TL;DR: Prediction markets are booming, but the guardrails aren’t keeping up. After Kalshi took in a record $871 million in Super Bowl bets, Washington signaled it may let the industry shape its own oversight, continuing the long-standing US tradition of trusting tech companies to police their own behavior.

What happened: At the same time New Yorkers were lining up to enter the city’s first free Polymarket grocery store/PR stunt yesterday, the Commodity Futures Trading Commission—the federal agency that oversees prediction markets and derivatives—effectively handed the industry a seat at its own regulatory table. New CFTC Chairman Michael Selig announced an Innovation Advisory Committee tasked with examining how “breakthrough innovations” are “transforming markets.” In plain English: A group originally established to oversee cattle futures has decided it’s out of its league and needs online betting companies to write their own playbook.

The 35-member panel is stacked with insiders: Kalshi CEO Tarek Mansour, Polymarket CEO Shayne Coplan, plus top execs from FanDuel, DraftKings, Robinhood, and Coinbase. Noticeably missing? Any industry skeptics.

How we got here: In 2018, online sports betting was illegal almost everywhere. Today, 90% of sports bets are placed on phones. Prediction markets saw how much people liked wagering on hyperspecific outcomes—like whether LeBron would hit a three-pointer at exactly 12:37 in the first quarter—and stretched that model to everything: politics, celebrity gossip, war, and plenty of tech product launches. The typical user is a young man; one recent analysis claimed that prediction market traders lose money at a faster rate than traditional gamblers do.

The house might have an edge: With more money and users flowing in, so is suspicion that the game is rigged. During the Super Bowl, more than $47 million was wagered on halftime show bets. The next morning, Kalshi’s CEO was asked on CNBC whether a backup dancer betting on the opening song would count as insider trading. His answer: If the information is considered “nonmaterial” and socially acceptable to share, it’s “totally fair game.”

That ambiguity (read: sketchiness) gets at the heart of prediction markets. Unlike securities law, CFTC anti-fraud rules typically require deception or breach of duty, a higher bar than the Securities and Exchange Commission’s. That means people can often bet on outcomes they have advance knowledge of. One Polymarket trader reportedly made nearly $410,000 betting on Nicolás Maduro’s capture hours before it was announced. Was the bettor lucky or a government employee with prior knowledge?

Not everyone’s buying in: Even as the industry surges, unease is growing. Among men under 30, the share who say legal sports betting is bad for society has more than doubled in three years—from 22% to 47%. Meanwhile, at least eight states have taken action against prediction markets, issuing cease and desist letters, filing lawsuits, or opening investigations. But until anything changes, people will bet more, lose more, and feel worse about it—which, so far, appears to be part of the business model. —WK

When your smart shades get in the way, knock

Ever tried to enter a sliding door from the outside only to realize your electric shades are pulled down on the inside, leaving you fumbling around trying to figure out how to manually lift them without breaking anything? Tech Brew reader Mike from Lake Havasu City, Arizona, has—and he has a techy solution for this problem.

“We use SmartThings for our home automation, so I was able to put a vibration sensor on the glass sliding door,” he says. “A knock on the door triggers the shade to open.”

How it works: Samsung’s SmartThings is a smart home system that connects, monitors, and controls a variety of smart devices and appliances, all centered on a free app you can download to your iPhone or Android. It lets you set up “scenes” to control multiple devices at the same time. Plus, it’s compatible with hundreds of brands and can connect to most smart devices via wi-fi.

The setup: Open the app and click the + icon → select Add device → tap By brand or Scan nearby → select your device brand or scan its QR code and follow the prompts. Once you’ve added your vibration sensor, repeat the steps to add your shades. To get your sensor and shades talking to each other, go to the Automations tab and create a new Routine. For a vibration sensor like Mike’s, you can set an If trigger by selecting the device and setting it to detect a knock. Next, set a Then action by selecting your shades and setting them to open.

The caveat: Certain brands of sensors and shades, like Zigbee and Z-Wave, do not use wi-fi and require a physical hub to act as a bridge, like the Aeotec Smart Home Hub or Samsung SmartThings Station. So, this Life Hack could get a bit expensive depending on the brands you prefer.

No more standing in the dark trying to circumvent your shades after popping out for a late night snack or stroll. Instead, a simple knock will clear the way, which, as Mike puts it, is “pretty cool.” —CM

If you have a tech tip or life hack you just can’t live without, fill out this form and you may see it featured in a future edition.

ChatGPT can ace elite math competitions—but ask how to get your car to a car wash 50 meters away, and OpenAI’s flagship model suggests you… walk.

The chatbot’s answer is amusing, confusing, and also a little vindicating at a time when there's been a lot of AI-will-replace-us handwringing. GPT-5.2 Thinking, the advanced reasoning model for complex tasks, gives several reasons: Walking is more fuel efficient. It would only take one minute. And given you're about to wash the car, driving it would just add 50 meters' worth of dirt again. Solid logic, no flaws whatsoever.

After hitting ChatGPT with the "are you sure?" follow-up—which often gets AI models to change their mind—the model clarifies: Walk over to scout the car wash, then head back and drive the car. And if you ask whether it's normal to carefully scope out a car wash on foot instead of peering out your window or simply googling if it’s open, ChatGPT insists that, actually, its thinking is "slightly more rational than average." In our testing, Anthropic's Sonnet 4.5 model also suggested walking, while Opus 4.6, Grok, and Gemini 3 Thinking all tell you to turn on the engine.

Plenty of other AI "gotchas" have stumped models in the past, like counting the number of r’s in "strawberry." These riddles expose AI’s core limitation (besides spelling): Models optimize for internally consistent answers, not real-world practicality. Even as AI surpasses humans on formal benchmarks, common sense remains harder to replicate for AI models and your ex, alike. —WK