Posts and short essays on where deep-tech systems are heading and what it takes to scale them — across company maturity, Physical AI, defence, investment, XR and digital twins.
This collection brings together selected CompoundWorks thinking published on LinkedIn. It is the owned index for the arguments that connect the strategic theses, scaling frameworks and commercial work across the site.
Curated by argument, not chronology.
The visible problem is not always the real constraint. When the constraint is unclear, diagnosis has to come before intervention — otherwise even good people can end up solving the wrong thing.
Read on LinkedIn ↗Deep-tech doesn't scale through tools, methods, or talent alone. It scales through maturing systems. A system-level model for escaping the Scaling Trap — and the foundation for reading Technology, Organisation and Trust as one coupled scaling system.
Read on LinkedIn ↗OKRs and governance rituals can look like Level 3. Then pressure hits, founders override the formal process, and the system snaps back. Level 2 protects speed through informality. Level 3 protects scale through structure.
Read on LinkedIn ↗A company can position itself as a platform before the architecture underneath it is ready. The gap stays hidden until a new customer, market or integration turns the platform promise back into bespoke delivery.
Read on LinkedIn ↗Most teams using AI agents feel faster, but throughput hasn't improved. That's not an AI problem — it's a system maturity problem. When generation becomes cheaper, the bottleneck moves to specification, context, verification, integration and trust.
Read on LinkedIn ↗Y Combinator's Fall 2026 call names four separate priorities that all point at the same transition: AI moving from generating outputs to participating in operations. Not one universal digital twin — the minimum sufficient shared reality for the action at hand, bounded by authority and retained as evidence.
Read on LinkedIn ↗Five systems read the same corridor differently — one sees a clear path, another flags it restricted, a digital twin shows a technician who's already moved on. A world model asks what might happen next. An operational twin asks what's happening now, and what's allowed for everyone.
Read on LinkedIn ↗Demand and technological relevance can be real while scale-readiness remains assumed rather than proven. The same risk applies in M&A: the question is not only whether the venture can scale, but whether the incumbent can absorb it without destroying the value it bought.
Read on LinkedIn ↗European defence is shifting toward software-led systems and consolidation. The strategic question is whether rising investment and M&A close capability gaps — or simply move hidden architecture, delivery and scale-readiness risks into post-transaction integration.
Read on LinkedIn ↗A digital twin of the battlefield can represent the battlespace. An operational twin has to govern action inside it — architecture, decision rights and a trust layer for when the representation is authoritative enough to act on. Train as we operate, not train then deploy.
Read on LinkedIn ↗XR, real-time 3D, simulation and digital twins are moving from representation to coordination. As humans, AI agents, robots and autonomous systems share the same spaces, the deeper opportunity is the spatial infrastructure that lets them act on shared operational context.
Read on LinkedIn ↗The device question is the wrong one for founders and investors. The strategic question is which layer of the spatial system a company can credibly own — visualisation, enabler, infrastructure or coordination — and whether the architecture can hold that position.
Read on LinkedIn ↗The Insights collection captures current observations and arguments. The longer-lived CompoundWorks theses and frameworks remain separate reference pages — updated as the thinking matures, not buried in a chronological stream.
If the decision is already clear, start there. If the situation is still coupled or uncertain, we can create the smallest read that makes the next decision clear.