rememberlenny

People don’t realize how crucial Ben was in the forming of OpenAI as it is known today. This is an extremely underrated post.

Congrats. This is very well executed.

This is amazing! Congrats on the launch!

Pro tip: Read the actual court decision and not just the headlines.

Tech law journalism is a telephone game that usually distorts what was actually said. People then form strong opinions on the headlines. Chaos ensues.

Example: “AI art cannot be copyrighted - US Federal Judge”

That’s not what the judge decided. The decision said you can’t assign a copyright to an AI. This doesn’t mean the thing you create with AI can’t be copyrighted. You write a prompt, press a button, cause something to be created, etc., and you’re the author not the AI.

https://www.documentcloud.org/documents/23919666-thalervperl...

From @AndrewMayne on Twitter

Love this.

Its the older completion models, not the older chat completion models.

In March, I was confident edge level usage of large models wouldn’t be possible soon.

Distilling large models into smaller parts, has gotten better than I anticipated.

In case it's not obvious, this is a concrete departure from rudimentary "prompt engineering".

Dialog-based interfaces, such as a context-aware conversation, are better at conveying human intent and provide a more natural way to interact with model capabilities.

With all the excitement around Stable Diffusion, it’s nice to have an easy API to plug into for product development.

The API supports three endpoints (generation, edits, and variation), but doesn’t support fine-tunes - which are capturing peoples attention now with deepfake avatars.

Im curious to see what people build with this.

Here is the prompt for "A still of Kermit The Frog in Dragon Ball Z (1989)"

https://user-images.githubusercontent.com/1332366/171921054-...

The reason we made this is because you often need to juggle multiple tools when doing anything in video. Rather then using one tool for video collection, another tool for editing, and another for sharing clips - we wanted to have one tool for everything.

I believe we are the only video collection tool that also lets users design around the collected video clips in a free-form drag-and-drop editor and share out final videos.

Sharing the latest iteration of Milk Video with HN.

Milk Video makes it easy to gather video testimonials, edit and share them in one place. We created a robust browser based video composition tool for teams that produce webinars, and recently added a feature to gather video testimonials via a link.

The overall application runs via a React application with a Ruby on Rails API. To gather video recordings, we use Daily.co as a recorder and stream the recordings using Mux. To composed new video with the collected recordings, we made a video rendering engine that breaks the source video material into frames, and screenshots individual frames of the composed final video. Once the frames are created, we reconstruct the video and add the audio tracks on for the final product.

If you have anyone in a Marketing team at the company you work at, I'd greatly appreciate sharing it with them and seeing what they think!

Thank you!

Porter customer support is insanely good. If you are trying to migrate from Heroku to AWS, they make it incredibly easy and go out of their way to engineer alongside you.

I was deeply impressed as a user and have nothing but positive things to say about their team.

This looks very nice and the YouTube/Loom integration is very nice.

What target user group are you aiming to serve? Given the number of video tools in the past two years, I find that the vertical served largely determines what features are most valuable.

This is not to be mistaken with the company https://cohere.io/ that lets you monitor your web users in real-time.

Can I run this in my Google glass to preemptively detect Rickrolling?

Thank you Mark for your selfless work to push this project forward. RTK Toolkit was an absolute game changer and has been a foundation that our product would suffer immensely without.

Classic. I remember this from middle school.

Delayed response here.

I was trying to count the number of unique graffiti pieces on trains across the US, and measure the distance individual paintings traveled, and the duration they lived before being painted over.

This is timely.

I just compiled a few resources I found useful on groking computer vision and deep learning.

Short list I’ve enjoyed recently:

- PyImageSearch blog: https://pyimagesearch.com/blog/

- Fastai library and course: https://www.fast.ai/

- Yan LeCun’s 2021 Spring NYU class: https://m.youtube.com/playlist?list=PLLHTzKZzVU9e6xUfG10TkTW...

We did something very similar here, using computer vision, to track video feeds of moving trains containing graffiti. Much more complicated in one sense, but also much simpler.

I love this kind of digital collaboration with the humanities.

The quoted StackOverflow post is gold:

"A mass-murderer space rock happens about once every 30 million years on average. This leads to a probability of such an event occurring in the next second to about 10-15. That's 45 orders of magnitude more probable than the SHA-256 collision."

Link: https://stackoverflow.com/questions/4014090/is-it-safe-to-ig...

Very neat primitive. Congrats on getting it out!

This feels like the missing tool in our remote team workflow. I love that you recognize there is a need for proper syncable docs that help manage project luck-off/monitoring processes.

Can you share a bit about your upcoming prioritized integrations?

Tuple is hands down the best pairing software.

We use it because the latency is the best compared to other pairing tools, and it has the best UX for remote pairing.

We even use it for pairing when we have people in the same room, and we need two people on their own machines.

The playground examples are very neat. Great job!

Important note -

I expected at most 100 people to try this out today, but we had over 3000 new users sign up.

It’s possible our download rendering is not working as desired, so I will revisit the core issues and fix accordingly!

Thanks for taking the time to try it out!

I’m noting these and will fix these respectively.

I really appreciate your taking the time and sharing this.

If you’d be open to chatting, I’d appreciate any more you’d be open to sharing: https://calendly.com/rememberlenny/15-min

Awesome!

Up to chat?

https://calendly.com/rememberlenny/15-min