what i'm on, right now.

shipping

building llm-batcher, an openai-shaped proxy in front of claude. it translates the openai and anthropic shapes both ways, microbatches concurrent requests into short admission windows, and watches itself through an opt-in /metrics snapshot (p50/p95/p99 latency plus an estimated dollar cost). distributed-systems instincts pointed at llm serving.

   openai client                        claude
        |  POST /v1/chat/completions    ^
        v                               |  /v1/messages
+---------------------------------------------------+
|                    llm-batcher                    |
|        translate  |  microbatch  |  observe       |
+---------------------------------------------------+
                          |
         GET /metrics  ->  p50 / p95 / p99  +  $cost