mxstbr

There are people that can do this, that are "social radars." It's rare, but I've seen one and it's mindblowing to observe. 100% hit rate.

(that doesn't mean we should do it in courts to convict people)

…writes a person who's never been a CEO before.

Which is not to say that AI won't be immensely helpful for running companies; just that once you're in the role for two days you realize that replacing that role entirely isn't going to happen.

> AI can win at chess now. You think being a CEO is harder than chess? Come on.

Being a CEO isn't like playing chess. Being a CEO is more like a really complex version of self-driving cars. Why? Because it involves interacting with other humans, all the time. It's all people problems.

Chess isn't. Chess is decision-making with a finite (if immense) set of possibilities. Very different set of problems.

Apparently they bought themselves out of Meta again?

While I'd take the metrics they talk about with a big grain of salt (they heavily depend on the specific startup being evaluated), the round sizes are in the right ballpark. Our[0] actual round sizes were:

1. Pre-seed: $1M (led by System.One)

2. Seed: $4M (led by Boldstart)

3. Series A: $25M (led by Tiger Global)

Note that all of these were all raised in 2021 & 2022 before the investment market crash, but even now I regularly see[1] rounds of similar sizes.

[0]: https://stellate.co

[1]: https://mxstbr.com/investing

I built a TypeScript-based DSL for Karabiner Elements that allows me to work with hyperkey sublayers, thus enabling me to have keyboard shortcuts for pretty much everything I do across my entire Mac: https://github.com/mxstbr/karabiner

Demo: https://www.youtube.com/watch?v=j4b_uQX3Vu0

I built something similar myself with Karabiner Elements: https://www.youtube.com/watch?v=j4b_uQX3Vu0

TL;DR: Capslock is Hyper key when held (Escape when pressed). Then I have hyperkey sublayers to allow for having essentially infinite commands: Capslock + O ("Open") + S ("Slack"), Capslock + S ("System") + U = Volume Up, J = Volume Down, etc.

This is what I had hoped Anki would eventually become. I'm sure there are some advanced features that power users are going to miss in Mochi, but for my simple "don't forget things I want to remember" Mochi is perfect.

- Creating new cards is much easier - Great cross-platform & mobile support - Modern, easily usable interface

Esther commented on Twitter[0]:

> The worst take you could have from watching me go all-in on Twitter 2.0 is that my optimism or hard work was a mistake. Those who jeer & mock are necessarily on the sidelines and not in the arena. I’m deeply proud of the team for building through so much noise & chaos.

[0]: https://twitter.com/esthercrawford/status/163004021988091494...

I don't think Sourcegraph is in big trouble here. Their whole play is enterprises, who likely have code spread across many different hosts. On top of that, their code search is still miles ahead of GitHub's.

We've been using https://bwamp.me for the past year at Stellate to make our larger meetings more engaging. It's essentially a soundboard with some common sounds (e.g. clapping) and some less common sounds (e.g. cow moos) that has added a lot of interactivity to our meetings — in particular for the speaker.

However, the biggest downside of Bwamp is that it's a website that every participant has to open, not a very convenient user experience.

Sendbloop is a reimplementation of Bwamp as a Zoom app right inside Zoom.

Note: I have no affiliation with Sendbloop other than having wanted "Bwamp inside Zoom" for a long time[0] and now being a paying customer.

[0]: https://twitter.com/mxstbr/status/1613163267957956620?s=20

Definitely styled-components[0].

#257th most starred repository on GitHub, used by millions of developers to ship millions of websites — you've very likely visited websites that are built with it!

[0]: https://github.com/styled-components/styled-components

How does Unthread compare to Abbot? (https://ab.bot)

This reminded me of the video by DIY Perks building an "artificial sunlight" for their office: https://www.youtube.com/watch?v=6bqBsHSwPgw

The amount of effort is unbelievable, but I'd love to have something like this at home. I can only imagine the impact on my productivity during winter evenings…

Who is making “Discover Weekly” for podcast episodes?

I want personalized episode recommendations based on past listening behavior; I don't have time to dig through 30 podcast feeds to find an episode that interests me, which is why I usually only listen to specific episodes that friends recommend.

I bet recommendations across all podcasts (not just ones I know of) that match my interests to others with the same listening behavior would massively increase the amount of time I spend listening to podcasts.

I'm an SCA (Speciality Coffee Association) certified barista and have, as part of my work travels, drank speciality coffee all around the world in many different coffee shops.

Most of it is terrible.

It's not that I have the best palate or anything like that; it's that most of the places have undertrained staff behind the counter and serve under- or overextracted coffee all day! Most of it is either acidic or bitter to the point where you're losing all the qualities of the coffee beans and upsets my stomach.

I've only found a handful of shops around the globe that genuinely care about the craft and will serve genuinely good coffee every time. (e.g. Prufrock in London, Tim Wendelboe in Oslo) The rest of the coffee shops are worse than useless, I'd rather go to a Starbucks or even McDonald's for my coffee than upset my stomach with that crap.

Fig recently shipped something related, which translates plain English into bash commands with AI: https://fig.io/user-manual/ai

Thanks for the nice words! I stopped maintaining ThemeUI when I left Gatsby in 2021.

In terms of getting the word out, it's really all based on my personal audience. I haven't found a repeatable marketing channel outside of just… tweeting about it, which is also visible in declining sales as I saturate that channel.

Would love to hear your thoughts!

I make https://bedrock.mxstbr.com, selling a full-stack Next.js & GraphQL boilerplate for building SaaS products, which comes with:

* A complete TypeScript toolchain with end-to-end typesafety from the database to the client

* User authentication

* Subscription payments

* Teams

* Invitations

It's also completely unstyled, which is somewhat unusual in this space. I've found that most of the time, people would just delete all the styling anyway, so instead, the frontend only contains the minimal code necessary to make the functionality work.

I originally launched it in February last year, and it's still going strong now!

How does this compare to KeystoneJS, another TypeScript CMS? https://keystonejs.com/

@Manish: is this your continuation of the Dgraph technology?

Context: Manish was the founder of Dgraph but recently left the company behind it. The company got new leadership and recently announced new funding: https://dgraph.io/blog/post/funding-20220720/

We do at Stellate: https://stellate.co/blog/one-job-one-pay

The discussion he's talking about in this post: https://news.ycombinator.com/item?id=31859072

Our GraphQL Edge Cache is based on Fastly's infrastructure, so we do use Varnish under the hood for the cache storage. In order to properly support GraphQL at the caching layer you have to understand GraphQL so you can do fine-grained invalidation.[0]

Essentially, our GraphQL Edge Cache is similar to a GraphQL client. It looks at the request body with the query & the response with the typenames and tags the cached query result with all the objects contained within. E.g. a getBlogPost query that fetches the blog post and its author will be tagged with BlogPost#asdf123 and User#gjkd489. Then we can invalidate whenever those specific objects change.

[0]: https://docs.stellate.co/docs/purging-api

Our C-level compensation is transparent internally[0] and anchored to the market data, just like all other roles.

I will say that for anyone else thinking about doing this, there is a lot less market data available for those positions. We had to combine market data from multiple providers to get to a good baseline.

[0]: https://www.sueodio.com/pay-transparency/

We are internally publishing levels and sub-levels with explicit skills & traits for each.

One of the easiest ways that we can hold ourselves accountable with minimum friction is through pay transparency. https://www.sueodio.com/pay-transparency/

We’re rolling it out in the next week 2 weeks. We will begin publishing salary bands on every job posting. Internally, we will publish levels and sub-levels for every job.

We’ve already begun using this framework and have made offers to candidates with surprising non-obvious results. Here are some examples:

- We offered one person $27K more than their asking salary because the data showed that they were underestimating their market value.

- We offered someone 30K below what they were currently making

- We’ve made a comparable market offers

We are putting a leveling system in place as we speak. Once we have that ironed out, in a few weeks, all job posts will show a salary range.

We are anchoring our pay to the 60th percentile of San Francisco. According to the most recent market data, this is higher than the average European pay.

This video is amazing.