The headlines started appearing in mid-August 2025. "The Race to Build the AI Software Engineer." "Dueling Visions of Autonomous Coding." "OpenAI's Codex vs. StrongDM's Factory: Who Will Win?" The tech press had found its narrative, and the narrative was war.
Justin read the articles with the expression of a man watching someone describe his house from the outside and get the floor plan wrong.
The timing was coincidental. OpenAI had launched its agentic coding platform in the same window that the factory had begun producing visible results. Multiple AI coding startups had announced products within weeks of each other. Cognition's Devin had been making waves. The press, pattern-matching as the press does, had drawn a battle map and placed everyone on it.
The battle map was wrong, but it wasn't entirely wrong. All of these projects shared a premise: that AI agents could do more than autocomplete code. They could plan, implement, test, and iterate. They could operate with increasing autonomy. The convergence was real.
"It's convergent evolution," Jay said during a standup, after Navan had shared a particularly breathless article. "Multiple organisms evolving similar solutions to the same environmental pressure. Birds and bats both evolved wings independently. They're not competing. They're responding to the same physics."
"The physics being LLM capability curves," Justin said.
"Exactly. Everyone who was paying attention saw the same inflection point. Everyone who had the resources to act started building at roughly the same time. The press calls it a war because wars make better headlines than convergent evolution."
The factory's approach was different from the others, and the differences mattered. The factory wasn't building a product to sell. It was building a methodology to use. The spec-driven, non-interactive model with satisfaction metrics and digital twins was architecturally distinct from the interactive agent assistants that most companies were shipping. But the press didn't draw those distinctions. The press drew a line between "AI that codes" and "AI that codes" and called it a rivalry.
Navan tracked the coverage in a spreadsheet. Not because Justin asked him to, but because he found the pattern-matching interesting. The articles that understood the factory's approach were almost always written by people who had read the factory.strongdm.ai documentation. The articles that mischaracterized it were almost always written from press releases and competitor comparisons.
"The press is using gene transliteration," Navan announced one morning. "They're copying the syntax of one story—AI coding tool launches—and pasting it onto every new entrant. They should be semporting. Different semantics, different story."
Jay grinned. "You just used a factory term to critique journalism."
"The terms are portable."
Justin let the coverage run its course. He didn't issue corrections. He didn't engage with the war framing. The factory's work would speak louder than any headline, and it would speak for longer. Wars end. Convergence continues.
Convergent evolution is the perfect frame. Birds and bats. Dolphins and ichthyosaurs. StrongDM and Cognition. Same environmental pressure, independent solutions. The press wants rivalries. Reality has patterns.