cloudking

My guess, lots of user data + user corrections. When we correct what our phones mishear, that data helps train the speech recognition models to get more accurate over time.

How would you implement your solution technically though? There isn't a reference set of images in the model, it's just weights.

I'm pretty happy with https://www.mockplus.com

I bet it likes salt & vinegar microchips

Yeah this is the reason. I want a solution that works from any browser/device without the need to use extensions.

I'd settle for that too!

Idea: 1000 karma should get custom CSS as a setting.

I was going to say, has OP downloaded a game recently? You can fit 10-20 modern games on this card tops.

It's great that there are many comments about being able to run models locally, but I rarely see comments about what folks are using them for. Would love to learn more about use cases and problems being solved.

Wonderful hack, the overall response latency is the only thing that hurts the UX, if you can get the response time down would be epic. Nice work.

In my experience in multiple roles (product manager, project manager, software engineer) and dozens of successful launches with large engineering teams, the ideal kind of PM has a deep understanding of users, how software should work best for them AND technical experience.

The problem with a purely technical team and no user-centric PM guidance is they'll end up building an engineering designed product, that doesn't resonate with non-technical users. As pure engineers, we love engineering, and may over-engineer solutions given the opportunity. Unless we are building some developer tools or highly technical product, this doesn't work for the majority of non-technical products.

You need a PM who can separate the technical design from the user experience, but also understands how to bridge the two when it comes to actually building the product. No technical experience rarely helps, as you spend a lot of wasted time debating why things should or can't be built.

As a PM with technical background, I'm not going to waste engineering time proposing radical user experiences that I know will require excess resources and time to build. At the same time, deeply understanding what users want, I'm not going to build a product that is too technical for them to understand how to use. We understand the limits of our technical stack and how to reduce user frustration. That is the advantage of having a balanced PM with both user and technical experience.

+1 can't access any Discord servers

This is most likely the reason, and the teams involved in launch approvals (legal, safety, privacy etc).

Probably because they know more about you on a personal level than the other services do. A lot of people have Facebook and Instagram accounts using their real names, and real interests linked to their real life friends. They know what you're talking about with each other and they know what you're doing on the web through their tracking pixels.

I wonder what Apple will do with the number registrations that came from Androids.

When can we send our gaussian heads to the Zoom meeting?

Kudos for the landing page, this is how AI products should be presented.

1) Direct access to the product

2) Multiple demo examples

I've seen so many launches here recently that have no demos, require sign up and basically gate all the usefulness of the product behind walls.

+1 using AdGuard on Home Assistant Raspberry Pi

Motion features are the worst, I find they always make shows and movies look uncanny. Have turned this off on many friends and family devices, who seem to not notice until I turn it off.

How does Huggingface work from a business model perspective? They let people use a lot of free compute resources running models on their platform, surely that costs them money?

It's a very cool technical feat, but not something I would personally pay for. I'll just spend the 1-2 minutes to watch the video for free. Not trying to discourage you, just giving honest feedback. Launching the early landing page is a good idea to validate further.

I hope they upgrade Duet AI to use this model then, because the current models in Google Workspace are quite lackluster. On the positive side, the UX is integrated quite smoothly.

This is the Unifi device hackers can recommend to their friends and family, but not actually use themselves because they already have the more sophisticated set ups.

It looks like it can connect to local models via LiteLLM proxy.

Related issue: https://github.com/danny-avila/LibreChat/issues/1131

LiteLLM: https://github.com/BerriAI/litellm

Guide: https://github.com/danny-avila/LibreChat/blob/main/docs/inst...

What are some good alternatives to Google Shopping? I'm looking for a meta search that will find the best deals across the web, but from legit sites only.

This is cool! I've also found GPT-4 is really good at generating complex CSS animations

Nice, thanks for the suggestion!

Speaking of Zoom. Are there any mics or webcam/mic combos that have really good directional noise cancellation to only pick up sound from the person in front of them? I'd like to stop using Bluetooth earbuds (best solution I've found), but need only my voice to be picked up.

GitHub Co-pilot can take into context your codebase.

Awesome work! You should break down the development in a blog post