How can we make Squiggle / Squiggle Hub more valuable?
Mini
11
200
resolved Oct 9
26%25%
Run 1:1 user interviews with members of your target audience
13%13%
A way to call nodes from other people's graphs
13%13%
EA Forum Posts of use cases
11%11%
Estimation workshops
8%8%
Squiggle: a squiggle logo that is a squiggle, because branding
8%8%Other
8%8%
Crowd forecasting integration
4%4%
Office hours, for support
3%3%
Prizes for good models
1%1.3%
Improve stability / reduce bugs
1%1.3%
Make Squiggle faster
1%1.3%
Better Squiggle dashboards
1%1.3%
Squiggle Hub: Commenting support
1%1.3%
Squiggle Hub: LLM integration

We recently released Squiggle Hub, and are interested in making Squiggle and Squiggle Hub useful under EA/longtermist goals.

Looking for suggestions on ways to best improve.

https://forum.effectivealtruism.org/posts/rK3NgqLg3HHDzyLah/announcing-squiggle-hub
https://squigglehub.org/
https://www.squiggle-language.com/

Get Ṁ600 play money

🏅 Top traders

#NameTotal profit
1Ṁ63
2Ṁ10
3Ṁ4
Sort by:

I originally didn't mean this as a formal forecasting question, but as a survey. It asked me to formally resolve - I just put the current market prices for each one, felt like this was better than resolving as N/A.

I haven’t used Squiggle more than 5 minutes, so grain of salt and everything. But I think it might be neat if you use could crowd forecasting sites as an input. Say, have a Bernoulli distribution according to the current probability of a binary Manifold market or a continuous questions on Metaculus as a continuous input distribution.

@1941159478 Yep, this is an area I'm interested in. I'd flag that this is fairly doable with something like ObservableJS, or just using Squiggle in a local JS script. Building it into SquiggleHub would be cool, though a bit tricky.

If people have more specific thoughts on how they would use this, I'd be interested!

@OzzieGooen Oh yeah fair enough. Just saw there’s a version of Squiggle for Python too. That’s what I tend to use and it’s easy to extend, so personally I’m fine already.

 

My dream scenario is that much like someone checking 538 several times a day before an election, people can now keep checking their carefully factored out P(doom) and just need to reload to get new information added in.

Run 1:1 user interviews with members of your target audience

30m user interviews are super powerful for understanding how your users (current and prospective) actually use and think about your product. People I'd interview:

  • Current Squiggle and Squigglehub users

  • Past Guesstimate users

  • People who spend a lot of time making models eg with in-depth EA forum posts

  • Grantmakers

  • People outside EA whose job is to stare at spreadsheets all day

Ideally at least half the interview is just you watching them do actions, without providing corrective guidance (aka give them goals, have them talk aloud, let them fail, take notes).

Other books I'd rec on this topic: Don't Make Me Think; The Mom Test; Talking to Humans

Another important part of user interviews is to try and understand the top problems of users; your goal should be to locate at least a few people for whom Squiggle will immediately solve a burning pain point.

@Austin Thanks so much! That's useful, will try.