Product Labs
We advise because we build.
Product Labs is where we turn our own questions into working systems. Some begin as research, some become internal tools, some become client accelerators, some become products — and all of it keeps our advice grounded in the reality of building, integrating, testing and operating software.
Research
Questions
Experiment
Prototypes
Incubation
Repeatable systems
Production
Products & accelerators
Experiment
Context & Knowledge Systems
How teams organise knowledge for humans and agents — conventions, architecture decisions, project memory, retrieval, graph-style structures, reusable prompts and source-of-truth docs an agent can actually use.
ExperimentLocal AI & Private Engineering Systems
Private, local and self-hosted AI workflows — local models, private coding environments, internal context stores, and hybrid approaches that keep sensitive knowledge inside controlled boundaries.
Incubation
Agentic Engineering Workbench
A practical environment for AI-assisted and agentic software delivery — repository-level instructions, context packs, task decomposition, automated checks and review gates — including a harness that makes weaker local models reliable.
IncubationAgentic Integration & MCP Experiments
How agents safely access tools, data and enterprise systems — multi-agent coordination, the Model Context Protocol, agent gateways, and tool / guardrail / handoff patterns.
Production
Domain Reasoning Products
Products that combine structured domain knowledge with conversational AI — encoding specialist knowledge as machine-readable contracts, resolving it deterministically, and serving it through APIs and apps.
ProductionConnected Device & Edge Operations
Practical IoT and edge operations — edge devices, Android-based terminals, ESP32-class hardware, remote diagnostics, secure connectivity, device flashing and field-deployable control systems.
ProductionProduct Discovery & Learning Systems
How AI can support structured learning — guided journeys, practice loops and personalised feedback — rendered across web, print and game environments from a single content bank.
Why these themes
These streams are not a survey of trendy topics. Each pairs a question the industry is actively asking with working code we build and operate — so our views are tested against the reality of shipping, not slideware. The shape they trace is the shift the whole industry is living through: the centre of gravity is moving from the model itself to the system around it.
Standalone chat interfaces are giving way to agentic systems wired to tools, data and workflows — formalised by open standards such as the Model Context Protocol and the Linux Foundation’s Agentic AI Foundation. As that happens, value moves down the stack: from models to agents, tools, context, governance, provenance and, finally, production. Our labs sit on exactly that path — which is why what we learn here flows straight back into the frameworks and the advice.
Models→Agents→Tools→Context→Governance→Provenance→Production