Most teams that tried to bolt an LLM onto a production workflow have lived the same failure mode. The prompt asks for JSON, the model returns JSON wrapped in an apology, a markdown fence, and an extra sentence about being a helpful assistant. The parser breaks. Someone adds a retry loop. Latency climbs from 300 …
Yusuf Gökdemir
Ankara robotics engineer who migrated to Berlin for synth festivals. Yusuf blogs on autonomous drones, Anatolian rock history, and the future of urban gardening. He practices breakdance footwork as micro-exercise between coding sprints.