Technical Deep Dives

In-Depth Technical Analysis

Deep technical insights on infrastructure architecture, data center technologies, and emerging patterns that shape how we build scalable systems.

From Intersections

Notes from our Substack

Living essays from the Studio team — published as we work through ideas. Subscribe on Substack →

Intersections · 12 MAY 2026

The Toy Story moment for Physical AI

How the sim-to-real gap collapsed in eighteen months, why the physical AI stack is moving faster than anything in our lifetimes, and the six places we want to build next.

Read on Substack

Intersections · 22 APR 2026

The second grammar

A map of the breakthroughs, the gaps, and a call to the people now standing at the intersection of physics, software, and the physical world.

Read on Substack

Intersections · 20 APR 2026

The Best Surgeon in the World Can Only Be in One Room

How Mentix is building the infrastructure to preserve expert surgical knowledge before it disappears

Read on Substack

Intersections · 07 APR 2026

Safety systems that don't know they are broken

How oil and gas facilities lost track of their own defences — and what a 20-year veteran of the field is building to fix it

Read on Substack

Intersections · 05 APR 2026

The $560 Billion Blind Spot

Data centre finance runs on risk models that don't survive contact with the physical world. One company is building the fix.

Read on Substack

Intersections · 31 MAR 2026

The Surgical Training Gap No One Talks About

Surgical training is stuck in 1890. The infrastructure to fix it finally exists.

Read on Substack

Infrastructure Architecture

Convenience–Control Tradeoff

Every infrastructure decision trades convenience for control. The pattern recurs across cloud, data centers, inference, and enterprise software — and reveals where the highest-value contracts land.

Read Deep Dive

Multi-Tenancy · Inference Infrastructure

The Noisy Neighbor Problem

In shared infrastructure, one tenant's workload spike degrades performance for everyone else. The problem compounds with latency-sensitive inference and bursty agentic workloads.

Read Deep Dive

AI Evaluation · Model Verification

Domain Experts as Eval Builders

LLMs are general. Verification is specific. The people who know what "correct" looks like should be defining the tests — not ML engineers, not prompt hackers.

Read Deep Dive

AI Evaluation · Model Verification

Where Domain Evals Matter Most

Not all domains need expert-built evals equally. The value scales with two forces: the cost of being wrong and the difficulty of checking.

Read Deep Dive

Organizational Design · Hiring

The Wall Is the Job Description

The constraints that define a role aren't written in the job description — they're the walls you hit when you try to solve the problem.

Read Deep Dive

Infrastructure Intelligence · Deep Tech Scaling

The Platform Trap

Why deep tech ventures confuse pilot revenue with product-market fit — and what separates a software business from an expensive consultancy.

Read Deep Dive

Infrastructure Intelligence · Pod 01

Airport Operations

A map of compounding concepts governing real-time constraint optimisation across multi-stakeholder airport systems.

Read Deep Dive

Infrastructure Intelligence

Heat Is The New Constraint

Hyperscalers deploy $300B+ in 2025 capex, but AI chips exceed 1000W TDP. Air cooling can't keep up. Liquid is mandatory.

Read Deep Dive

Industrial AI · Venture Building

Consultancy-to-Platform Transition

The most common failure mode for deep tech startups: building something genuinely valuable but delivering it through services instead of software. Revenue grows linearly with headcount. Margins stay thin.

Read Deep Dive

Industrial AI · Full Stack Architecture

The Physics-to-Software Pipeline

The full stack for industrial AI runs six steps from physical sensor to human decision. Each transition is where companies live or die. The core challenge is not the AI — it's bridging the gap between messy physical reality and clean algorithmic assumptions.

Read Deep Dive

Deep Tech · AI · Edge Inference

Model Compression & Edge AI

The question is not "how small can we make it" — it is "what do we lose, and does it matter for the deployment we care about?" Software foundations, hardware physics, and the constraints where compression stops being optimisation.

Read Deep Dive

Data Insight · Global South · Edge AI

AI Adoption at Scale

0.03% of people on Earth pay for frontier AI — and the same people will be first to leave when capable models run on-device. An interactive look at who actually uses AI, who doesn't, and where the real market lives.

Read Deep Dive

Decarb Industry · Bio-Derived Materials · Green Chemistry

Natural Antimicrobials & Sustainable Materials

Bio-derived actives, functional packaging, and the 2–5 year regulatory gauntlet that decides who scales. A map of the four converging fields and the seven first principles to evaluate any bet in the space.

Read Deep Dive

AI-Native Computing · Platform Paradigms · Agent Infrastructure

The Computer Is Being Reinvented

Not improved, not given a sidebar — rebuilt from first principles. From human-as-runtime to human-plus-agent-swarm, the six native primitives the new computer assumes, and twelve startup shapes mapped to each one.

Read Deep Dive

About Technical Deep Dives

Our Technical Deep Dives explore the architectural patterns, constraints, and innovations shaping modern infrastructure. These analyses inform our investment thesis and help us identify where the highest-value opportunities emerge in the infrastructure stack.