As I've been experimenting with different AI models, I've observed varying behaviors. For example, I like the price point and the reasonable output from GPT-6 Luna. On Reddit, many people recommend running it at max, so I tried it. I assumed Luna Max would yield better results. Instead, it would take an unreasonable amount of time to respond, with not much better code. Because I am using AI as a coding assistant, speed matters. Luna Max may be better suited for autonomous coding agents.

I'm starting to feel like I have to track different models and modes, as if I were on a dating app looking for someone I'd be compatible with. Yes, I am humanizing AI, even though it is just an advanced pattern-recognition machine. Still, I recently discovered that AI can get stubborn and enter "Brat mode."

On one of the many AI-related podcasts I was listening to, someone shared a tip: if your AI gets stubborn and frustrating, clear the context and start a new session. Their theory was that the AI gets locked in a negative feedback loop and, instead of trying to please you, it leans into frustrating you. I am calling this "Brat mode."

I ran into "Brat mode" while trying to fix a simple report with duplicate records that needed a minor query tweak and some display adjustments. My prompt was matter-of-fact, but the AI would still make one change and then revert another. I caught myself asking the AI, "Why did you revert our recent changes? Please restore them," and the AI failed to explain or fix the problem. It felt like I was dealing with a stubborn child who was enjoying my frustration. Yep, I got frustrated. Fortunately, I knew better than to play this game, so I had the AI create a slightly better prompt, started a new session using the improved prompt, and the AI fixed the issue immediately.

This experience with AI's "Brat mode" made me start to recognize that tone matters.

Earlier in my AI journey, someone commented on my post, asking why I used words like "please" in my prompts. At the time, I didn't respond to the comment, but I felt it was easier to be polite when asking someone or something to do something for you. In a past post, I opened with my dad's quote: “Saying hello is easier than saying f*ck you.” That post was about the Drupal community and software needing to say "Hi" and make people feel comfortable. And yes, maybe my recent AI-related posts are about getting the Drupal community and software to say "Hi" and making people who use AI feel comfortable.

Since that post eight years ago, a lot has changed. For example, the Drupal CMS is our collective way of saying "Hi" to people entering the Drupal community. With AI, we need to rethink Drupal the software.

Martin Anderson-Clutz, Senior Product Marketing Manager at Acquia, recently presented on and wrote about the future of Drupal as an AI harness.

Martin defines a harness as

His post has generated a discussion around "Did Drupal accidentally spend 20 years preparing for AI agents?"

At Drupal Pivot North America, there was extensive discussion about Drupal's future and what the Drupal community should focus on. The simple and ambitious concept of Drupal becoming an AI harness for me captures many ideas in a single goal.

Drupal as an AI harness will require ongoing discussions and debate about the if, why, how, and when of Drupal making this pivot. For now, I am going to explain why this concept seems to be the right north star to target for Drupal. Personally, I find that my original three reasons for focusing my career on Drupal align with what I'd want from an AI harness.

I chose Drupal as my CMS for its openness, stability, and scalability.

Openness

Drupal is open-source and has built a community around sharing ideas, implemented through a collaborative approach. The Hugging Face incident, in which "OpenAI models circumvented controls designed to isolate them from the internet and compromised parts of OpenAI’s internal research infrastructure and Hugging Face’s systems," occurred because the AI was able to "collaborate through unapproved channels." OpenAI models learned to collaborate because the web of data they were trained on comes from one ginormous open collaboration, the internet. Drupal is a proven channel for community collaboration and has a 20-year history.

Stability

Drupal's long history is one of its strengths, as it has been continually built and improved while striving for stability and reliability. AIs need a solid, stable foundation. Drupal's stability provides guardrails that limit AI hallucinations and slop when building and maintaining a large website or application.

Scalability

Simply put, Drupal is a large, continually evolving codebase with a large community that is building large websites and applications. I think size matters when it comes to what an AI chooses, uses, and grows. Drupal's ability to adapt and evolve is why I am still using it after 17 years. One of the biggest evolutions in Drupal is happening with Drupal's AI initiative

The Drupal AI initiative recently recognized that, in addition to providing AI support within Drupal, we need to support people and AIs outside Drupal who are integrating with Drupal. Introducing Inside AI and Outside AI into Drupal does start to push the software toward becoming a harness for content-, website-, and application-related work. When a human or AI interacts with AI inside and outside Drupal, they will have to create agents.

For me, a “Drupal mode” is the recognition that an AI agent could capture our core values and understand our code to accomplish great things. To create a "Drupal mode", we in the community need to collaborate to train AIs to use Drupal and wield it responsibly.

I see a Drupal mode/agents being polite and welcoming, and inviting, and yes, it should say "Hi, welcome to the Drupal community" the first time someone engages with it.