Proof, not adjectives.

Practising engineers can't sign off on an AI answer that changes every time and cites nothing. So we built the opposite: deterministic results, traceable to a clause, with every assumption on the table.

680

Byte-identical cases

Determinism is verified, not asserted: a regression suite runs hundreds of cases and checks the outputs are byte-for-byte identical, every build. Same input, same output — no model drift.

42

Standards implemented

Every result maps to a clause in DNV, API, ISO, ASME, BS and more. The standard reference travels with the number, so a checker can verify it against source.

100%

Shows its work

Every deliverable carries an assumption ledger and explicit scope limits. If something wasn't checked, it says so.

The assumption ledger

Every Deckhand deliverable ends with the inputs it used, the governing clause, and what it did not check. No silent defaults; no hidden scope.

This is what makes a result defensible in review — and what generic chatbots can't give you.

result: min wall = 9.5 mm (util 0.81)
standard: DNV-ST-F101 (2021) §5.4.2
inputs: OD 12.75", P_d 10 MPa, X65
assumed: SMYS 450 MPa; t_fab −12.5%;
corrosion allowance 1.0 mm
not checked: local buckling, fatigue
reproducible: yes — hash-stable

From real projects

Computational Fatigue Assessment

Gulf of Mexico platform life extension

85% time reduction 23% accuracy improvement
Read Case Study

Automated FEA Post-Processing

Subsea equipment qualification pipeline

90% time reduction 50+ reports/week
Read Case Study

All case studies

See it on your own question

Bring something from your real work to Open Deck and check the result against your own methods.

Try Open Deck — free