< Back to the archive

Like what you see? Subscribe here and get it every week in your inbox!

Issue #271 - May 19, 2024

Here are the top threads of the week, happy reading!

Top comment by ezekg

I have a single code base [0] for the self-hosted editions and for the SaaS. I split features up across two editions: community edition (CE) and enterprise edition (EE), and then also two modes: singleplayer and multiplayer (single- vs multi-tenant). The switching is done via environment variables, KEYGEN_EDITION and KEYGEN_MODE. The SaaS offering actually runs the EE edition in multiplayer mode, and is right now the only instance of the EE edition in multiplayer mode (other EE customers are entitled to singleplayer only).

I've been running this set up for about a year [1] and it's working well. Having a single code base was a requirement before I made the project fair source [2]; the fair source ELv2 license lets me add feature gates to facilitate this, while protecting me from forks giving away EE features for free (while still allowing forks).

Updates are pushed to the SaaS offering daily. I cut self-hosted releases bi-annually unless it's for a critical fix.

[0]: https://github.com/keygen-sh/keygen-api

[1]: https://keygen.sh/blog/all-your-licensing-are-belong-to-you/

[2]: http://fair.io/

Top comment by AnarchismIsCool

I feel it. I'll add that I'm also rather scared of how AI's compute and data hungry nature is de-democratizing tech and...reality in a big way. A handful of giga-scale companies hold all the cards. The actual technology behind AI is mostly a small set of clever tricks for encoding data in a way that neural nets can work efficiently on, the nets themselves aren't magic and the process of making them is well understood, but unless you have the money to shell out for tens of thousands in hardware and have access to literally everyone's private data to train on you're just going to be very second tier for eternity.

Top comment by holman

I used it pretty regularly for non-work things (primarily really great TV or movies I wanted to watch, or live soccer games). Found I couldn't really use it for work stuff (the interactions, though not fatal, aren't in a state that I found a net benefit currently).

Am I using it now? Not at all. But I blame a new puppy in the household upending my schedule rather than the Vision Pro. Big fan of the puppy release, though, definitely one of my favorite innovations as of late.

Top comment by pembrook

Usage-based pricing works great for core infrastructure type-stuff (think AWS, sendgrid, twilio), where the growth of your customers business inevitably means their spend with your business will also grow. Great incentive alignment for both parties.

Subscription pricing works best for end-user tools that help a certain function do their job better (ongoing). Why? If your target customer has to ask their boss for more budget every time they want to use your product — your business is dead in the water.

Top comment by joshvince

Ruby on Rails has recently started actively marketing itself as the solo developer framework. I've worked with it for a long time now (among a bunch of other languages and frameworks) and I have to say the recent updates to the frontend 'story' are really compelling for a solo dev.

One of the compelling reasons to write JS on the server was to have the same language and ecosystem in the client and server, and things like Turbo for Rails really go a long way towards delivering on that for Ruby (although of course you still need to understand javascript and browsers.)

That said, if you're already familiar with something else, then pick the most mature batteries-included framework in that language. Languages are just a tool for the job, it's incredibly unlikely as a solo dev that you can pick a "wrong" one.

Top comment by throwaway38375

You have the same person causing two different problems.

1. With regards to the CTO, treat them like an employee. Prove that their performance as an employee is not good enough, and fire them. They now have no reason to interact with the company on a day to day basis.

2. As a 5% owner, treat them like a shareholder. Put together a reasonable offer to buy them out, and do so. If they don't want to sell then that's fine, you'll only have to interact with them during shareholders meetings.

The first action will solve a lot of your problems. The second action can happen later on when you've got the money to do so.

Top comment by plaidfuji

As others have noted, Yann LeCun is looking beyond autoregressive (next-token-prediction) models. Here’s one of his slide decks that raises some interesting new concepts:

https://drive.google.com/file/d/1BU5bV3X5w65DwSMapKcsr0ZvrMR...

Top comment by meiraleal

I got laid off last month but I'm in no hurry to find a new job. I'm actually not even sure if I want one. Working as a frontend software engineer became too boring, I don't want to work with React anymore. I want to get back to work as a full stack engineer creating full applications but that aren't many jobs for that.

So I'm creating apps for myself with the intention of turning one of them into a profitable business. It is been fun to catch up with LLMs and other models.

Top comment by asicsp

I use pandoc to convert markdown to pdf+epub (https://learnbyexample.github.io/customizing-pandoc/). For web version, I use mdbook (https://github.com/rust-lang/mdBook) mainly because it comes with user choosable themes and search by default.

However, if I had to start from scratch, I'd probably look at Quarto (https://quarto.org/) or Mau (https://project-mau.github.io/index.html)

Top comment by PaulHoule

Speaking as someone who learned to edit video on tape:

I disagree with the people who think you are overthinking it. If you are planning to do something this big you’d better have a plan.

How many hours on camera is this going to take? How much off-camera think time do you expect this to take? I’d imagine if I was coding something this big I’d make at least one major wrong turn that I’d want to if not have to redo.

I think you should try a mini-project that is not part of the series so you can test your setup before you make something that you want to promote. You want to design things so that you get an interesting series even if you quit early.

I’m a little skeptical of the turn to video in the last ten years. I mean, 10 years ago if I got stuck in a video game I could find the answer in a FAQ in less than a minute. Today I have to find the right video in a collection of 60 one hour videos and then seek to the right point in the video. From the viewpoint of a literate person this is a huge step backwards but I’m left with the feeling we’re on the path to Fahrenheit 451.

Who has time to watch it? Is a recruiter going to watch a video that is longer than Game of Thrones?

Myself as a photographer I’ve found that I want to share my works in progress and show people a bit of my behind the scenes work but I’ve found people don’t have time for anything less than the final polished works. People have never been rude to me the way they are to the Midjourney artists who always post the nine images they got without any filtering but I know the engagement isn’t here.