Problem first
Start with an operational need, not a technology looking for somewhere to land.
AI × ENERGY × ENGINEERING
Practical AI for
real industrial problems.
I build, test and benchmark AI against genuine engineering and operational challenges—not what a slide deck says it might do.
OPERATING PRINCIPLEEVIDENCE / 01
BUILD → TEST → SHOWOpen methods. Working outputs.
ABERDEEN / SCOTLAND CURRENTLY BUILDING
THE POSITION
“The useful question isn’t what AI could do. It’s what happens when we give it a real industrial problem and inspect the result.”
Make it inspectable
Show the prompt, source material, output and limitations—not just the polished conclusion.
Build the thing
Working applications reveal more than demonstrations, diagrams or claims ever can.
BENCHMARK 03 / MAINTENANCE
The Monday
Maintenance Meeting.
Eight models confront the same 92-order offshore backlog: expose hidden risk, test readiness and build an executable seven-day plan.
08 MODELS92 ORDERS693 CRAFT HOURS
Explore the benchmark →
BENCHMARK 04 / PRODUCTION LOSS
Production
Loss Atlas.
Three models turn a 90-day loss portfolio into a reconciled account of what happened, why it happened and what must change.
03 MODELS90 DAYS01 LOSS TRUTH
Explore the benchmark →
BENCHMARK 02 / WELL INTEGRITY
Eight models.
One difficult decision.
A fictional North Sea well. An imperfect evidence pack. A real test of whether AI can reason through uncertainty, barrier condition and competing production pressure.
08 MODELS01 EVIDENCE PACKZERO MANUAL FIXES
Explore the benchmark →
HALCYON P-17 / DECISION WORKBENCH
BENCHMARK 01 / HSG253
Nine models.
One industrial challenge.
Each model received the same source document and base challenge: turn HSG253—an 88-page safe-isolation guidance document—into useful digital training. These are first-pass outputs, with no manual fixes.
01OpenAI
Launch application ↗
02Claude
Launch application ↗
03Gemini
Launch application ↗
04DeepSeek
Launch application ↗
05Qwen
Launch application ↗
06Mistral
Launch application ↗
07GLM
Launch application ↗
08Perplexity
Launch application ↗
09Kimi
Launch application ↗WHY HSG253?
Read the method and compare all nine outputs
It is complex, consequential and recognisable to industry. The test is not whether a model can summarise it. The test is whether it can turn technical guidance into something people can understand and use.
GLEN'S TOOLBOX
Work from the field.
Applied ideas spanning production performance, decision quality, operational learning and industrial AI.
Production performance systems
Making potential, constraints, losses and opportunities visible—and owned.
Operational decision tools
Turning complex well and production choices into transparent, repeatable decisions.
Industrial data and agents
Connecting AI to the context held in production data, engineering systems and operating history.
Technical learning
Exploring better ways to help people navigate standards, guidance and operational knowledge.
EARLIER BUILDS / 2025—26
Prototypes that shaped the work.
Earlier applications preserved as evidence of the path: the original problem, the first working response and what happened next.
WELL / 01BENCHMARK LIVE
Well Integrity Monitoring
The original monitoring prototype became an eight-model test of evidence-led well integrity decision quality.
DECISION / 02REBUILD PLANNED
5 Whys RCA
A guided investigation workflow connecting causal questioning, categorisation, corrective actions, an RCA library and reporting.
Open original prototype ↗
SAFETY / 03PRODUCTION ADOPTION
Safety Intelligence
Turns safety-observation records into themes, hotspots, trends, queries and a concise operational brief. The approach subsequently moved into production.
PROD / 04REFRESH PLANNED
Well Surveillance
A traffic-light view of well performance using operating baselines, production indicators and lift-specific parameters.
Open original prototype ↗
ASSURE / 05REFRESH PLANNED
OCA Intelligence
Browser-based analysis of assurance exports, open items, ageing and recurring patterns using workbook or demonstration data.
Open original prototype ↗ABOUT
Operator's context.
Builder's mindset.
I’m Glen Milne, a production and operational excellence leader with more than 20 years in the energy industry.
My background spans production optimisation, asset management, hydrocarbon accounting, emissions, operational excellence and digital transformation. That experience shapes how I approach AI: grounded in the work, alert to the risk, and focused on measurable usefulness.
I use this site to publish experiments, benchmarks and working tools openly. It is an evidence base—not a sales brochure.
Connect on LinkedIn ↗20+YEARS IN ENERGY
28MODEL OUTPUTS PUBLISHED
01RULE: SHOW THE WORK