OpenAI's Strongest Model Is Here. Washington Capped Who Can Use It.

Claude vs Gemini. GPT-7 vs Llama 5. Which AI lab ships AGI first. These are live Kalshi markets with real money on both sides, updated in real time as releases land. The person who follows model cards and tracks evals has a genuine edge here. If that's you, trade it.

Trade the model race

3 things that happened while you were busy

1. OpenAI launches its GPT-5.6 series, but only to a chosen few.

The new GPT-5.6 series includes Sol, OpenAI's strongest model yet, Terra, a cheaper daily driver, and Luna, a budget option. At the US government's request, the company is opening access to a small circle of partners first and framing the slow rollout as a safety measure. It mirrors the path of Anthropic's Mythos, which is back in limited preview, with Fable rumored to follow this week.

2. Chinese models keep closing the gap with America's top labs.

Tulongfeng, a new model from the Chinese cybersecurity firm 360 Security Technology, was billed by its founder as a Chinese answer to Mythos for spotting software bugs. Z.ai's GLM-5.2 is piling on more pressure, matching the performance of top US models at a fraction of the price. Read the details.

3. Google starts rationing Meta's access to Gemini.

Google has capped how much Meta can use its Gemini models, blaming a shortage of compute. The crunch is severe enough that Google agreed to pay SpaceX $920 million a month for access to 110,000 Nvidia GPUs. It is a sign that the AI infrastructure buildout still has not kept pace with demand, and that the compute shortage is starting to reshape which companies work together.

AI is minting winners and gutting losers in the same stock market.

Memory chips surge. Semiconductor and memory chip firms have barely been able to lose this year. Last week Micron, up 296 percent on the year, said quarterly revenue quadrupled to nearly $42 billion. SanDisk, up 780 percent, watched last quarter's sales jump 251 percent to $6 billion, and SK Hynix, up 294 percent, has grown so fast it is now South Korea's most valuable company. These names have been the exception, though, while many marquee tech stocks sit far below their highs.

Software's bad year continues. Earlier in 2026, fears that AI would gut enterprise software, a scenario some called the SaaSpocalypse, sent Oracle, Salesforce, and ServiceNow tumbling.

Months on, all three are still down more than 50 percent from their highs, though the White House move to restrict OpenAI's GPT-5.6 models sparked a sharp bounce last Friday.

Big Tech stumbles. The giants have not been spared. Meta is down 29 percent, Amazon 15 percent, Alphabet 16 percent, and Microsoft 28 percent from recent highs. One big reason is spending. Together these companies plan to pour roughly $725 billion into AI infrastructure this year, an increasing share of it funded by debt. Investors have pulled back until they see clearer proof the spending is paying off.

But the market is green. For all the red, the S&P 500, Dow Jones, and Nasdaq are still up 7 to 9 percent on the year, nowhere near recession territory. Even so, expect volatility to be the defining theme for the rest of 2026, when a single model launch, feature, or leaked memo can swing the market hard in either direction.

What people are actually watching and sharing

Skill upgrading. A viral post lays out how to build a personal productivity system around Claude Cowork, so the setup handles your daily production while you focus on the decisions. Steps and prompts are included, and it has pulled in 4.5M views.

Historical figures. Ever wonder what Socrates, Mozart, or Cleopatra would look like today? A set of viral AI images reimagines them in the modern day, and the thread has 1.5K upvotes.

Token efficiency. Coinbase CEO Brian Armstrong shared how the company keeps its AI bill flat even while burning through far more tokens. His main tactics: defaulting to open-weight models, smarter routing, and keeping context lean. The post has 3M views.

PDF viewer. A Redditor used Claude to build a small tool that makes navigating PDFs on a computer far easier, and commenters say it should have been the default years ago. You can try it here.

Heat wave. Just how brutal is the record heat dome sitting over Europe right now? A batch of AI-generated videos captures how extreme it feels, and one has 1M views.

Turn ChatGPT into a PhD-level researcher on any topic

When you want more than a surface-level answer, this prompt makes ChatGPT reason like an academic. It weighs competing theories, checks the evidence on each side, flags the open questions, and tells you where a field is still unsettled instead of handing you one confident take. It is handy for real research, prepping a debate, or pressure-testing an idea before you commit to it.

How to use it. Swap [TOPIC] for whatever you are digging into, like remote work productivity, the history of jazz, or the future of nuclear energy. To go deeper, ask it to cite the strongest study on each side, or to name the researchers most associated with each viewpoint so you know where to read next.