Howdy Knit neighbors! Hope you put your algo textbook away because there are no technical developments this time.
Advocacy stuff
Lots of networking this week. Some better than others, and I’ve noticed I’ve been demoing to tell the story more. In terms of what could be better, I think relying on the demo might mean the pitch narrative is still not that tight. As for the demo, I’m observing that Knit does not tell its own story. It relies on me to frame and motivate it. The question I am asking myself is how do I invite engagement?
I’ve been thinking for a while now about putting together an MVP SaaS / web app. The least I think would be hosting for run history and data. The data storeis not technically necessary but providing a simple one means external dependencies don’t need to be set up to get started. I think this scaffolding lets us frame ideas (like populating a nav bar with links that don’t go anywhere). And it gives a pathway to create a more motivating demo; the lead hypothesis is a data pull request workflow.
There’s still plenty of work to do on the core though! So it’s not obvious that switching to a web app is the most worthwhile investment right now. Likely I’ll keep it constrained and spend a week or two this month building a totally insecure web app prototype. Worst case it can be a demo aid.
Tool stuff
Some people dream about coding a stable sort for an interview. I dream about Bret Victor’s hands (guarantee the guys behind last issue’s Muse app do too). Unfortunately for the next paragraph to make much sense, you’ll need to read his part about hammers; honestly if you just look at the pictures you’ll get the idea (I recommend the whole article though!). We’ll wait.
OK, back? I believe the hammers that most data practitioners use today are handles without heads. This is what it feels like to work with data today.
Knit at its core wants to be the hammer head. Alone it is still not an ideal tool, but hook up some handles and we’re cooking with fire! (Nailing with Knit??)
Circling back to engagement, another possible path is to build more ergonomic handles for Knit. I’ve been calling these frontends, and one or two could be refined to be much more usable (the template frontend that is similar to dbt, or the Jupyter-style notebook frontend). I don’t plan to put much energy here yet, because I think it could be a big distraction to build what amounts to less mature versions of existing UIs. I think Knit should try to stand out from what’s available instead of showing it can be just as good or slightly better.
Open source stuff
I have a copy of Working in Public whose cover makes my table really pretty, mostly because it’s been closed since I got it. It talks about open source dynamics and helped surface some discussion today around maintainer burnout and project sustainability. I think there’s tricky things to consider, because it’s becoming clear that we’re doing open source wrong. VCs have also been largely unsuccessful backing commercial open source (but big tech has done well with Apache projects). This is all to say that I’ve been questioning some open source best practices(like hosting on GitHub). Maybe I should read the book and engage some contributors before worrying about my (empty) pull request tab.
Next time: SQL!