THE_NEW_RULES

The Invisible Oligarchy: Who Really Controls Software Development?

Chapter 2: The Algorithmic Gatekeepers

"Less than 100 individuals control training data decisions for models used by 97% of developers worldwide. We've accidentally created the most concentrated power structure in software history."

A handful of engineers at OpenAI, Google, and Anthropic now influence how millions write code. Their decisions about training data become our architectural patterns. Is this the most dangerous concentration of technical power we've ever seen?

Questions for Debate:

The Power Concentration Crisis

  • Should a few dozen people have this much influence over global software architecture?
  • How is this different from the old days when Microsoft or Sun set standards?
  • What happens when these gatekeepers have blind spots or biases?

The Democratic Illusion

  • We think we're making free choices, but are we just selecting from pre-approved options?
  • Can open-source AI models really compete when they're trained on the same biased data?
  • Is the promise of AI democratization actually creating unprecedented centralization?

The Accountability Vacuum

  • Who's responsible when AI recommendations lead to security vulnerabilities at scale?
  • How do we audit decisions made by training data curators we've never heard of?
  • What recourse do we have when entire programming paradigms get excluded?

Share Your Experience:

Evidence of Influence:

  • Can you trace a technical decision in your codebase back to AI recommendation bias?
  • Have you noticed your team converging on the same patterns since adopting AI tools?
  • What patterns has AI pushed that you wouldn't have chosen otherwise?

Fighting the Current:

  • Have you successfully implemented patterns that AI tools consistently recommend against?
  • What strategies work for getting AI to suggest non-mainstream approaches?
  • How do you maintain architectural diversity when every suggestion points the same way?

The Uncomfortable Questions:

For the Industry:

  • Are we comfortable with this level of concentrated influence?
  • Should AI training data curation be regulated like other critical infrastructure?
  • How do we preserve innovation when statistical patterns drive recommendations?

For Individuals:

  • How much of your recent code reflects your decisions vs. algorithmic suggestions?
  • Can you still evaluate trade-offs independently, or do you defer to AI recommendations?
  • Are you developing your own judgment, or outsourcing it to models?

The Historical Parallel:

Remember when everyone worried about Google controlling web search? Now a smaller group controls how we think about code.

Is this progress or have we traded one monopoly for an even more powerful one?

Loading comments...