Dash0 Acquires Polar Signals for Continuous Profiling

Traditional observability systems commonly combine logs, metrics, and traces to locate an incident or resource bottleneck. Continuous profiling adds a different signal: sampled execution data that can identify where an application spends CPU time, allocates memory, or executes GPU work. Dash0 said it will integrate that data natively into SignalStore and connect it to the agents that build and operate code.

Business Insider reports that Dash0 is developing agentic capabilities intended to let AI systems monitor and improve running software. The publication reports that a proposed feature called AutoTune is designed to identify inefficient code and generate improvement suggestions. Dash0's announcement similarly states that putting profiling data into Agent0 is intended to support autonomous optimization use cases.

For ML infrastructure teams, GPU profiling is especially material because aggregate GPU-utilization metrics often cannot distinguish data-loading stalls, kernel inefficiency, synchronization overhead, memory pressure, and expensive model-serving paths. Code- and kernel-level evidence can narrow the debugging path, although any automated code change remains subject to normal validation, benchmarking, and deployment controls.

A separate joint post by Dash0 founder Mirko Novakovic and Polar Signals founder Frederic Branczyk states that Dash0 is adopting Polar Signals' Great Lakes database to replace ClickHouse in SignalStore. The founders describe Great Lakes as the fifth iteration of a database effort at Polar Signals, built to handle the high-cardinality nature of profiling data. According to the post, earlier internal iterations used a columnar design built on Parquet and Apache Arrow.

That database work matters because continuous profiles create a large volume of stack-trace samples with many dimensions for filtering and aggregation. Systems handling comparable data types commonly face a tradeoff among ingestion cost, query latency, retention, and flexible cardinality. A backend designed around profiling workloads could therefore be as consequential to operating economics as the profiler itself, though Dash0 has not published migration timing or performance benchmarks in the materials reviewed.

Business Insider describes Dash0 as a New York-based observability startup valued at $1 billion. The acquisition joins a broader observability-market push to connect runtime telemetry with coding agents, where the practical test is whether automated recommendations improve performance without introducing regressions or increasing operational risk.