tash_2s

I also ended up using max effort/reasoning for both coding and general chat. They don't spend too much extra time on simple tasks these days.

Love this. I hope it works with Japanese kanji too, because sometimes I forget the exact character but remember a similar one.

Yeah, having a system that understands the situation and intervenes could be useful in a lot of cases.

I feel the same. As AI gets better, I think personal AI for the physical world will become much more useful, and glasses could be a natural interface for it.

Thanks! I need to test it in low-light environments, and making it work reliably across a variety of kitchen settings is definitely one of my goals.

Totally. I think this kind of thing sits right on that line: it can help someone (hands-free guidance, training, accessibility, staying in flow), and it can also slide into a pretty dystopian "score the worker" surveillance HUD. My intent here is the former: personal "real-world speedrun" / practice tooling, not a manager dashboard or productivity policing.

Cooking feels like a perfect fit for smart glasses (hands busy, lots of short steps), but I have not seen many apps that work reliably in a real kitchen. It feels like the hardware is finally getting to the point where this should become practical soon.

Right now it runs on Rokid Glasses using the OpenAI Realtime API.

I'll take another look at EgoBlur to see if it's a good fit, thanks. When I checked it briefly before, I thought it was focused on post-processing (non-realtime) and didn't integrate with matching easily, but it's definitely solid tech.

When multiple people are in view and the system detects consent, the current implementation assumes the person closest to the camera is the one giving it. This is not ideal, so active speaker detection is planned.

Verbal consent is just one example. Depending on the situation, other interfaces may work better, such as having a predefined list of friends who are always consented.

With proper implementation, on-device processing on a smartphone is feasible. On-glasses processing would be challenging, especially with battery constraints.

The 720p 30fps figure is from a PoC implementation, so there is still significant room for improvement. And yes, the demo is on an Apple Silicon M2.

Thanks! Beyond face blurring, I'm planning to detect and redact other private information like license plates, name tags, and sensitive documents.

Most of these features are aimed at protecting bystanders, but I'm also interested in exploring privacy protection for the wearer. For example, automatically shutting off recording in bathrooms, or removing identifiable landmarks from the video when they could reveal the wearer's location, depending on the use case.

I think mainstream adoption of smart glasses could be slowed more by privacy concerns than by hardware limitations. Remember Google Glass? While the hardware keeps improving, I want to make sure we're also addressing the software side.

Sounds great, love the idea. Since it's based on a single prompt, can you manually edit the generated flow afterward?

This is an indie open-source blockchain game, running on-chain, led by the community.

The game is coded in Rust, compiled to Wasm. Both the blockchain and its web frontend use that.

Most "blockchain games" use closed-source centralized servers. They just change some in-game assets into NFTs, and that's all. On the other hand, this open-source project fully runs in the standalone blockchain and is led by the community.