Creating Content Was Never the Bottleneck

There's a great post I just read arguing that writing code was never the bottleneck in software engineering. The real constraints were always code review, knowledge transfer, testing, debugging, and the coordination overhead of working together. AI makes code easier to produce—but it doesn't make it easier to understand, trust, or maintain together as a team.

The same logic applies to translation.

Creating translated content is no longer the bottleneck. AI can produce fluent drafts quickly. But the hard parts—trust, quality assurance, team coordination, and the cost of scaling across languages and projects—remain just as difficult as they ever were. Maybe more so, now that production is cheap and the hype is extensive.

When translation was slow, the slowness itself acted as a filter. Teams had time to review as they went. Decisions accumulated gradually. The process, however painful, built shared understanding.

Now, with AI, drafts can appear faster than teams can absorb them. More output doesn't mean more comprehension. It often means more pressure on reviewers, more ambiguity about who is responsible for what, and more surface area for errors to hide in plain sight.

The marginal cost of producing a translation is approaching zero. The cost of trusting it? Higher than ever.

This is the problem we're working on at Frontier R&D. Not how to translate faster—AI already does that. But how to make trust, accountability, and quality assurance scale alongside the new velocity. How to keep humans in the loop without making them the bottleneck. How to build systems where speed and responsibility coexist.

If we don't solve that, fast translation just becomes fast confusion.