{"id": "art-001", "title": "How object storage changed database design", "body": "Separating durable state from compute lets databases scale to zero and recover instantly. Object storage like S3 offers eleven nines of durability at a fraction of the cost of replicated SSDs.", "category": "databases", "year": 2023}
{"id": "art-002", "title": "A gentle introduction to vector search", "body": "Vector search finds the nearest neighbours of an embedding in high dimensional space. Cosine similarity and dot product are the most common distance metrics for semantic retrieval.", "category": "search", "year": 2022}
{"id": "art-003", "title": "BM25 is still hard to beat", "body": "Despite the rise of neural retrieval, the classic BM25 ranking function over an inverted index remains a strong baseline for keyword and full text search workloads.", "category": "search", "year": 2021}
{"id": "art-004", "title": "Hybrid retrieval for RAG pipelines", "body": "Combining lexical BM25 scores with dense vector similarity using reciprocal rank fusion improves recall for retrieval augmented generation systems.", "category": "search", "year": 2024}
{"id": "art-005", "title": "Why write-ahead logs matter", "body": "A write-ahead log makes writes durable before they are indexed. After a crash the database replays the log to recover every acknowledged write.", "category": "databases", "year": 2020}
{"id": "art-006", "title": "Compaction strategies for LSM trees", "body": "Merging many small sorted files into fewer large segments reduces read amplification. Levelled and tiered compaction trade write amplification against query latency.", "category": "databases", "year": 2021}
{"id": "art-007", "title": "Tide pools of the Pacific coast", "body": "Tide pools host anemones, hermit crabs, and sea stars that survive crashing waves and shifting tides. Each pool is a tiny self-contained ecosystem.", "category": "nature", "year": 2019}
{"id": "art-008", "title": "How schools of fish move as one", "body": "A shoal of fish coordinates without a leader. Each fish follows simple local rules about distance and alignment, producing fluid collective motion.", "category": "nature", "year": 2020}
{"id": "art-009", "title": "Sourdough starters explained", "body": "Wild yeast and lactic acid bacteria ferment flour and water into a living starter. Regular feeding keeps the culture active for baking bread with deep flavour.", "category": "cooking", "year": 2021}
{"id": "art-010", "title": "The science of searing steak", "body": "The Maillard reaction between amino acids and sugars at high heat creates the browned crust and savoury aromas that make seared meat delicious.", "category": "cooking", "year": 2022}
{"id": "art-011", "title": "Knife skills every home cook needs", "body": "A sharp chef's knife and a stable cutting board make dicing onions and julienning carrots faster and safer than any gadget.", "category": "cooking", "year": 2023}
{"id": "art-012", "title": "The James Webb telescope's first year", "body": "Infrared observations revealed unexpectedly mature galaxies in the early universe and detailed atmospheric spectra of distant exoplanets.", "category": "space", "year": 2023}
{"id": "art-013", "title": "Why Mars rovers drive so slowly", "body": "With round-trip light delays of up to forty minutes, rovers plan short autonomous drives and verify each move, trading speed for safety on alien terrain.", "category": "space", "year": 2022}
{"id": "art-014", "title": "Orbital mechanics for beginners", "body": "To catch up with a spacecraft ahead of you in orbit, you slow down and drop to a lower, faster orbit. Intuition from driving does not apply in space.", "category": "space", "year": 2020}
{"id": "art-015", "title": "Caching hierarchies in modern systems", "body": "Memory caches serve hot data in microseconds, local NVMe disks in hundreds of microseconds, and object storage in tens of milliseconds. Good systems exploit all three tiers.", "category": "databases", "year": 2024}
{"id": "art-016", "title": "Copy-on-write: snapshots without copies", "body": "By referencing immutable files instead of duplicating them, copy-on-write branching creates instant snapshots of large datasets. Writes diverge while shared history stays intact.", "category": "databases", "year": 2023}
{"id": "art-017", "title": "Approximate nearest neighbour indexes compared", "body": "Graph indexes like HNSW excel in memory, while inverted-file IVF indexes with coarse centroids suit disk and object storage because they read a few contiguous lists per query.", "category": "search", "year": 2023}
{"id": "art-018", "title": "Tokenisation pitfalls in full text search", "body": "Lowercasing, unicode normalisation, and stemming decisions change what matches. A search engine should document its analyser so users can predict results.", "category": "search", "year": 2020}
{"id": "art-019", "title": "Coral reefs and ocean acidification", "body": "Rising carbon dioxide lowers ocean pH, making it harder for corals to build calcium carbonate skeletons. Reef ecosystems shelter a quarter of marine species.", "category": "nature", "year": 2022}
{"id": "art-020", "title": "Migration routes of the Arctic tern", "body": "The Arctic tern flies from pole to pole each year, the longest known migration, navigating by sun, stars, and the Earth's magnetic field.", "category": "nature", "year": 2018}
{"id": "art-021", "title": "Fermentation beyond bread", "body": "Kimchi, miso, and yoghurt all rely on controlled microbial fermentation to preserve food and develop complex sour and umami flavours.", "category": "cooking", "year": 2024}
{"id": "art-022", "title": "Metadata filters make vector search useful", "body": "Real applications rarely want the nearest neighbours from everything. Filtering by tenant, date, or category before ranking keeps results relevant and access control intact.", "category": "search", "year": 2024}
{"id": "art-023", "title": "Black holes that sing", "body": "Pressure waves rippling from a supermassive black hole in the Perseus cluster correspond to a B-flat fifty-seven octaves below middle C.", "category": "space", "year": 2019}
{"id": "art-024", "title": "Stateless compute and the end of pet servers", "body": "When durable state lives in object storage, any compute node can serve any tenant after warming its cache. Deployments become cattle, not pets.", "category": "databases", "year": 2024}