Not trained.
Encoded.
StarFarer's intelligence isn't scraped from the internet. It is hand-encoded by senior engineers — standards, statutes, the documented lessons of major accidents, and the reasoning chains that connect them. This is what that looks like.
Four domains. One reasoning core.
Every domain is encoded at depth — not keyword-matched, not summarised. StarFarer reasons within these frameworks the way a senior engineer does: knowing which standard governs, when they conflict, and what the incident record says about the gap between what the code requires and what actually goes wrong.
Solar · Wind · Battery Storage
Utility-scale PV performance, AC and DC system design, BESS fire propagation and thermal runaway, interconnection and protection, offshore wind hazard profiles.
Production · Compression · Storage · Offtake
PEM and alkaline electrolyser safety, hydrogen dispersion and jet-fire modelling, area classification, storage and compression risk, fuel cell systems.
Hybrid · Islanded · Remote · Resilience
Hybrid generation dispatch, islanded operation and black-start, protection coordination, rural electrification, remote asset monitoring and fault analysis.
HAZOP · HAZID · LOPA/SIL · Consequence
Structured hazard identification, layer of protection analysis, safety instrumented system design, dispersion and consequence modelling, major accident hazard assessment.
US Federal · European Directives · UK Statutory
The full regulatory stack across the US, Europe and India — encoded so StarFarer knows not just what the standard says, but what the regulator actually enforces and how.
The record that governs everything else
Major accidents encoded not as summaries but as reasoning — what was missed, what the standard said, what the gap was. The lessons that changed the codes are inside the codes StarFarer reasons with.
The accidents that wrote the rules.
Every major standard traces back to a failure. StarFarer knows both — the rule and the event that made it necessary.
Bhopal
MIC release. 3,700 immediate deaths. Rewrote industrial hazard legislation globally — the origin of Seveso II, EPA RMP, and the modern major hazard framework.
Piper Alpha
167 lives. Permit-to-work failure and SIMOPS. The Cullen Report built the foundation of modern offshore safety cases and COMAH.
Texas City
15 dead, 180 injured. Process safety management gaps at BP's refinery. Catalysed OSHA PSM enforcement and the Baker Panel reforms.
Buncefield
Largest peacetime explosion in Europe. Tank overflow, vapour cloud ignition. Rewrote overfill protection requirements across the EU and UK.
Hand-encoded. Not generated.
The difference between StarFarer and a language model answering engineering questions is the difference between a library and a search engine. One retrieves. The other knows.
Not a replacement. Not the only one, either.
Before you evaluate StarFarer against anything else, it's worth being direct about what it isn't — and what else exists in this category.
What you can ask for on day one.
Specific outputs, not categories. Each of these is a real request StarFarer answers today — grounded in your plant data once it's loaded, cited to standard either way.
Dispersion & jet-fire numbers
Hydrogen dispersion envelopes, jet-flame lengths and radiation distances, vent adequacy — computed to NFPA 2, ISO 26142 and API 521, with the method shown. Not estimated. Computed.
Draft worksheets, cited
Causes, consequences, safeguards, gaps and prioritised recommendations for every node in your plant — before your study team sits down. They validate; they don't start blank.
Blind-spot checks
A replacement compressor with a different surge curve. StarFarer re-runs your propagation graph and flags the failure paths a datasheet comparison would never surface.
Gap registers
Every applicable standard, clause by clause: met, not documented, ambiguous. Inspection preparation becomes fixing, not discovering.
Aging-risk triage
Which vessel, which PSV, which line needs the budget now — ranked by your plant's actual inspection findings and degradation trajectories, not vendor urgency.
Novel-hazard candidates
Failure paths crossing system boundaries in configurations no incident database has ever recorded — because they're unique to how your plant is put together.
Committee analysis
Operations, instrumentation and process safety perspectives argued independently, synthesised by a chairman — with minority reports where they disagree, because suppressed disagreement is how things get missed.
Near-miss mining
Your alarm historian, read systematically for the first time: the sequences that came close to a serious incident. The incidents you almost had, found in data you already own.