2026-08-06

I describe the workflow I use to study core concepts, hand-code meaningful components, and use AI for guidance, simplification, and testing. Distsys Forge packages that workflow so I can repeat it across new projects.

2026-08-01

Quickspin used to be a CLI that told Docker to start containers. It now has a control plane that owns sandbox lifecycle, with the CLI as its first client.

2026-07-24

I am building my first agent sandbox platform!

2026-07-22

A Kubernetes-flavored manager/worker orchestrator in Go: a marginal-cost scheduler, a health-checked Docker task runtime, and a disciplined concurrency model that treats lost updates as seriously as data races.

2026-07-22

A distributed in-memory key/value cache in Go, hand-written once and then rebuilt by AI against the same interface — so the two implementations can be run and diffed file-for-file to study concurrency, consistent hashing, and error handling.

A vibe coding web app that takes a prompt and generates full web applications, with live-streamed agent thoughts and sandboxed execution across multiple AI providers.

• Vercel AI SDK
• Inngest
• E2B Sandboxes
• OpenAI / Anthropic / Gemini
• PostgreSQL
• Prisma
• tRPC
• Next.js

A YouTube video transcription and RAG chat platform with vector embeddings, pgvector similarity search, and multi-turn conversational AI.

• pgvector
• Gemini Embeddings
• LangChain
• RAG
• Inngest
• Prisma
• Next.js

An n8n/Zapier-like workflow automation platform with a visual node editor, topological execution engine, and AI provider integrations.

• React Flow
• Topological Sort Engine
• Vercel AI SDK
• Inngest
• OpenAI / Anthropic / Gemini
• PostgreSQL
• Next.js

A C++23 LLM inference engine built from scratch, featuring arena allocation, mdspan-backed tensors, and cache-line aligned memory.

• C++23
• Arena Allocator
• std::mdspan Tensors
• Llama 3 / Mistral
• CMake
• GoogleTest

A Kubernetes-style container orchestrator in Go with a manager/worker control plane, an E-PVM marginal-cost scheduler, a health-checked Docker task runtime, and a disciplined concurrency model that treats lost updates as seriously as data races.

• Go
• Docker Engine API
• Marginal-Cost Scheduler
• bbolt
• chi
• cobra
• Generics

A distributed in-memory key/value cache in Go, hand-written and then rebuilt by AI behind the same interface so the two implementations can be diffed file-for-file across LRU eviction, sharded concurrency, consistent hashing, and HTTP request forwarding.

• Go
• Consistent Hashing
• Sharded LRU
• TTL Min-Heap
• net/http
• Standard Library Only

A distributed append-only commit log in Go with Raft consensus, Serf cluster discovery, memory-mapped indices, and leader-aware gRPC load balancing.

• Go
• HashiCorp Raft
• HashiCorp Serf
• gRPC / Protobuf
• Memory-Mapped I/O
• Casbin ACL

Built a Python transportation-network optimization project using NetworkX and A* pathfinding to simulate traffic across weighted road graphs and identify the highest-value unconnected roads to add.

• NetworkX
• A* Pathfinding
• Traffic Simulation
• Greedy Optimization

Built a Python network-flow optimization project using NetworkX to simulate flow assignment across directed graphs with edge-capacity and flow-conservation constraints, benchmarked against Edmonds-Karp maximum flow.

• NetworkX
• Simulated Annealing
• Edmonds-Karp Max Flow
• Boltzmann Criterion

Built a Python grid-world policy optimization project using NetworkX to model a reward-based environment with terminal states, obstacles, and probabilistic action noise to learn optimal navigation policies.

• NetworkX
• Value Iteration
• Stochastic Transitions
• Policy Optimization