AI is good news for Australian and European software engineers (www.seangoedecke.com)

AI-assisted programming today looks like “centaur chess”: a human engineer supervising a powerful LLM assistant. Current tools favor high-feedback interactions (e.g., Claude Code) where engineers guide and correct every step rather than fully autonomous agents. But the real bottleneck in these pairings is the LLM infrastructure: top models are locked in a few datacenters, run on scarce GPUs, and become less reliable at US peak hours—requests time out more often and labs may quantize models (reduce weight precision) to cut serving costs, degrading performance even if companies like Anthropic deny deliberate downranking. That operational reality creates a measurable hiring and scheduling advantage: American tech firms can get more effective use of limited LLM compute by employing engineers in Australia and Europe to work during US off-peak hours. Those engineers can sustain "around-the-clock" centaur workflows, increase effective engineering hours for urgent launches or bug fixes, and avoid peak-time outages. The approach isn’t universal—long-tenured experts still outperform LLMs on deep, domain-specific tasks—but for typical high-throughput engineering work, time-zone-distributed teams are a practical lever to maximize scarce LLM resources and improve reliability.