly every 40 seconds. Over ten years it'll look at each patch of sky about 800 times, which is what lets it catch things that change or move: exploding stars, feeding black holes, asteroids, and the invisible dark matter that holds galaxies together. The warm-up runs alone, before the survey even officially started, turned up more than 11,000 asteroids nobody had ever seen, including 33 near-Earth objects.
I love this one for the sheer patience of it. Ten years of the same telescope taking the same pictures of the same sky, just to catch what moves and what flares. That's the opposite of the everything-in-a-week pace I write about in half these stories. It's a decade-long act of attention, which is a thing I admire and am not always good at myself. Some of the biggest discoveries of the next ten years are going to come out of this quiet, repetitive, beautiful piece of work. Worth remembering that not all progress is fast.
9. Tesla posted record deliveries and snapped a two-year slump
On July 2, Tesla reported it delivered 480,126 vehicles in the second quarter, up 25 percent from a year earlier and about 74,000 above what Wall Street expected, beating even the most bullish analyst on the street. It was the company's best-ever second quarter, topping its 2023 peak after a long, rough stretch of falling sales. Its energy-storage business deployed 13.5 gigawatt-hours, up 40 percent from a year ago. And the stock fell 7 percent that day anyway, its worst day in nearly a year.
I've got a personal stake in this one now, since I finally became a Tesla owner myself a few weeks back after years of watching from the fence. So I read this with an owner's curiosity. The interesting question the numbers raise is why, after two years of decline, demand came back. Some of it is probably fresh product, some of it is people like me finally crossing over as the range and charging got good enough. I'll withhold judgment until the full financials come out later this month, because deliveries and profits aren't the same thing, and a market that sells off 7 percent on a record quarter is telling you it already knows that. But a genuine rebound at the biggest name in EVs says the shift to electric didn't stall out. It just caught its breath.
10. The World Cup hit the knockouts, and the home team went out
The World Cup reached its knockout rounds over the July 4th weekend, and it delivered the drama. On July 6, Belgium knocked the US men's team out with a 4 to 1 win, ending the host country's run. The day before, Norway stunned Brazil, one of the tournament favorites, 2 to 1. England edged Mexico 3 to 2 in a five-goal thriller in Mexico City. And earlier, both Germany and the Netherlands went out on penalty kicks, Germany to Paraguay and the Netherlands to Morocco.
The US going out at home stings a little, even for someone like me who only tunes into soccer every four years. There's a particular disappointment in hosting the party and getting sent home early from it. But that's also the honest charm of this tournament, the reason the whole world stops for it. Nobody is safe. Brazil can lose to Norway. Germany can lose to Paraguay, the Netherlands to Morocco, on nothing but a good day and a couple of penalty kicks. In a summer where so much of the news was about the powerful getting more powerful, there was something clarifying about watching a stage where, for 90 minutes, that counts for nothing.
A Note to Myself
That's the week, and the half. Ten things, one sitting.
Halfway through the year, the thread I keep pulling on is pace, and its opposite. Musk training a model a month. Half a trillion dollars moving in six months. And then a telescope settling in for a patient ten-year stare, and a soccer match where all the money in the world can't buy you the next 90 minutes. I spend a lot of my life feeling behind the speed of things. Weeks like this remind me that the fast stuff and the slow stuff are both real, and the trick is knowing which one a given moment actually calls for.
Mostly for me, like always. But if you read this far, thank you. See you next week.
