Author: Jezca Felarca

ChatGPT still doesn’t give you a built-in way to import or export its memory features, so if you want to move saved context between chatbot services, you’re still doing it by hand. In the current settings, you can switch memory on, check your Memory Summary, delete stored items, and manually copy details like your writing preferences, recurring projects, favorite formats, personal interests, tone, job role, goals, and ongoing tasks into another chatbot.

ChatGPT Download

What ChatGPT still won’t let you do is officially export a tidy memory profile, or bring that profile back into ChatGPT with a single click. If remembered context is part of how you work every day, that gap is a real limitation.

ChatGPT memory also isn’t the same as your full chat history. The email export you can request from ChatGPT usually arrives as a broader JSON or HTML archive that includes account data, not a memory-only file. Anthropic’s Claude and Google’s Gemini already have memory-related import workflows, and Claude now includes a built-in tool that can pull memories in from other providers.

You should still check anything you move over carefully. Some remembered items may have been generated by the system, and exported data can include sensitive personal or business information. In ChatGPT, you can view your Memory Summary in the app, and on supported platforms, you can also request a broader account data export by email.

Android 17 and Wear OS 7 are out, and if you’re using a supported Google Pixel device, you can install them now instead of waiting for other Android brands to get them sometime in 2026. Google is also rolling out its June 2026 Pixel Drop alongside both updates.

For Android 17, Google says the big additions include Bubbles floating windows for easier multitasking, Screen Reactions so you can record your screen and your selfie camera at the same time, and a 50/50 split-screen gaming mode built for foldables. The update also expands parental controls, adds one-time precise location access, lets apps ask for permission to only selected contacts, and puts biometric protection on the Mark as lost option in Google Find Hub.

Wear OS 7 has a shorter feature list, but there’s still plenty in it. Google says it brings Live Updates to the watch face, improves the controls you can use from your wrist, and delivers up to 10% better battery life than Wear OS 6.

The June 2026 Pixel Drop adds a separate batch of features: Gemini Omni for video creation, Lyria 3 for music generation, Take a Message, and Quick Share with AirDrop compatibility on the Google Pixel 8a and Google Pixel 9a.

Some things still aren’t here yet. Google says multi-step app automation and Create My Widget are expected later this summer, and a few features will stay limited to premium devices.

If you use a Pixel every day, this update is worth grabbing. Some features are still coming later, and others won’t reach lower-end devices or non-Pixel phones.

You can download the updates now on supported Google Pixel devices, while the broader Android and Wear OS rollout continues through 2026.

OpenAI just rolled out something called Deployment Simulation ahead of the GPT-5 family launch. It’s a pre-release testing method meant to answer a pretty practical question: once a model is actually out in the world, how often is it likely to mess up?

ChatGPT Download

The setup is different from the usual synthetic-prompt testing. OpenAI takes real, anonymized chats from an older model, replays them, and has the unreleased model write the next reply. According to the company, that approach got the direction of error trends right 92% of the time.

The idea is also supposed to cut down on the distortion you get when a model can, in effect, tell it’s being evaluated. Using de-identified conversation logs means the test is built around the kind of messy, unpredictable prompts people really type. OpenAI says that helped surface issues it hadn’t seen before, including something it calls “calculator hacking,” and shifted the emphasis toward how often failures are likely to happen in real use, not just whether a failure can happen in theory.

If you pay attention to AI safety, this one’s worth watching.

OpenAI is clear about the limits, too. Deployment Simulation isn’t meant to replace red-teaming or targeted evaluations, and the company says it should sit alongside both. Rare failures can still slip through. So can new attack techniques and odd behavior that doesn’t show up often enough to get caught this way. OpenAI researchers also ran the method on the public WildChat dataset so outside auditors would have a version they could use without private logs, though OpenAI says it’s probably less accurate than the internal-data version, especially as pressure keeps building from the US National Institute of Standards and Technology (NIST), new European Union rules, and safety institutes in other countries.

The full research paper is up on OpenAI’s website.