Start with the work
Each build begins with a recognisable operational or engineering problem—not a generic software brief.
BUILD EXPERIMENTS / 01 / 2026
Six industrial software prototypes. One AI model. No conventional development team. Working engineering experiments built over two days, then independently audited for functionality, technical quality, engineering credibility and commercial potential.
THE EXPERIMENT
What can an experienced industrial professional build in a few hours when modern AI handles much of the conventional software production?
Each build begins with a recognisable operational or engineering problem—not a generic software brief.
The outputs contain simulations, calculations, analytical engines and investigation workflows—not static mock-ups.
Independent audits examine what works, what does not and what would be required before any real deployment.
SIX APPLICATIONS
Each card opens the live standalone application. They use fictional or synthetic data and remain prototypes.
Interactive safe-isolation trainer with a live fuel-gas filter scenario, selection logic, case files and knowledge checks.
Launch application ↗Well integrity management and diagnostics interfaces for monitoring pressures, barriers, anomalies and emerging conditions.
Launch application ↗Process-safety assurance and barrier intelligence built around evidence, performance standards and emerging weakness.
Launch application ↗Production-systems simulation and optimisation across wells, networks and process constraints.
Launch application ↗Evidence-led industrial incident investigation, connecting evidence, hypotheses, causal reasoning and corrective action.
Launch application ↗Hydrocarbon accounting, allocation and reconciliation with material-balance logic and explainable variance analysis.
Launch application ↗THE REALITY CHECK
The scores are independent prototype assessments, not certifications of accuracy, safety, cybersecurity or operational readiness.
Strong accounting and reconciliation capability. The audit also identified a CSV formula-injection vulnerability requiring remediation.
Excellent investigation-training workflow. Its scenarios and evidence are predefined rather than generated from live incident records.
Credible evidence-based assurance analytics. Conclusions are produced from synthetic datasets and need validation against governed operational sources.
Genuine numerical simulation and optimisation. It is not a calibrated or validated production digital twin.
Effective interactive technical learning. Physics and safety logic are intentionally simplified and site procedures still take precedence.
Functional monitoring interfaces. Failure-classification logic is incomplete and annulus-pressure guidance is inconsistent.
THE BIGGER PICTURE
The bottleneck is moving. Software production is becoming cheaper. Problem selection, domain judgement and assurance are not.
Working logic can now be explored before a conventional business case, procurement cycle or delivery team is assembled.
Teams can inspect a functioning application and its weaknesses instead of debating abstract claims on slides.
Domain specialists define the problem, spot false confidence, test operational logic and decide what deserves investment.
CONCLUSION
The cost of building software is collapsing.The value of knowing what to build isn’t.
These six builds do not prove that industrial software is solved. They show that credible ideas can now be made tangible, tested and challenged at a speed that changes how digital work should begin.