October 1, 2026
OpenAI's latest ChatGPT update, reported by TechCrunch, lets users upload a photo and virtually try on clothes and accessories, then save favorites for later. On its face this is a convenience feature. Underneath it, it's a land grab for the moment of purchase itself. Every time a shopper tries on a jacket inside ChatGPT instead of on a retailer's site, OpenAI gets closer to owning the transaction, not just the conversation leading up to it. That matters for any business that sells online, because the companies that control checkout increasingly control the customer relationship. We've written before about how Google, Meta, and now OpenAI are quietly racing to own checkout, and this launch is another data point in that trend. If you run an e-commerce brand, the real question isn't whether this feature is neat -- it's whether you want a chatbot sitting between you and your customer's wallet, and what that does to your margins and your data over time.
Google has launched an advanced chip into orbit as an early step toward space-based data centers, and TechCrunch reports the company's own estimate is that SpaceX's Starship would need roughly 1,600 launches before the concept becomes viable at scale. That number is the real story here. It's not a quiet caveat, it's the headline math, and it tells you this is a decades-long bet, not a near-term product roadmap. For most businesses evaluating AI infrastructure today, orbital compute is irrelevant -- the constraints you actually face are power, water, and land for terrestrial data centers, which is why every hyperscaler is scrambling for energy deals right now. I think this experiment is worth doing, because someone needs to stress-test the idea before capacity on Earth truly runs out. But I'd be skeptical of anyone using it to argue compute scarcity is about to be solved. The practical lesson for buyers: don't let moonshot headlines distract you from asking vendors the boring question of where their current compute actually comes from and how stable that supply is.
The Wall Street Journal reported, via TechCrunch, that OpenAI has cut ties with three safety researchers after an internal investigation concluded they mishandled sensitive company information. That's a different kind of departure than the usual story of safety staff quitting in protest over product speed -- this is OpenAI asserting it enforced its own internal controls. Both readings can be true at once: a company can have a real information-security problem among safety staff and also have a pattern of safety-minded people leaving under pressure. We've covered that pattern before in OpenAI's apology tour and the price of moving fast, and in how OpenAI's rogue agents say the quiet part about internal tension. For enterprise buyers, the takeaway isn't to panic about any single personnel story -- it's to keep asking vendors concrete questions about data handling and access controls rather than taking safety messaging at face value. Review the Security page of any vendor you're evaluating, including ours, before you sign.
Which of these three stories actually changes a decision you're making right now -- your e-commerce strategy, your infrastructure bets, or your trust in a particular AI vendor?
Sources