Key Takeaways

  • Frontier labs bought tens of thousands of Mac mini and Mac Studio devices to run computer use reinforcement learning loops in parallel.
  • Open-source software like Blender wins model proficiency over proprietary rivals like Cinema 4D because researchers can clone free instances a million times without licensing friction.
  • Slack receives superior agent support because AI researchers use it internally all day, turning their own workplace workflows into direct reinforcement learning environments.
  • Closed suites like Adobe Premiere risk falling behind as creative builders migrate toward open tools that AI models actually know how to operate.

The Licensing Trap for Proprietary Software

If you want an AI agent to master 3D modeling, you need to let it fail millions of times in a sandbox. That requires spinning up thousands of virtual environments simultaneously, running actions, and scoring outputs.

This simple math creates an immediate problem for proprietary software. John Coogan pointed out the physical scale of these operations: “Because Blender is open source, can be downloaded and copied a million times into reinforcement learning environments. We know that OpenAI bought, quote, 'tens of thousands of Mac minis and Mac studios to do computer use training on.'”

Try doing that with closed enterprise tools. Coogan highlighted the contrast with video editing: “I wonder the the interesting thing is that most video is still edited in Adobe Premiere. It's a closed source Adobe, you know, they have their guardrails. They probably they might not want to, you know, say, 'Oh yeah, sure. Take 10,000 licenses, no problem. Pay us $10 a month for that.'”

Proprietary vendors built their business models on per-seat subscriptions and DRM protections. That licensing architecture prevents AI labs from spooling up massive training clusters. As a result, models get trained on the software that is free to clone. Over time, models become fluent in Blender and clumsy in Cinema 4D.

The Dogfooding Flywheel in Frontier Labs

Agent capabilities do not emerge in an abstract vacuum. They mirror the exact digital tools that AI researchers stare at every day. When labs run reinforcement learning on daily human tasks, they run them on their own operational stack.

“And specifically he said Slack is better for work with AI agents,” Coogan noted. “And why is that? Well, it's because the frontier labs all use Slack. And so they have Slack as a reinforcement learning environment.”

This creates an accidental moat. The software stack of frontier labs becomes the software stack where autonomous agents work best. If researchers coordinate on Slack, build with Python, and render in Blender, the resulting models will execute commands inside those specific interfaces with high reliability.

Coogan expects this dynamic to alter commercial software adoption: “I do wonder if people will wind up shifting from Cinema 4D to Blender because the models are better at Blender.” When an agent can execute ninety percent of a 3D pipeline in an open-source tool, paying for closed software that the agent cannot touch becomes a liability.

What to Do With This

Audit the core software stack your team uses this week. If your creative or operational workflows rely on closed tools with strict licensing walls, test their open-source equivalents inside your automation pipeline. Move your experimental agent workflows to scriptable, open-source software like Blender before proprietary lock-in strands your team with tools AI cannot operate.