Insights
Analysis from the data.
113 evidence-graded features across 12 research pillars — every figure traced to a source, every claim labelled by evidence strength.
AutonomyCan LLMs actually command an autonomous ship? The honest verdictReasoning about navigation isn't commanding a ship. The LLM can think about COLREGS; it can't hold the conn.Read the analysis →MachineryAI predictive maintenance for marine enginesAnomaly autoencoders now flag degradation 97–637 minutes early on a real vessel — the AI application closest to paying for itself.Read the analysis →RegulationThe MASS Code era has begun — what the IMO's autonomous-ship code actually regulatesThe first global framework for remotely operated and autonomous cargo ships took effect 1 July 2026 — scope, verification, and the road to a mandatory code by 2032.Read the analysis →
Computer vision for autonomous ships: the lookout problemPerception · 2026-10YOLO for maritime object detection: benchmark vs bridgePerception · 2026-10AI detection in fog and darkness: the vision stack's hardest testPerception · 2026-10AI navigation-light recognition: the night-watch COLREGS problemPerception · 2026-10AI detection of small vessels: the rare-class problemPerception · 2026-10Vision Transformers for ship detection: global context, scarce dataPerception · 2026-10AI maritime cyberattack detection: the regulated frontierMaritime security · 2026-10Adversarial attacks on maritime AI: the sensor stack's weakest layerMaritime security · 2026-10AI detection of GNSS jamming: the denial problemMaritime security · 2026-10LLM cybersecurity assistants aboard: the SOC-in-a-box questionMaritime security · 2026-10Remote master vs. autonomous master: the command questionAutonomy research · 2026-10AI navigation in congested ports: the density wallAutonomy research · 2026-10AI navigation through narrow channels: the physics the rules don't coverAutonomy research · 2026-10AI decision-making in restricted visibility: the caution problemAutonomy research · 2026-10How to build a maritime foundation modelAI research · 2026-10RAG vs fine-tuning for maritime AI: the freshness-vs-consistency tradeAI research · 2026-10Hallucination in maritime AI: fluent, confident, wrongAI safety · 2026-10The AI copilot for shipmasters: advice, not ordersThe workforce · 2026-10The AI copilot for marine engineers: the memory the watch lostThe workforce · 2026-10When the autonomous ship leaves its envelope: the designed exitRegulation · 2026-10AI software assurance: certifying the uncertifiableRegulation · 2026-10MASS connectivity and the ROC: the link is the systemRegulation · 2026-10AI risk assessment under the MASS Code: the HAZID that never endsRegulation · 2026-10What is a maritime digital twin — and what isn'tDigital twins · 2026-10The propulsion digital twin: where the technology is already provenDigital twins · 2026-10DQN vs PPO for collision avoidance: the false algorithm debateAutonomy research · 2026-10Computer vision + AIS for collision avoidance: covering each other's blind spotsPerception · 2026-10Multi-objective voyage optimization: the trade-space problemVoyage optimization · 2026-10AI routing + ECA avoidance: regulation as an optimization constraintVoyage optimization · 2026-10Which maritime jobs automate first — and which resistThe workforce · 2026-10AI fatigue detection: the camera that watches the watchkeeperThe workforce · 2026-10Training seafarers for AI ships: the missing curriculumThe workforce · 2026-10Quantized LLMs aboard: the compute budget that decides what's possibleShipboard AI · 2026-10AI at sea with intermittent connectivity: the offline-first constraintShipboard AI · 2026-10AI-assisted bridge watchkeeping: the co-lookoutThe workforce · 2026-10AI-assisted engine-room watchkeeping: the monitor that reads driftMachinery · 2026-10Vision-language models for ships: the screen that explains itselfPerception · 2026-10AIS + SAR + optical: the satellite fusion that watches the seaPerception · 2026-10Human trust in autonomous ships: the calibration problemHuman factors · 2026-10AI copilots for chartering: the assistant in the inboxCommercial · 2026-10AI copilots for port operations: coordinating the choreographyCommercial · 2026-10AI copilots for maritime compliance: the research assistant that must cite its sourcesRegulation · 2026-10Maritime LLM vs general-purpose LLM: the build-or-retrieve questionAI engineering · 2026-10Maritime AI with tool calling: from chatbot to operatorAI engineering · 2026-10Digital twins for autonomous navigation: the sea you can afford to fail inAutonomy · 2026-10Real-time ship digital twins: the synchronization problemEngineering · 2026-10Digital twins for voyage optimization: rehearsing tomorrow's passageNavigation · 2026-10Port digital twins: simulating the terminal before it happensEngineering · 2026-10Digital twins + predictive maintenance: the engine's live baselineMachinery · 2026-10Multi-agent AI for fleet management: a fleet of agents for a fleet of shipsOperations · 2026-10AI speed optimization: the fuel knob nobody uses fullyEfficiency · 2026-10AIS-trained route prediction: the ocean's memory of where ships goAI · 2026-10Routing under forecast uncertainty: optimize against the ensemble, not the pointNavigation · 2026-10AI for auxiliary machinery: monitoring the equipment that actually failsMachinery · 2026-10LLM vs RL for ship navigation: the reasoner and the reflexAutonomy · 2026-10AI visual COLREGS interpretation: reading the rules from the pixelsPerception · 2026-10How the MASS Code regulates AI decision-makingRegulation · 2026-10MASS Code 2026 → mandatory 2032: the roadmap and its risksRegulation · 2026-10What the MASS Code means for seafarers: the occupation being rewrittenRegulation · 2026-10AI literacy for maritime officers: the competency that isn't taught yetHuman factors · 2026-10AI collision avoidance with real ship dynamics: the gap between the model and the seaAutonomy · 2026-10AI inference on shipboard hardware: compute that survives the seaEdge AI · 2026-10TinyML for marine sensors: intelligence at the instrument itselfEdge AI · 2026-10Cybersecurity requirements for AI ships: autonomy's attack surfaceRegulation · 2026-10The future role of the shipmaster: authority without a bridgeHuman factors · 2026-10MASS vs conventional automation: the difference is who decidesRegulation · 2026-10Human responsibility aboard AI ships: who answers when the software errs?Regulation · 2026-10LLMs in maritime operations: the 2026 state of playAI · 2026-10Making an LLM safe for the bridge: the guardrails, not the modelSafety · 2026-10Digital twin vs AI model: the difference that mattersEngineering · 2026-10ML vs traditional weather routing: the physics model vs the learned oneNavigation · 2026-10Can LLMs actually command an autonomous ship? The honest verdictAutonomy · 2026-10Can AI make collision-avoidance decisions safely? Both sides of the evidenceSafety · 2026-10AI cybersecurity for autonomous ships: when the ship itself is the attack surfaceSecurity · 2026-10AI intrusion detection for ships: the network's watchkeeperSecurity · 2026-10AI detection of abnormal vessel behaviour: the pattern that shouldn't be thereAI · 2026-10Cloud vs edge AI for vessels: where the model should liveEdge AI · 2026-10AI decision support vs automation: the line between advise and doSafety · 2026-10Can AI predict the safest ocean route? Safety as the optimization objectiveNavigation · 2026-10Collision avoidance under uncertainty: deciding with incomplete knowledgeAutonomy · 2026-10AI buoy detection: the small object that matters where the water narrowsPerception · 2026-10AI night-time maritime vision: the half of the voyage the light doesn't coverPerception · 2026-10Radar + camera fusion: the sensor pair that covers each other's blindnessPerception · 2026-10AI nautical-chart interpretation: reading the densest document on the shipPerception · 2026-10AI engine-degradation prediction: the trajectory, not the thresholdMachinery · 2026-10AI generator-failure prediction: the machine whose failure blacks out the shipMachinery · 2026-10AI turbocharger monitoring: the signature of the 30,000 RPM rotorMachinery · 2026-10AI bearing-failure prediction: the one failure mode that's everywhereMachinery · 2026-10AI for pumps and compressors: watching the machines nobody watchesMachinery · 2026-10AI fuel-injection monitoring: the per-cylinder read the aggregate gauge missesMachinery · 2026-10Ship-to-twin synchronization: the engineering that keeps the virtual vessel trueDigital Twins · 2026-10Can AI read a radar screen? The multimodal test that separates understanding from captioningMultimodal · 2026-10Navigation-GPT: inside the emerging architecture for LLM ship navigationAI research · 2026-10Agentic AI for autonomous shipsAI research · 2026-10LLMs + digital twins: the language layer shipping's digital models neededDigital twins · 2026-10AI digital twins for merchant shipsDigital twins · 2026-10Reinforcement learning for COLREGS-compliant collision avoidanceAutonomy research · 2026-10Multi-agent AI for multi-ship collision avoidanceAutonomy research · 2026-10AIS + radar + computer vision: building a multimodal maritime AIPerception · 2026-10AI predictive maintenance for marine enginesMachinery · 2026-10Agentic AI for marine engineering diagnosticsMachinery · 2026-10AI detection of AIS spoofing and GNSS attacksMaritime security · 2026-10Small language models for shipsShipboard AI · 2026-10AI-based ETA and fuel predictionVoyage data · 2026-10AI storm avoidance and extreme weather routingWeather routing · 2026-10Can AI explain its navigation decisions?Human factors · 2026-10Human-in-the-loop AI on the ship's bridgeHuman factors · 2026-10Will AI replace seafarers? A function-by-function analysisThe workforce · 2026-10The MASS Code era has begun — what the IMO's autonomous-ship code actually regulatesRegulation · 2026-10Can an LLM be trusted with COLREGS decisions? What the research actually showsAI research · 2026-10AI weather routing: machine learning vs. traditional voyage optimizationVoyage optimization · 2026-10Edge AI at sea: running maritime intelligence without the cloudShipboard systems · 2026-10Suez vs. Cape of Good Hope: what the diversion actually costsVoyage economics · 2025-01
Estimates are labelled as such inside each piece. Nothing here is navigation, compliance or commercial advice — see our editorial standards.
Get in touch
Numbers behind the writing.
The same sea-route data powers the free calculators.