Blog
Notes from the build.
Computer vision at the edge: smart cameras that decide locally
How to run detection and tracking on the camera or a box beside it: hardware classes, model choice, frame rate, camera placement, privacy and monitoring.AI agent memory explained: short-term, long-term, in between
A model remembers nothing between calls. How context, history, summaries and long-term stores make up agent memory, and when an agent really needs it.Browser and computer-use agents: automating work that has no API
What browser and computer-use agents can do today, where they beat scripts and RPA, where an API is still better, and how to pilot one safely in a sandbox.How to build long-term memory for an AI assistant
A build guide to long-term memory: what to save, how to write and store it, retrieval, contradictions, decay, isolation, the tools and how to test it.LLM application security: a practical checklist
A phase-by-phase security checklist for LLM features, based on OWASP's 2026 lists: threat model, output handling, tools, secrets, cost limits and monitoring.Scaling an LLM app: rate limits, queues and backpressure
What changes between ten users and ten thousand: provider rate limits, queues and priorities, caching, retries with jitter, load shedding and capacity deals.Serverless vs containers vs Kubernetes: where to run what
Functions, managed containers and Kubernetes compared on cost, cold starts, long-running work, GPUs and team size, with a decision table and three scenarios.Event-driven architecture for AI workloads: a practical guide
Why model calls belong behind queues and events: async document and agent jobs, webhooks, fan-out, idempotent retries, backpressure and a worked pipeline.Long-running AI agents: checkpoints, compaction and recovery
What breaks when an AI agent works for hours, and how to fix it: checkpoints, crash recovery, idempotent actions, context that stays useful, and testing.Designing an agent harness: loop, tools, context, permissions
How to design the software around an AI model: the agent loop and its stop conditions, tool results, context assembly, permissions, sandboxing and tests.