Table of Contents
Quick heads-up—this week’s big story isn’t about a flashy new model release. It’s about privacy, and it actually affects how you should think about “share” features in ChatGPT.
OpenAI disables a ChatGPT sharing feature after private chat content showed up in Google results. Yeah—this is the kind of headline that makes you immediately wonder: “If I shared something, could it end up searchable?”
- OpenAI Shuts Down ChatGPT Sharing
- According to reporting shared on X, OpenAI disabled a popular ChatGPT sharing capability after it unintentionally exposed private conversations via Google search results. The core issue wasn’t that OpenAI “leaked” data out of nowhere—it was that some shared content was discoverable in a way it shouldn’t have been.
- What stood out to me is how fast these things snowball once search engines index pages. If something becomes publicly crawlable, you can’t really “un-ring the bell.” Even if OpenAI removes the shared items afterward, caches, third-party indexing, and people who already saw the content can keep it circulating.
- What this likely means in practice:
- Sharing ≠ guaranteed private. Even if you think you’re only sharing with a person (or a small group), you should assume search indexing is the risk scenario.
- Google indexing is the multiplier. Once content is indexed, it can surface in future searches even after the original page changes.
- Safety settings matter. If there are toggles around visibility, indexing, or “public link” behavior, treat them like they’re real-world publication buttons.
- Action you can take today:
- If you’ve ever used ChatGPT’s share/public link option, double-check that you’re comfortable with what’s in those chats (names, emails, phone numbers, company info—anything identifying).
- Be cautious about sharing drafts that include sensitive details. I try to keep personal identifiers out of prompts when I’m using shareable outputs.
- Assume anything shareable could become searchable. That’s not paranoia—it’s just how the web works.
- Source: OpenAI Shuts Down ChatGPT Sharing (X post)
Note: This article is based on the linked report. If you want the most authoritative wording (exact dates, the precise feature name, and what OpenAI said about remediation), check OpenAI’s official status/communications when they publish updates.
- Gemini 2.5 Deep Think
- Google’s Gemini 2.5 Deep Think is positioned as a “more deliberate” model that uses multiple internal agents to work through a problem—basically, it’s designed to spend more effort on reasoning instead of rushing to an answer.
- From Google’s announcement, the big practical differences are:
- Better at multi-step tasks like coding help, structured research, and tougher math problems.
- Slower than quick chat (in my experience with “deep reasoning” modes, you usually trade speed for thoughtfulness).
- Availability is currently limited—Google mentions it’s for Ultra subscribers, with API access coming soon.
- If you’re deciding whether it’s worth it, ask yourself: do you need fast answers, or do you need fewer mistakes and more careful reasoning?
- Source: Google Blog: Gemini 2.5 Deep Think
-
Reddit’s Revenue Surges
TechCrunch reports that Reddit’s revenue climbed sharply, landing at almost $500 million for the quarter discussed in the piece. The coverage ties the rise to advertising momentum and highlights how AI is being used in the advertising stack.
- Here’s the part I’d pay attention to if you’re an advertiser or marketer: it’s not that AI “creates money” by itself. It’s that AI can help with targeting, creative iteration, and matching ads to what users are actually discussing.
- Why it matters: If Reddit is seeing results, other platforms will copy the same playbook—meaning AI-driven ad optimization is likely to keep spreading.
I’m not going to pretend every new AI app is amazing. So I pick tools based on one simple thing: do they actually help with a specific task? Here are a few that look useful on day one.
- TextWisely– I’d use this when I’m polishing messages and want quick fixes. It helps correct mistakes, rewrite for clarity, summarize long text, and even translate—handy if you’re sending emails or drafting posts without wanting to switch tools every 5 minutes.
- OdysseyGPT– This is the kind of tool I’d reach for when I have a pile of notes or a document and I want it turned into something I can actually navigate—like a “knowledge center” with organized info and answers. Great for research projects where you keep circling back to the same sources.
- StylePal– If you’re stuck staring at your closet, this sounds like it could help. It’s built for style ratings and outfit suggestions, and that “regardless of the event” angle is useful when you don’t want to think too hard about where you’re going—you just want options.
- HumanTone– I’d try this for turning stiff drafts into something that reads more naturally. It also mentions SEO-friendly output, so it’s a candidate when you’re writing for search and want the tone to feel less robotic.
- AIVocal– This one’s for anyone turning scripts into voiceovers. Multiple voices and languages are the obvious selling points, and it’s exactly the kind of tool you’d use for podcasts, short videos, or localized ads.
- Enclave AI– I’m interested in tools that reduce friction. If it really supports “no registration” and straightforward privacy controls, it’s the kind of app you’d open quickly when you just want to talk without signing up for yet another account.
- Slax Note– For meetings or lectures, transcription-to-notes is the whole point. If it turns long sessions into simple notes you can share, that’s a practical workflow upgrade—especially when you’re trying to capture action items.
- iColoring AI– This is more “fun” than “work,” but it’s still a real use case: converting photos or text into coloring sheets and letting you add color. I could see it being useful for kids’ activities or quick creative projects at home.
Today’s prompt:
"Write a practical privacy checklist for using AI tools safely. Cover what to avoid sharing in prompts, how to think about public links/indexing/search exposure, and a step-by-step routine someone can follow before they hit ‘share.’ Include a short example prompt that shows what NOT to include (like names, emails, or internal project details) and a revised version that’s safer."
