Root-cause analysis for automotive software

We move developers from a bug ticket to the next step. Within hours, not days.

From a log line to the code that caused it, with the evidence to prove it.

Basset causal chain graph view
Zoomed causal chain segment: anchor to MARKER
The causal chain from the symptom back to the code, across process boundaries.
The promise

With the logs you already have, you always end up with an evidence-backed next step.

01

The exact line of code, backed by evidence.

A causal chain, as in the view above.

02

A set of candidates with confidence levels and evidence.

Candidate list with confidence levels
03

A highlighted area where visibility fades and further instrumentation is required.

This is slow, painstaking work only your experts could do manually. Now anyone on your team can move from a bug ticket to the next step.

The mechanism

How it works, in one pass.

Basset ingests your source and matches each runtime log line to the exact statement that produced it. No debug symbols needed. It then rebuilds the causal chain across process boundaries from what actually happened in the run, and you walk it clickably into the code. Every step is deterministic and inspectable: no hidden AI invents the answer.

The Basset workspace: chain, code, and logs
One workspace: the chain, the code it points to, and the logs that prove it.
Walk-control menu on a chain node

You drive the walk. The tool shows its evidence at every hop.

The research

Built on how debugging actually works.

Before building Basset we studied the research on how developers actually find root causes. Three findings shaped the product:

R1

Developers spend roughly half of debugging time hunting for the right code, not fixing it.

So Basset attacks the hunt: from a log line straight to the statement that wrote it.

R2

Ranked lists of "suspicious lines" fail in practice; developers need the causal story, not a leaderboard.

So Basset shows a walkable chain of what happened, with evidence at each step.

R3

Engineers distrust tools that hide their reasoning.

So every Basset answer traces back to real code and the real run, and the tool says honestly where its visibility ends.

It handles more than one kind of failure.

A permission denial traced across processes, or a hard crash: seed the walk from any log record, including a stack frame.

A crash trace: walk connected to its cause along a real edge-path
A crash trace: the walk connected the failing record to its cause along a real edge-path.

What Basset is not.

not a log viewer

It starts where viewers stop.

not a cloud service

It runs where you decide, and your source and logs stay under your control.

not an AI oracle

Deterministic evidence first, any AI sits on top to explain, never to invent.

Bring us a bug that cost you days.

Basset is in development and we are taking a small number of early partners. You bring real tickets with logs, we prove it on your hardest cases, and early partners pay roughly half the standard price for the first year.

Talk to us about the evaluation