Thread by @ShriramKMurthi

1/ Several people have asked me to summarize my exploration of the low-code/no-code space. Here's what I learned. Note that this is VERY temporal: what's true today may not be in one month (especially with so much VC money sloshing about). Also, not tagging any companies. 🧵

2/ The principal dimensions of the space that I can discern are:

- new databases/tables
- data analytics, including visualization
- CRUD
- one-offs

I'll discuss these not in this order, but focus on CRUD, which to me is the most personally intriguing of the lot.

3/ The one-offs are things like vision model generators (Lobe), medical viz (MeVisLab), notebook-based data-driven storytelling (Observable). All very nice, but not much (yet) comparable to them, so they're specialized point solutions.

4/ The data/tables and analytics are often interwoven and can be hard to separate: some one, some the other, most somewhere between. These are products like Tableau, Sigma, Airtable, Budibase, and Retool. Some focus on the data, some are more code-oriented.

5/ I didn't have a good data analytics example in mind when I set out, so I mostly glossed over these. (But I do have a concrete analytics need now [🚴, caused by weaknesses in @FulGazApp], so I may return to them.) What I did notice is they are glossy and quick. But:

6/ What is not clear is how flexible they are, nor how good they are if you care about things like statistical rigor. Some are probably much more so, some definitely less so. Other things surprised me, but they are more general than this group, so more below.

7/ Now we get to the CRUD-ish group. Here, I see two spaces. One is the tools that are built around dataflow. They're all vaguely reminiscent of the late, great Yahoo Pipes (and others over decades). My notes say n8n, Integromat, Automate, Node-RED all fall here (E&OE).

8/ Some of these have support for IoT devices. This is good: I've programmed IoT (eg, particle.io) with C variants and it can be painful. In fact, I have an old particle system that needs updating; I may try one of these if they can work with particle.

9/ Popping back up two levels, back to CRUD. I've been interested in this a long time. In the mid- to late-noughties, I'd built several CRUD apps (conference, job search, grad apps, etc… managers) and was spending my time on the wrong things. I felt only three things mattered:

10/ 1. Data model. 2. Access privileges. 3. UX. Everything else should be automated. We built a system, Alchemy, based on Alloy, for 1 (and some of 2).
cs.brown.edu/~sk/Publicatio…

11/ As for 3, we were working on Flapjax, and I was especially intrigued about treating user actions as abstractions (eg, section 2.4, drag-and-drop as an abstraction). So it seemed it'd be great to wire the two together, and even prove things.
cs.brown.edu/~sk/Publicatio…

12/ Unfortunately, none of this went anywhere. There was no real interest at all in Alchemy, and Flapjax couldn't get any interest out of industrial users: reactivity was just too weird for JSers. So I moved on to other things. But I've always been intrigued!

13/ So now on to the main no-code/low-code CRUD space. This is where most of the action is. And I'm really fascinated by it. It's terrific and terrible at the same time. And I'm starting to see some real structure in this space. Most of all: I'm optimistic!

14/ I'm optimistic because it really does democratize some aspects of programming. We didn't need to write code for many of these things. And now we really don't need to. You don't need fancy math and synthesis and what not. You just need to … not write code!

15/ Increasingly table stakes:

- linking to backing stores (GSheet/Airtable/…)
- lots of std UI widgets
- adaptation across resolutions/devices
- integrating workflow automation (eg Zapier)

This is GOOD! Eg, fewer crappy home-grown, ill-tested widgets is great for all users.

16/ It's AMAZING that on so many systems now, within 5 minutes, you can connect to a backing store, drop in good widgets, link up some logic, generate an app, run it on multiple devices, and have updates from one show up soon on the other…for NO pain.

17/ I really can't emphasize how awesome this is. Talk about lowering the floor! I'm so delighted by these, I'm likely going to have a lab in my intro class starting next fall that has students use them. Gatekeepers excluded, we should all be delighted by this development.

18/ Okay, so what are the differentiators in this space? From what I can tell:

- programmability
- "enterprise"
- open source/self-hostable
- ML integration
- DB sophistication

In turn:

19/ Programmability: Many of these have ways of opening up the box for extension with code. Typically React, JavaScript, … but not exclusively. I'm surprised I haven't run into any that offer a block interface. Someone needs to occupy that space. Use Snap! (@moenig)

20/ "Enterprise": Do you view this as a serious, long-term asset or as a throw-away item? Everything from version control to collaboration to review. Various people grumbled to me that absence of this is a big weakness. But some do understand the need and offer it.

21/ Open-source/self-hostable: self-evident. Most are totally closed. A handful are not. [Not limiting to CRUD, here are some: Budibase, n8n, Draftbit, Baserow, Appsmith.] Clearly a big differentiator, especially since many in the crowded CRUD space will go poof in 2-3 years.

22/ ML integration: I'd have expected to see more ML in the mix, e.g., to help with system design/production. Saw very little, and what I saw may have been very lightweight stuff. I expect this will be a big growth area in this space, especially with lots of low-hanging fruit.

23/ DB sophistication: Most are very unsophisticated, giving you no more than a Google Sheet-equivalent. Now, I'm a big fan of treating Sheets as your DB (that's what @PyretLang does), but it's only okay for starting purposes. So many issues that it can't help with.

24/ Some systems let you use an RDBMS, and let you break out with your own queries. Especially the more corporate-facing ones. As an extreme case, Retool seems to be centered around queries. Of course, once you have an RDBMS, you can do all sorts of things.

25/ What I was surprised by is the unsophistication of the rest. There's little to no support to understand your data model, write down integrity constraints, even express data validation (beyond widgets), etc. This will create very brittle apps. Big hole here.

26/ The quality of documentation also varies widely. Some are stunning. Some are rubbish. Some are sub-rubbish (eg, their own docs are full of broken links). I suspect documentation quality is directly proportional to the amount of VC money being pumped in.

27/ I'm really tempted to dunk on the offerings of some major companies (eg, Amazon and Google), but I'll take the high road. But I've already written publicly about Amazon Amplify; see here:

https://twitter.com/ShriramKMurthi/status/1475463825507991553

28/ Overall, I'm very excited to see what happens in this space. There's a ton of cloning, and that will shake out. This can greatly reduce the barriers to entry for people. The trick is how to grow from here; those who've heard my CS Ed talks know I think a lot about this issue.

29/ But unlike many other "low floor" entry points, this one can produce truly useful things in truly short amounts of time. We shouldn't discount the value of that.

Okay, I've spent an hour just writing this thread! I'm done. Thank you for coming to my Gartner report. (-: •

PS: One thing I forgot to add: there are now dev shops that help you build a NC/LC app. That's some very rich irony, and yet at the same time quite sensible.