Ready to build what's next?
Whether you're exploring AI integration, rethinking your product strategy, or need a builder in your corner — let's talk.
Start a conversation →
Perspectives on AI, product strategy, builder culture, and what it takes to ship what's next.
Re-identification, jersey OCR, and team affiliation must all be solved from broadcast footage at once. Identity is what separates tracking from understanding.
Read more →Nostalgia engages the brain's memory, reward, and self-reflection networks. That makes it a psychological resource and a design challenge for immersive media.
Read more →Four concrete moves every product team needs to make as AI-native software becomes the default: tool surfaces, evals, retrieval, and mixed human-agent UX.
Read more →Multi-object tracking in football breaks linear motion assumptions. Association quality, not detection, is the real engineering bottleneck.
Read more →AI-native products shift SaaS economics: value moves to workflow completion, costs turn variable and model-driven, quality needs continuous management.
Read more →Reconsolidation means retrieving a memory can change it. For AI systems that reconstruct the past, every replay is a potential rewrite.
Read more →Ball detection is a deceptively hard problem in sports computer vision. Motion blur, occlusion, and pitch line confusion push specialised models far beyond what general-purpose detectors can handle.
Read more →From YOLO to football-specific transformers, player detection is evolving to handle the unique challenges of broadcast football footage.
Read more →We analysed 7,000+ IT staff augmentation companies to map productised services readiness in a $128 billion market. Most providers are indistinguishable. The few that have productised are pulling away.
Read more →When agents call tools, execute code, and act on external data, traditional security guarantees break. Trust engineering is the discipline that fills the gap.
Read more →Football analysts don't need the highest mAP score. They need reliable, fast, and actionable intelligence from the video already in front of them.
Read more →Most software has no API. Computer use agents close that gap by seeing and operating UIs directly, unlocking the long tail of automation that APIs can't reach.
Read more →When agents execute, the UX problem shifts from usability to trustability. Here are the four patterns that make delegation interfaces actually work.
Read more →How sports video intelligence evolved across three eras: from Hawk-Eye officiating to elite tracking data to AI-native single-camera analysis.
Read more →Software development has always been a workflow. As AI agents take on the work, the SDLC maps precisely to the agent architecture it helped inspire.
Read more →The dashboard was built for direct manipulation. As agents take on routine tasks and queries, it is becoming a secondary oversight surface.
Read more →Game state reconstruction is converging as the organising goal of sports AI, shifting evaluation from isolated models to system-level pipeline metrics.
Read more →Human memory is not a recording. It is a reconstruction shaped by emotion, identity, and context. That changes how we should build technology around it.
Read more →Monocular broadcast cameras create occlusion, blur, and resolution constraints that define the real engineering constraints of sports CV.
Read more →Football dominates sports CV research for structural reasons. Understanding why reveals what transfers to other sports and where the gaps remain.
Read more →We treat content publishing as a software engineering problem: git-driven, schema-validated, human-approved. Here is how the pipeline works and why it is designed the way it is.
Read more →The AI pipeline that converts football video into structured game state, from player detection and tracking through tactical reasoning and natural language queries.
Read more →From vector similarity to knowledge graphs: how retrieval evolved in AI-native systems, and why retrieval quality is now a core product dependency.
Read more →Agent workflows don't fit inside a request handler. They pause for humans, retry on failures, and branch across tools. Orchestration frameworks are the new application server.
Read more →We built Grain CLI to answer a question nobody could: how do you trace AI-generated code back to the conversation that created it? Here's what we found.
Read more →APIs were designed for developers reading documentation. Now their primary consumers are AI agents. That shift changes how you should design them.
Read more →The most important user of your software may no longer be a person. When agents become your primary operators, everything changes: API design, permissions, UX, and what it means for a product to be usable.
Read more →Most products calling themselves AI-native are really AI-enhanced: they've added a model to an unchanged architecture. The distinction isn't semantic. It determines whether your product survives the next platform shift.
Read more →AI-native software isn't SaaS plus a chatbot. It's a fundamentally different architecture built around three layers: record, context, and action. Most product teams haven't separated them yet.
Read more →The frameworks that built the last generation of SaaS products won't survive the AI shift. Here's why builder-led teams need to rethink product strategy from first principles.
Read more →Whether you're exploring AI integration, rethinking your product strategy, or need a builder in your corner — let's talk.
Start a conversation →