Accelerating Drupal Core development

A LinkedIn post (by Jay Callicott) made the case that Drupal core development needs to accelerate to meet modern (AI-driven) expectations, and adopting AI-DLC is the way to get there. 

"Hot take [..] Drupal Core team needs to adopt AI-DLC [..] (as defined by AWS). AI does the code writing you are doing the orchestration. Who is with me??"

Increasing the velocity of evolving Drupal is a valid and worthwhile goal. The community had already identified the speed of Drupal Core development as an issue. Their solution was to move more quickly outside of Drupal core, in a (non-core) version of Drupal named "Drupal CMS".

Pinpointing the Velocity Bottleneck

Before I go too far - this is not a criticism of the Core maintainers' work or velocity. Drupal Core development is not at all stagnant or slow, and new features and improvements are added in every release.

But I do agree with Jay - we would stand to benefit from increasing the velocity of Drupal Core development. Why do issues sometimes take months, or even years, to complete?

IMO, it's not because we don't write code fast enough.

If not code, what is the bottleneck?

One random example: https://www.drupal.org/project/drupal/issues/2630732 - ten years, 169 comments.

It's not that the code took a long time to produce. It's that the implementations were functional, but had tradeoffs that needed to be resolved. AI-generated code would have the same problem.

But I think we can all agree that, regardless of cause, issues requiring 169 comments over ten years is problematic in the new age.

Retaining Drupal's Strengths

One of the selling points of Drupal core (for large orgs) is the thorough focus on security, accessibility, and stability. Switching to a stated philosophy of "AI does the code writing" would hurt that brand and trust.

In fact, it could further legitimize vibe-coding to current users of Drupal: "If they're using AI to write core, let's just use AI to build exactly what we want. Why is Drupal's AI-code better than mine?"

I am, of course, not arguing that core maintainers should not use certain tools. Drupal core maintainers have consistently shown they're great stewards of quality, and I trust them to make informed decisions about what tools they use and how they use them.

How to Accelerate Core Development

The bottleneck is actually related to collaboration and concurrent availability of core maintainers and contributors.

And that aspect actually is partly addressed by the AI-DLC approach:

"[..]teams unite in collaborative spaces for real-time problem solving, creative thinking and rapid-decision-making[..]"

The Drupal community already does this - collaborating in designated spaces in real-time - periodically through contribution day(s) & contribution spaces at various Drupal conferences.

The issue is that only happens a handful of times a year, and it requires a whole conference and venue to make it happen.

So here's the idea: we fund real-time collaboration for core development outside of the context of Drupal conferences.

We don't need a conference for people who are not colleagues to work together in real-time. We need funding, agreed-upon times, and a coworking space (or a Zoom call!)

Here is an example of how it could work:

1. Drupal Association (DA) creates a "real-time core collaboration" fund
2. "Real-time collaboration days" take place during the last week of each month
3. Groups of e.g. 2-5 individuals submit requests to 'claim' the funding for the upcoming month
   • Groups which include at least 1 core maintainer are prioritized
4. The request includes budgetary requirements.
   • Example "Real-time collab funding" request:
     • We are 4 people, including a core maintainer.
     • We want to work together full-time for 3 days.
     • 2 are employer-sponsored, 2 of us need funding ($100/hr x 8 hrs x 3 days)
     • We plan to focus on the following core subsystem: __
     • Travel/lodging funding is not required.
     • Coworking space funding is not required.
5. To fund the initiative, companies contribute:
   • Employee availability during designated "real-time" collaboration days
   • Funding to the DA (funders can request, not force, their funding be used on specific issues)
6. In return, companies are credited as being sponsors for all issues worked on during the collaboration days, and are given a shout out at DrupalCon.
7. If travel/lodging/etc is needed, participants organize it for themselves.

The cost is a fraction of running 'contribution days', but the impact could be noteworthy.