apodolny

I like the approach of providing a discounted version of the API that is used to train vs. the full price version. Seems reasonable and transparent.

For anyone reading this story as a cautionary tale on having children, you should know that this is not my lived experience of parenting (or the described experience of many of my peers). The way he describes it is the exact opposite of my experience:

> "if you feed me a BIG tablespoon of shit every day and then switch to just a little teaspoon daily, I won’t suddenly start liking the taste of shit."

I genuinely love time spent with my child. I had a similar experience to him and Felix in my 20s and am glad I did, but I'd never want to go back there. That was right for me then and I deeply feel that parenting (alongside a software engineering career I love) is the most joyful, rewarding place to be in my 30s.

> "To keep the trend honest we computed it on a fixed panel of 594 companies that appear in our data every year, 385k reports in total, so the decline you see is the same companies getting stingier."

Isn't there a selection bias here? The average tech company that has stayed in business since 2015 presumably grew quite a bit and can now be choosier.

We've found that rolling our own spam model can be very effective (especially at improving precision). Many sites have their own quirks around what material counts as spam which leads to false positives or negatives wherever your site differs from the norm. No need to go to GPT4 though. We've found even low compute algorithms like random forest perform quite well at the task. You do have to create your own training set, but even a few hundreds or thousands of manually sorted examples can work pretty well.

A little frustrated by the ambiguity of the $100M in the post. Is that capital raised, ARR, GMV? Makes a big difference in terms of understanding what kind of business they built.

Glad someone has written about this!

We talk about the Diátaxis model a lot at Scribe (scribehow.com) where we feel like the how-to guide gets very little love from most of the traditional documentation platforms, which feel like they were all designed for reference material and not much else.

Hey all, looking forward to hearing your feedback on Scribe - a browser extension and desktop app that automatically generates step-by-step guides for any GUI-based process you choose to record on your computer.

Me and my cofounder originally set out to build an RPA tool. Along the way, we discovered that the bigger problem most teams were facing had to do with capturing and sharing each individual’s specialized knowledge. Most of this knowledge wasn’t written down, let alone stored in some sort of shareable format. With hybrid and remote working models becoming the new norm, this struck us as a problem that really couldn’t wait any longer, so we decided to address this need instead.

We wanted to create something that felt quite different to the manual and tedious nature that’s often associated with putting together process documentation or how-to guides. That meant that Scribe had to be able to intelligently recognize the key steps in your process, while still feeling very lightweight and easy to use.

The tool generates instructions and screenshots based on your clicks and keystrokes to build your step-by-step guide--to do this, there were a few key technical challenges we had to solve: 1. Going from a set of keystrokes and clicks to actual steps 2. Recognizing when to take the screenshot (i.e. before you do a key action) 3. Reconciling different DPI screens (pixel densities) 4. Automatically identifying and combining similar (but distinct) steps

We have Basic (free) and Pro (paid) tiers. Basic gives you a Chrome extension to record web-based apps while Pro comes with a desktop recorder so you can record any program on your computer. I’d love to hear your feedback on either. If you’d like to check out Pro, you can use the code HN2021 after you sign up for a one-month free trial.