Joon

Finding files on Windows might get a bit easier, as you might not have to open every search result one by one. It looks like they are expanding the search bar to show file previews, modified dates, and file locations.

According to the Windows Insider announcement on October 30th, they are also adding sliders in the search bar to adjust volume and screen brightness. That should save some time toggling back and forth between settings.

It's currently only available in the Experimental channel, so it's not on all Windows 11 devices yet. It seems like it'll be really convenient to verify you've found the right file directly in the search bar.

When AI finds travel products, instead of a long response, it displays cards you can select immediately. On October 29th, the Google Developers blog introduced an example connecting A2UI with CopilotKit.

The AI sends the UI layout as data, and the app renders the screen using pre-built buttons or cards. From there, when you select a date or flight, the chosen value is passed back to the agent to trigger the next screen.

However, developers do need to pre-define the available UI components. It feels like you can keep the process in one window rather than receiving a chat response and having to go to a separate page to make selections, so I'd love to try it out.

I wish it would remember the last folder I visited. It’s pretty discouraging to have to start searching from the wrong place every time I come to upload a file.

Firefox 159, released on the 27th, has been fixed to open the last folder used for uploading rather than just the last one accessed. Choosing files should be this effortless.

If you are learning for the first time, I recommend defining 'what constitutes a finished product' before asking for code.

For example, if you're building a household ledger, decide how it should handle empty amount fields or double-clicking the save button. If you set these requirements first, you'll have a standard to check against—whether the code is written by AI or yourself. You can then approach learning syntax as needed to implement and refine those specific conditions.

The KAIST robot demonstration at the parliamentary audit just moonwalked out on its way off stage lol

I find it hard to feel at ease just because iris authentication has passed.

For instance, if someone is actually trying to send money because they believe an urgent request from someone else, even if the identity check is done, the question of whether that request is legitimate still remains.

I wonder if verifying identity is the only thing banks should be asking first. For an unusual transaction, a moment to pause and think about who and why you are sending money to might be more helpful.

That said, if they pop up a ton of warning messages, even I would just skip through them. I care more about what is said and at what moment rather than the number of times it appears.

Even if they consider introducing iris authentication, I'd like to see it paired with a mechanism that makes you pause and rethink the transaction. I believe confirming your identity and deciding whether it is safe to send money should be kept separate.

Claude Duet has started its research preview in the US. They say it needs approval before sending an email, but I wonder if the recipient and attachments will show up on that screen as well.

Anthropic's public benefit fund announcement includes a condition requiring shareholder approval for the 5% equity pledge. If they are helping people whose work is being transformed by AI, I wonder what the first round of support will look like...