Key Takeaways

  • Dan Bender and Kirby Montgomery at Code & Co. track the rise of dark factories, where automated AI agents generate software without human oversight or verified test suites.
  • Production costs for individual software features dropped toward zero, but Bender points out that the lifetime total cost of ownership for AI-written code remains unmeasured.
  • CTOs face immediate pressure from private equity boards to cut engineering headcount based on headlines, despite retaining full personal accountability for production outages.
  • Effective engineering teams replace key-person hero engineering by writing developer knowledge into durable instruction files rather than letting unmonitored code run loose.

The Cost per Feature Collapsed but Maintenance Is Unknown

Code generation tools make writing software look nearly free. A developer prompts an agent, an entire feature appears, and the commit gets pushed to staging. Dan Bender calls the uncontrolled version of this dynamic a token furnace, where companies burn capital running prompts without calculating what happens when that code breaks six months later. “Suddenly the cost of a feature has gone down so much,” Bender noted. “On the other hand you don't want to become this like token furnace kind of factory where you just burn through tokens mindlessly, plus the total cost of ownership when it comes to software engineering is still unknown.”

When an engineering department does not understand how generated code functions, the initial speed gain turns into deferred technical debt. Kirby Montgomery defines these uncontrolled systems as dark factories, describing them as setups where “vibe codes being written and you don't really know if it works or not. The robot said it was tested, said it works.” When software engineers stop reading code line by line, the cost does not vanish. It moves from upfront developer hours into debugging, unexpected architectural failures, and emergency patches during peak production hours.

Executive Pressure Clashes with System Stability

Private equity deal teams and operating partners see public market headlines and expect immediate headcount reductions in technical teams. Montgomery points out that CTOs now sit between aggressive board directives and platform reality. “There is a reality though where every CTO is getting bombarded by every LinkedIn article, Wall Street headline,” Montgomery said. “Hey, we're building a dark factory this week. Hey, can we fire half of our engineering? And overall, they're still accountable to stability.”

Cutting developers to capture immediate margin expansion creates structural risk across software portfolios. The engineers who built legacy architectures understand the edge cases that agents fail to catch. When sponsors strip headcount before establishing automated verification, stability falters. Bender explains that the actual upside of autonomous development lies in process continuity rather than headcount reductions. “If your agentic engineering workflow and the harness you build around it is done good, then you move beyond hero engineering,” Bender observed. “There's less key person risk and then things you learn are written into instruction files.” Turning tacit technical knowledge into machine-readable operating standards de-risks software assets; firing engineers without those standards simply introduces unmonitored failures.

Why It Matters

Board mandates to slash technical headcount ignore the unmeasured balance-sheet liability of autonomous code generation. In buyout diligences, technical audits are moving prior to the letter of intent to price the cost of untangling unverified AI features before close. Sponsors who force margin expansion through engineering cuts will likely see their multiple contraction come at exit when buyers audit system stability and find unmaintainable platforms.