Skelf Research BlogResearch articles from Skelf Research — an independent AI research lab investigating machine reasoning, computational intelligence, and safe AI systems.https://skelfresearch.com/en-gbcompere: Multi-Armed Bandit Ranking with Fewer Comparisonshttps://skelfresearch.com/blog/cross-cutting/compere-vs-bradley-terry/https://skelfresearch.com/blog/cross-cutting/compere-vs-bradley-terry/How compere uses bandit algorithms to rank items with minimal pairwise comparisons — applications in search evaluation, recommendation, and tournament design, vs Bradley-Terry and TrueSkill.Tue, 23 Jun 2026 00:00:00 GMTformal-optimisationmulti-armed-banditMABBradley-TerryTrueSkillElorankingcomperecomparisonembedcache vs Redis vs GPTCache: Caching for Embedding Computationshttps://skelfresearch.com/blog/cross-cutting/embedcache-vs-redis/https://skelfresearch.com/blog/cross-cutting/embedcache-vs-redis/A practical comparison of embedcache, Redis vector cache, and GPTCache for eliminating redundant embedding recomputation in RAG pipelines.Tue, 23 Jun 2026 00:00:00 GMTsafe-computingcachingembeddingsRedisGPTCacheRAGembedcachecomparisonl0l1 vs Text2SQL Tools: A Privacy-Preserving SQL Co-Pilot for Your Own Databasehttps://skelfresearch.com/blog/cross-cutting/l0l1-vs-text2sql/https://skelfresearch.com/blog/cross-cutting/l0l1-vs-text2sql/A practical comparison of l0l1 with generic Text2SQL tools — when the privacy guarantee matters, when the learning from query patterns matters, and when a vanilla LLM SQL bridge is enough.Tue, 23 Jun 2026 00:00:00 GMTllm-cognitionSQLText2SQLcopilotdata-analysisprivacyl0l1comparisonllamafu vs llama.rn vs flutter_llama_cpp: On-Device LLM for Flutter in 2026https://skelfresearch.com/blog/cross-cutting/llamafu-vs-llama-rn/https://skelfresearch.com/blog/cross-cutting/llamafu-vs-llama-rn/A practical comparison of llamafu, llama.rn, and flutter_llama_cpp for running LLMs on Flutter — what each does well, what each doesn't, and which to pick.Tue, 23 Jun 2026 00:00:00 GMTedge-intelligenceon-devicellama.cppFluttermobilellamafullama.rncomparisonmemista vs Pinecone vs Qdrant vs Chroma: Do You Need a Vector Database?https://skelfresearch.com/blog/cross-cutting/memista-vs-pinecone/https://skelfresearch.com/blog/cross-cutting/memista-vs-pinecone/A practical comparison of memista, Pinecone, Qdrant, Weaviate, Milvus, and Chroma for vector search — and why the answer is usually 'no, you need SQLite'.Tue, 23 Jun 2026 00:00:00 GMTsafe-computingvector-searchPineconeQdrantWeaviateMilvusChromaLanceDBSQLitememistaANNcomparisonmullama vs Ollama vs vLLM: Choosing a Local LLM Server in 2026https://skelfresearch.com/blog/cross-cutting/mullama-vs-ollama/https://skelfresearch.com/blog/cross-cutting/mullama-vs-ollama/A practical comparison of mullama, Ollama, vLLM, LocalAI, and LM Studio for local LLM serving — when to use which, and why you might pick the research-instrumented option.Tue, 23 Jun 2026 00:00:00 GMTllm-cognitionOllamavLLMLocalAILM Studiomullamallama.cpplocal-llmcomparisonperishable vs Cloudflare API Gateway vs HashiCorp Vault: Short-Lived Credentials for LLMshttps://skelfresearch.com/blog/cross-cutting/perishable-vs-cloudflare-vault/https://skelfresearch.com/blog/cross-cutting/perishable-vs-cloudflare-vault/A practical comparison of perishable, Cloudflare API Gateway, and HashiCorp Vault for ephemeral LLM API credentials — when zero-trust LLM access needs its own purpose-built tooling, and when a general-purpose solution is enough.Tue, 23 Jun 2026 00:00:00 GMTllm-cognitionzero-trustephemeral-credentialsCloudflareVaultAPI-gatewayperishableLLM-securitycomparisonpolymathy vs Haystack vs LangChain: Building RAG Pipelines in Rusthttps://skelfresearch.com/blog/cross-cutting/polymathy-vs-langchain/https://skelfresearch.com/blog/cross-cutting/polymathy-vs-langchain/A practical comparison of polymathy, Haystack, and LangChain for building RAG pipelines — when you want a Rust-native async chunking + embedding service, and when the Python frameworks are still the right answer.Tue, 23 Jun 2026 00:00:00 GMTsafe-computingRAGHaystackLangChainLlamaIndexpolymathyRustembedding-pipelinecomparisonroute-switch vs LiteLLM vs Portkey vs OpenRouter: LLM Routing in 2026https://skelfresearch.com/blog/cross-cutting/route-switch-vs-litellm/https://skelfresearch.com/blog/cross-cutting/route-switch-vs-litellm/A practical comparison of route-switch, LiteLLM, Portkey, and OpenRouter for LLM routing — when you need a gateway, when you need a learned router, and when you need both.Tue, 23 Jun 2026 00:00:00 GMTllm-cognitionLLM routingLiteLLMPortkeyOpenRouterMIPROv2cost-qualityroute-switchcomparisonsavanty: Bridging Natural Language and Mathematical Solvers in 2026https://skelfresearch.com/blog/cross-cutting/savanty-vs-nl2opt/https://skelfresearch.com/blog/cross-cutting/savanty-vs-nl2opt/How savanty turns English problem descriptions into mathematically guaranteed solutions. The case for LLM-to-formal-solver pipelines over pure LLM output.Tue, 23 Jun 2026 00:00:00 GMTformal-optimisationconstraint-satisfactionSMT solverZ3OR-ToolsMiniZincGurobisavantyNL2OPTcomparisonSkelf Research: The Independent Lab That Ships Research as Codehttps://skelfresearch.com/blog/cross-cutting/skelf-vs-big-tech/https://skelfresearch.com/blog/cross-cutting/skelf-vs-big-tech/How Skelf Research compares to Big Tech AI labs, AI startups, and individual researchers — what 'independent AI research lab publishing open-source software' means in 2026, and why the methodology is the moat.Tue, 23 Jun 2026 00:00:00 GMTcross-cuttingAI research labopen sourceopen scienceindependent researchhypotheses as softwareSkelf Researchcomparisonslorg vs Algolia vs Meilisearch vs Typesense: Deliberative vs Traditional Searchhttps://skelfresearch.com/blog/cross-cutting/slorg-vs-algolia/https://skelfresearch.com/blog/cross-cutting/slorg-vs-algolia/A practical comparison of slorg, Algolia, Meilisearch, and Typesense for search — when deliberative search (reasoning before retrieval) is the right answer, and when traditional search is still the right answer.Tue, 23 Jun 2026 00:00:00 GMTedge-intelligencesearchAlgoliaMeilisearchTypesenseElasticsearchdeliberative-searchagentic-searchslorgcomparisonukkin vs ChatGPT Operator vs Anthropic Computer Use: On-Device vs Cloud AI Agentshttps://skelfresearch.com/blog/cross-cutting/ukkin-vs-operator/https://skelfresearch.com/blog/cross-cutting/ukkin-vs-operator/A practical comparison of ukkin, ChatGPT Operator, and Anthropic Computer Use for autonomous mobile and desktop AI agents — privacy, autonomy, and the on-device vs cloud trade-off.Tue, 23 Jun 2026 00:00:00 GMTedge-intelligenceagentsmobileautonomyChatGPT OperatorAnthropic Computer Useukkinon-devicecomparisonwaremax: Deterministic Warehouse-Robotics Simulation for RL Researchhttps://skelfresearch.com/blog/cross-cutting/waremax-vs-rawsim/https://skelfresearch.com/blog/cross-cutting/waremax-vs-rawsim/How waremax gives warehouse-robotics research a deterministic simulator, a Gymnasium RL interface, and instrumented delay attribution. Why deterministic beats plausible for reproducible research.Tue, 23 Jun 2026 00:00:00 GMTroboticswarehouse-roboticsRMFSAMRRAWSim-OARENA-Simdeterministic-simulationGymnasiumMaskablePPOwaremaxcomparisonzviz vs gVisor vs Firecracker: Choosing a Sandbox for AI-Generated Codehttps://skelfresearch.com/blog/cross-cutting/zviz-vs-gvisor/https://skelfresearch.com/blog/cross-cutting/zviz-vs-gvisor/A practical comparison of zviz, gVisor, and Firecracker for sandboxing untrusted AI-generated code — performance, security, and operational trade-offs.Tue, 23 Jun 2026 00:00:00 GMTsafe-computingsandboxinggVisorFirecrackerWASMuntrusted-codezvizcomparisonmemorg vs Mem0: Structured Memory vs Vector Memory for LLM Agentshttps://skelfresearch.com/blog/cross-cutting/memorg-vs-mem0/https://skelfresearch.com/blog/cross-cutting/memorg-vs-mem0/A practical comparison of memorg and Mem0 for agent memory — when structured schemas beat vector recall, when they don't, and how to combine them.Mon, 22 Jun 2026 00:00:00 GMTllm-cognitionmemoryagentsMem0LettaZepstructured-memorycomparisonmpl vs MCP: A Compliance and Audit Layer on Top of the Agent Protocolhttps://skelfresearch.com/blog/cross-cutting/mpl-vs-mcp/https://skelfresearch.com/blog/cross-cutting/mpl-vs-mcp/MCP defines how agents talk. mpl defines what correct looks like. Why production agentic systems need both — and how to compose them.Mon, 22 Jun 2026 00:00:00 GMTllm-cognitionMCPA2Aagent-protocolcomplianceauditcomparisonpromptel vs DSPy: Declarative Specification vs Programmatic Compilationhttps://skelfresearch.com/blog/cross-cutting/promptel-vs-dspy/https://skelfresearch.com/blog/cross-cutting/promptel-vs-dspy/A practical comparison of promptel and DSPy for prompt engineering — when to use a declarative specification language, when to use a programmatic compiler, and how they compose.Mon, 22 Jun 2026 00:00:00 GMTllm-cognitionpromptsDSPydeclarativeprompt-optimizationcompilerscomparisonProgrammable Databases: Why We Built liath Twicehttps://skelfresearch.com/blog/safe-computing/programmable-databases-why-we-built-liath-twice/https://skelfresearch.com/blog/safe-computing/programmable-databases-why-we-built-liath-twice/The story of building a Lua-native database in Lua, then rewriting it in Rust with RocksDB — and what the performance tells us.Sat, 04 Apr 2026 00:00:00 GMTsafe-computingdatabaseluarustRocksDBstorage-enginesDeliberative Search: When the Engine Reasons Before It Retrieveshttps://skelfresearch.com/blog/edge-intelligence/deliberative-search/https://skelfresearch.com/blog/edge-intelligence/deliberative-search/Why traditional search retrieves first and ranks later — and how slorg inverts this by understanding intent before fetching results.Thu, 02 Apr 2026 00:00:00 GMTedge-intelligencesearchSvelteKitreasoningintentretrievalPersistent Memory for Long-Running Agentshttps://skelfresearch.com/blog/llm-cognition/persistent-memory-for-agents/https://skelfresearch.com/blog/llm-cognition/persistent-memory-for-agents/What happens when LLM agents need to remember across sessions — structured memory schemas, retrieval strategies, and the memory-context distinction.Wed, 01 Apr 2026 00:00:00 GMTllm-cognitionmemoryagentsretrievalarchitectureVector Search Without the Cloud: memista's SQLite-Backed ANNhttps://skelfresearch.com/blog/safe-computing/vector-search-without-the-cloud/https://skelfresearch.com/blog/safe-computing/vector-search-without-the-cloud/Building approximate nearest-neighbour search on SQLite in pure Rust — and why you might not need a dedicated vector database.Sun, 29 Mar 2026 00:00:00 GMTsafe-computingvector-searchSQLiteANNembeddingsrustEphemeral Credentials and Zero-Trust AI: Rethinking API Securityhttps://skelfresearch.com/blog/cross-cutting/ephemeral-credentials-zero-trust-ai/https://skelfresearch.com/blog/cross-cutting/ephemeral-credentials-zero-trust-ai/Why AI agents need scoped, time-limited credentials — and how perishable implements zero-trust patterns for LLM API access.Wed, 25 Mar 2026 00:00:00 GMTcross-cuttingsecurityzero-trustcredentialsAPIagentsPrompt Lifecycle Management: From Extraction to Deploymenthttps://skelfresearch.com/blog/llm-cognition/prompt-lifecycle-management/https://skelfresearch.com/blog/llm-cognition/prompt-lifecycle-management/A practical framework for managing prompts as versioned dependencies — tackling drift, regression, and reproducibility.Sun, 22 Mar 2026 00:00:00 GMTllm-cognitionpromptsversioningdevopsreproducibilityBetter Rankings with Fewer Comparisons: Multi-Armed Bandits for Efficient Orderinghttps://skelfresearch.com/blog/formal-optimisation/better-rankings-fewer-comparisons/https://skelfresearch.com/blog/formal-optimisation/better-rankings-fewer-comparisons/How compere uses MAB algorithms to rank items effectively with minimal pairwise feedback — applications in search and recommendation.Thu, 19 Mar 2026 00:00:00 GMTformal-optimisationMABrankingbanditsrecommendationsearchFormalising Prompts as First-Class Research Objectshttps://skelfresearch.com/blog/llm-cognition/formalising-prompts-as-first-class-objects/https://skelfresearch.com/blog/llm-cognition/formalising-prompts-as-first-class-objects/Why treating prompts as typed, portable artefacts changes how we reason about LLM behaviour — and how promptel implements this idea.Sun, 15 Mar 2026 00:00:00 GMTllm-cognitionpromptsformal-methodsspecificationportabilityFrom English to Optimal: How savanty Bridges Natural Language and Constraint Solvershttps://skelfresearch.com/blog/formal-optimisation/from-english-to-optimal/https://skelfresearch.com/blog/formal-optimisation/from-english-to-optimal/Describe optimisation problems in plain English and receive mathematically guaranteed solutions — no PhD required.Thu, 12 Mar 2026 00:00:00 GMTformal-optimisationNLPconstraint-satisfactionsolversoptimisationIntelligent LLM Routing: Spending Compute Where It Mattershttps://skelfresearch.com/blog/llm-cognition/intelligent-llm-routing/https://skelfresearch.com/blog/llm-cognition/intelligent-llm-routing/How route-switch uses MIPROv2 to automatically select the right model for each query — balancing cost, quality, and latency.Sun, 08 Mar 2026 00:00:00 GMTllm-cognitionroutingMIPROv2cost-optimisationmodel-selectionAutonomous Mobile Agents: ukkin's Architecture for On-Device AIhttps://skelfresearch.com/blog/edge-intelligence/autonomous-mobile-agents/https://skelfresearch.com/blog/edge-intelligence/autonomous-mobile-agents/Building AI agents that browse, observe, and automate tasks entirely on-device — the autonomy-safety spectrum on mobile.Thu, 05 Mar 2026 00:00:00 GMTedge-intelligenceagentsmobileautonomyflutterautomationBuilding mullama: What We Learned Replacing Ollama from Scratchhttps://skelfresearch.com/blog/llm-cognition/building-mullama/https://skelfresearch.com/blog/llm-cognition/building-mullama/A post-mortem on building a local LLM serving layer — llama.cpp integration, model management, and where existing tools constrain research.Sat, 28 Feb 2026 00:00:00 GMTllm-cognitionllama.cppinferencelocal-llmarchitectureRunning Language Models on Your Phone: The llamafu Experimenthttps://skelfresearch.com/blog/edge-intelligence/running-language-models-on-your-phone/https://skelfresearch.com/blog/edge-intelligence/running-language-models-on-your-phone/What happens when you run a full LLM on mobile hardware with zero cloud dependency — memory, latency, and model quality on consumer devices.Wed, 25 Feb 2026 00:00:00 GMTedge-intelligencemobilellama.cppflutteron-deviceinferenceSandboxing Untrusted Code in Zig: The zviz Architecturehttps://skelfresearch.com/blog/safe-computing/sandboxing-untrusted-code-in-zig/https://skelfresearch.com/blog/safe-computing/sandboxing-untrusted-code-in-zig/How zviz uses Zig's comptime capabilities to build gVisor-inspired sandboxing with near-zero runtime cost.Sat, 21 Feb 2026 00:00:00 GMTsafe-computingzigsandboxinggVisorsecuritycomptimeWhy We Write AI Infrastructure in Rust (and Zig, and Go)https://skelfresearch.com/blog/safe-computing/why-we-write-ai-infrastructure-in-rust/https://skelfresearch.com/blog/safe-computing/why-we-write-ai-infrastructure-in-rust/Language choice as research methodology — how memory-safe, deterministic-performance languages produce falsifiable systems claims.Sat, 14 Feb 2026 00:00:00 GMTsafe-computingrustziggosystems-programmingmethodologyCompiling Trading Signals: sigc and the Quantitative Hypothesis Pipelinehttps://skelfresearch.com/blog/formal-optimisation/compiling-trading-signals/https://skelfresearch.com/blog/formal-optimisation/compiling-trading-signals/From visual signal specification to verified Rust executable — how sigc turns alpha hypotheses into production-ready code in minutes.Sat, 07 Feb 2026 00:00:00 GMTformal-optimisationquantsignalscompilerrustverificationOpen Science in AI: Why We Publish Everythinghttps://skelfresearch.com/blog/cross-cutting/open-science-in-ai/https://skelfresearch.com/blog/cross-cutting/open-science-in-ai/The case for radical openness in AI research — reproducibility, falsifiability, and community trust through 25 open-source projects.Thu, 15 Jan 2026 00:00:00 GMTcross-cuttingopen-scienceopen-sourcereproducibilitymethodology