Spec review and test plans
Reads massive specs, flags ambiguities, extracts requirements and builds the test plan before bring-up.
A new frontier
in chip design.
Backed by
Deployed on-prem with fast-moving startups and some of the largest chip companies in the world.
One goal.
A fleet of agents.
Bronco plans the work, splits it across agents and runs them on
your compute through Slurm or LSF. Every finding comes back
root-caused and linked to the spec.
GOAL
Close coverage and fix bugs on wsma_fabric
Agent 1
REPRODUCE
CHECK WAVES
FIND-903root-caused
LINK TO SPEC
DRAFTING FIX
FIX
Agent 2
READ RTL
WRITE TB
BUILD SCOREBOARD
RUN ON SLURM
TESTS200/412 passed
DEBUG
Agent 3
CHECK COVERAGE
UNHIT BINS
WRITE STIMULUS
WAITING ON AGENT 2COVERAGE72%
RUN
Agent 4
NIGHTLY REGRESSION
BUCKET FAILURES
SIGNATURESASSIGNED
PLAN
Agent 5
READ SPEC
AMBIGUITIES
ASK ENGINEER
REQSEXTRACTED
FIX
Agent 6
FIX FIND-911
WRITE RTL
ADD SVA
PPA CHECK
RE-VERIFYING
FIX
Days of verification work, done in minutes.
No hourly billing. No per-seat math.
One fee for a service that keeps working.
15 min
to root-cause a failure that used to take an engineer days.
10x
faster to working UVM stimulus, checkers and scoreboards.
96%
functional coverage, from the ~80% where regressions usually stall.
Agents for
Every Task
They read massive specs, track requirements, write RTL and
testbenches, and drive your EDA tools. Every agent shares one
core that keeps improving across projects.
RTL and assertions
New blocks, fixes and SVA derived from spec and waves, PPA-aware and re-verified against simulation.
Testbench bring-up
Working UVM stimulus, checkers and scoreboards, wired up and checked against spec and simulation. Any methodology.
Regression triage
Nightly failures bucketed into unique signatures and assigned, before your DV engineers start work.
Simulation debug
Hand it a failing test. It reads waves, logs, RTL and spec, returns the root cause and proposes the fix. 100GB+ waveforms, block to SoC.
Coverage closure
Regressions plateau around 80%. Bronco closes the last stretch with targeted stimulus for unhit bins.
Protect Your IP. Accelerate Your IP.
Deploy on-prem with Bring-Your-Own-Model and Custom Model Options.
Day 0
Bring-up. One call, deployed inside your environment.
Weeks 1-2
Sample runs to check every tool, spec and design is connected.
Week 3+
Evaluation on a real block or full chip, against other AI and human efforts.
Ongoing
Upgrades and support from a dedicated AE team, in a shared Slack or Teams channel.
