rememberlenny

I would strongly advise against using Google's ML apis.

First, at my company Milk Video, we are huge fans of Assembly AI. The quality, speed and cost of their transcription is galaxies beyond the competition.

Having worked in machine learning focused companies for a few years, I have been researching this exact question. I'm curious how I can better forecast the amount of ML talent I should expect to build into our team (we are a seed stage company), and how much I can confidently outsource to best-in-class.

A lot of the ML services we use now are utilities that we don't want to manage (speech-to-text, video content processing, etc), and also want to see improve. We took a lot of time to decide who we outsource these things to, like working with AssemblyAI, because we were very conscious of the pace of improvement in speech-to-text quality.

When we were comparing products, the most important questions were:

1. How accurate is the speech-to-text API

1.a Word error rate

1.b Time attributed to start/end word

2. How fast does it process our content

3. How much does it cost

AssemblyAI was the only tool that used modern web patterns (ie. not Googles horrible API or other non-tech based companies trying to provide transcript services) that made it easy to integrate with in a short Sunday morning. The API is also surprisingly better than other speech-to-text services, because its trained for the kind of audio/video content being produced today (instead of old call center data, or perfect audio from studio-grade media).

Google's api forced you to manage your asset hosting in GCP, handle tons of unnecessary configuration around auth/file access/identity, and its insanely slow/inaccurate. Some other transcription services we used were embarrassingly horrible from a developer experience perspective, in that they also required you to actually talk to a person before giving you access.

The reason Assembly is so great is that you can literally make an API request with a media file url (video or audio), and boom, you get a nice intuitive JSON formatted transcript response. You can also add params to get speakers, get topic analysis, personal information detection, and it's just a matter of changing the payload in the first API request.

I'm very passionate about this because I spent so much time fighting previously implemented transcript services, and want to help anyone avoid the pain because Assembly really does it correctly.

It uses your user's Twitter account to make queries to avoid rate limits. Very normal. Also works really nicely.

While only slightly relevant to the content of the article, the author has an incredible tweet thread.

"Taylor Swift as classic programming textbooks, a thread."

https://twitter.com/jeanqasaur/status/1290883041418649600?s=...

Thank you for the clear explanation.

Could someone explain the significance of this, why it took so long, and what it opens open?

For anyone interested in raising money soon, but not immediately, the best thing you can do is start an weekly update email.

Start sending thing to all the people you interview for advice, supporters you know in the same space, and peers/coworkers. If you keep a regular email showing your progress, even though you may not have traction, potential investors will see the pace at which you are developing/iterating, and have a reference point to be able to cut a check.

I started this 3 months before I planned on raising any money, and it allowed for people who were on the email to confidently make recommendations on my behalf to investors they knew, and also made our company top of mind.

I have continued this for almost a year now, and its led to customer development opportunities, investment inbound, and a steam of good relevant advice.

I agree with the CLIP reference point. Feels very timely.

With remote work being so common, I regularly think about this section:

  Another trait, it took me a while to notice. 
  I noticed the following facts about people 
  who work with the door open or the door closed. 
  I notice that if you have the door to your 
  office closed, you get more work done today 
  and tomorrow, and you are more productive than 
  most. But 10 years later somehow you don't 
  know quite know what problems are worth working 
  on; all the hard work you do is sort of 
  tangential in importance. He who works with 
  the door open gets all kinds of interruptions, 
  but he also occasionally gets clues as to what 
  the world is and what might be important. Now 
  I cannot prove the cause and effect sequence 
  because you might say, ``The closed door is 
  symbolic of a closed mind.'' I don't know. But 
  I can say there is a pretty good correlation 
  between those who work with the doors open and 
  those who ultimately do important things, 
  although people who work with doors closed 
  often work harder. Somehow they seem to work 
  on slightly the wrong thing - not much, but 
  enough that they miss fame.

Possibly the best essay in the Stripe book of Richard Hamming's essays/articles.

Founder of a team of five here. For a sample size of one, we have a Rails based backend and it’s on our roadmap to migrate parts to Elixir.

We have a video based tool and the low compute plus multithreaded aspects of Elixir make it ideal for web based FFMPEG processes.

We are three engineers, all of whom are senior/principal level SWE.

If we were hiring someone next, we’d likely optimize for their frontend skills, given the app is the complicated part, and the API is trivial. That being said, we would expect them to have some advanced experience with a backend language, and to be able to pickup new things quickly.

I can confidently say we wouldn’t hire someone who has only done a few projects in a programming language, and hadn’t worked in a production environment. On the other hand, we would definitely hire someone without relevant experience in our entire stack, if we knew they have proven deep expertise in their own stack and the ability to learn/move quickly.

For anyone curious, this is actually a very big deal.

Fuchsia is a sort of secret Google project to redesign the operating system from scratch - with a focus of making multi-device interactions a base primative.

The goal is to make multi-device interactions easier and not dependent on pseudo HTTP initiated sockets/streams.

It’s very exciting and hopefully makes it possible to have all the devices you are working with easily work together. For example, pull up content on your phone, pass it to your TV, interact with it on your keyboard, sync with your smart watch, etc.

The opposite today would be depending on a series of memory issues or depending on browsers or socket/protocols to manage a series of interactions between devices.

> Disclaimer: word "cuda" is taken from Serbian language, it means "miracles".

This has to do with a Serbian word and not the Nvidia GPU programming language.

While not directly commented on the point above, I would like to publicly recognize dang for his Herculean effort to manage HN and process all of the Launch HN posts. He is an absolute champion and is a unmatched pillar that makes this community run.

Thank you dang!

Love this idea. Congrats on launching!

Feels like Substack, but for video.

In the "creator economy" realm, this looks like the missing part of YouTube - where creators can actually make money without being forced to make content optimizing for clickbait advertisement catered content.

I'm a huge fan of Rezi which has a very similar experience to my knowledge. They are able to reduce broker fee/rents because they can assure a reduced time where apartments are unrented.

https://www.rentrezi.com/

THIS. IS. INCREDIBLE.

Cool to see italki mentioned here.

italki.com was the first startup job I got out of college over 8 years ago in Shanghai, and where I met my current cofounder.

Thank you! The creator of this is a machine when it comes to well designed and quality open source projects.

Models for Work

1. Full snap back to pre-COVID in-office work

2A. Hybrid: 3 days in office for everyone, 2 flex days.

2B. Hybrid: Mainly back in office, but “remote” is considered its own office.

3. Remote only

Neat simple extension.

Can you elaborate on where the audio is stored and when it's removed from your buckets?

The author, Hunter Scott, has been working on everything from plants to farming on Mars to long-distance wireless power-charging electronics.

I spent the last decade tinkering with software/hardware projects. This book is exactly what I wish I had when diving in.

Note: I have no connection to the book. I just really like the author's work.

Edit: The digital book is free.

There is a company that takes this concept and lets you take notes in "spaces".

https://www.nototo.app/

This is an easy to use wrapper on ESBuild, the Go-based javascript compiler.

I have no affiliation to this project, but used it for our company frontend build process. This decreased our production package build time by 8x and makes our local development code change updates more than 10x faster.

I followed ESBuild because of Evan Wallace, but never knew how to actually get it working for a project.

Using Vite - I was able to replace a Create React App based project's Webpack builder, and was absolutely blown away by the improvements.

This is genuinely a cost saver. Since builds are faster, I decreased the number of concurrent builds to pay for. Outside of that, dev productivity is definitely up.

I absolutely love this. Congrats on building such an obviously missing problem. The efficiency/productivity obsession in me dances for joy at the idea of this.

The last company I worked at used the "Accelerate: The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations" book as their bible. It was a wonderful way to structure team rituals and review cycles.

I encounter creative folks who really don't like Slack's notification system, and hate how the workplace has become a bunch of impersonal pings.

Can you talk a bit about the tech stack? How is your video handled and are you using any services worth noting?

One thing that Slack is useful for is an accumulation of a searchable knowledge base. Do you have any plans to make the videos searchable and/or are closed threads/conversations deleted to the end user?

Overall - this makes so much sense. At the beginning of the Covid lockdowns, my wife started regularly using MarcoPolo to keep in touch with her friends. She loved the semi-serendipitous messages and intimacy exchanged. This is definitely missing in the workplace.

Congrats on launching.

I'm continually blown away by these small things. During your launch blog post, you showed how configuration tables can add tens of megabytes, and now the Swift<>Obj-C dictionary format can do the same?

It feels like these should be optimizations in Xcode, so its perplexing, but on the flip side, this horrifies me about my JavaScript/Ruby package management. At least on the Xcode side, there is some single party trying to improve.

They are actively working on this.

I worked on Google Fonts, and you’d be incredibly surprised how few people run the entire service. If I recall correctly, ether ratio is something like one Google Fonts team member is responsible for the equivalent opt hundreds of millions of requests a day.

The main focus they had in the variable fonts roll out was making the internet faster. The process of choosing when to/not serve variable fonts behind the scenes was oriented around the font-weight axis, so they could reduce the number of requests, and hence speed up everything from Google searches to YouTube and every other website using Google Fonts.

Sadly the UX around displaying axes isn’t straight forward. Since the Google Fonts wants to setup a conservative interface, that doesn’t result in users requesting an entire variable axes with every request, especially when it’s not being used - there needed to be more work than just showing the font’s design space on a specimen page.

There is an incredible amount of behind the scenes effort from an incredibly talented team.

Huge props to each of them.

I'll say this clearly: Assembly AI is the hands down best speech to text transcription service on price, quality, speed and support. Hands down. It more than pays for itself time and time again.

If we were free, then I might have this concern, but our average customer value is far beyond the cost to transcribe.

Actually this is insanely cheap, especially given how much they are regularly improving. Amazon costs over $1.25, RevAI costs over $2/hour, and Google Speech to Text is over $2.15/hour.

We have a shared Slack channel with their team and I can't convey how incredible they have been. Literally the moment we have a question, we get instant responses.

Also our users will not poor transcripts. Every word that needs correcting is time/money lost for them, so our goal is to give them the highest quality transcripts.

We want to focus on our key value prop. Transcription is not one of them. We focus on user experience, design options, and speed.

I’m curious - does React Native/Expo do any better job at this, with tree-shaking and package building?

"The Lyft app has hundreds of duplicate files, the largest consumer of space is a single asset catalog copied 73 times in separate bundles. Another asset catalog that is virtually identical except for the timestamp at which it was created is copied 67 times. Each of these contain nothing but 482 colors (colors can be stored in asset catalogs to simplify management of dark mode). With each one taking ~250kb these quickly eat up 35mb."

I read this as: Lyft installed Dark mode for 35mb. I can only imagine what my JavaScript modules are doing behind the scenes.