Posts

Complete list in reverse chronological order.

• How the Mannaeans Sat on a J2a Genetic Frontier

  Feb 18, 2026 5 min

  • mannaeans
  • j2a-haplogroup
  • iron-age-iran
  • zagros-mountains
  • ancient-dna
  • assyrian-empire
  • history
• Why Earth’s Long Evolution Hints Most Planets Never Finish

  Feb 18, 2026 5 min

  • bayesian-inference
  • astrobiology
  • evolutionary-transitions
  • great-filter
  • fermi-paradox
  • research
• How to Turn ‘Room Temperature’ Into a Real Test Condition

  Jan 2, 2026 4 min

  • test-temperature
  • experimental-design
  • measurement-uncertainty
  • environmental-testing
  • thermal-control
  • research
• Three Structural Failure Modes of Single-Vector Cosine Routers

  Jan 2, 2026 4 min

  • embeddings
  • cosine-similarity
  • routing
  • llm-systems
  • retrieval
  • tool-selection
  • research
• Turning Clipboard SQL into a Self-Healing mycli Workflow

  Dec 30, 2025 4 min

  • shell
  • macos-clipboard
  • mycli
  • stdin-vs-files
  • command-chaining
  • code
• Designing Claude Code Tools Using an MVC Mental Model

  Dec 29, 2025 3 min

  • claude-code-tools
  • agents-and-skills
  • mvc-mental-model
  • tool-architecture
  • ai-integration
  • code
• Why ‘Solid’ Matter Is Mostly Empty Space Inside Atoms

  Dec 26, 2025 5 min

  • atomic-structure
  • rutherford-scattering
  • quantum-mechanics
  • nuclear-physics
  • science-communication
  • research
• Why Eridu Gets Called The First City At All

  Dec 25, 2025 5 min

  • eridu
  • first-cities
  • mesopotamia
  • ubaid-period
  • urban-origins
  • archaeology
• Combining BM25 and embeddings instead of using semantic search everywhere

  Dec 25, 2025 5 min

  • search
  • bm25
  • embeddings
  • semantic-search
  • information-retrieval
  • hybrid-search
  • code
• Using cheap embeddings to route language model requests

  Dec 25, 2025 5 min

  • llm-routing
  • embeddings
  • model-selection
  • classification
  • similarity-search
  • system-design
  • code
• Using Postgres pgvector and HNSW Instead of Dedicated Vector Databases

  Dec 25, 2025 4 min

  • postgresql
  • pgvector
  • vector-search
  • hnsw
  • rag
  • infrastructure
  • database-design
  • code
• Why Cretan Y‑DNA Is Dominated By Haplogroup J2a

  Dec 25, 2025 5 min

  • crete-genetics
  • y-dna-j2a
  • eastern-mediterranean
  • population-history
  • paternal-lineages
  • genealogy
• Older models with better plumbing can outperform shiny stacks

  Dec 25, 2025 4 min

  • machine-learning
  • ml-engineering
  • production-systems
  • model-selection
  • data-quality
  • infrastructure
  • research
• Using cosine similarity and HDBSCAN for stable text clustering

  Dec 25, 2025 5 min

  • clustering
  • embeddings
  • cosine-similarity
  • hdbscan
  • nlp
  • unsupervised-learning
  • research
• Rethinking GPU necessity for model inference with CPU batching

  Dec 25, 2025 4 min

  • inference
  • cpu
  • batching
  • latency
  • performance
  • deployment
  • research
• Deterministic LLM pipelines as an alternative to multi-agent orchestration

  Dec 25, 2025 5 min

  • llm-orchestration
  • deterministic-pipelines
  • multi-agent-systems
  • software-architecture
  • ai-engineering
  • production-systems
  • research
• Designing retrieval systems around top-5 user experience

  Dec 25, 2025 6 min

  • information-retrieval
  • reranking
  • recall-at-k
  • llm-applications
  • caching
  • search-systems
  • research
• Improving retrieval relevance with error analysis and hard negatives

  Dec 25, 2025 5 min

  • retrieval
  • hard-negatives
  • error-analysis
  • machine-learning
  • rag
  • contrastive-learning
  • research
• Using vector similarity to deduplicate and select LoRA adapters

  Dec 25, 2025 5 min

  • lora
  • model-management
  • vector-similarity
  • model2vec
  • deduplication
  • clustering
  • research
• Using MiniLM Cross-Encoders Instead of LLMs for Reranking

  Dec 25, 2025 5 min

  • rag
  • retrieval
  • reranking
  • cross-encoder
  • minilm
  • llm-efficiency
  • research