mirror of
https://github.com/hanzo-apps/hanzo.ai.git
synced 2026-10-28 07:32:52 +00:00
228 lines
6.3 KiB
JSON
228 lines
6.3 KiB
JSON
{
|
|
"source": "https://papers.hanzo.ai",
|
|
"fetched": "2026-09-30",
|
|
"papers": [
|
|
{
|
|
"slug": "hanzo-context-memory",
|
|
"title": "Linkage, Not Similarity",
|
|
"subtitle": "What long-term agent memory has to retrieve, and how it is measured",
|
|
"date": "Sep 2026",
|
|
"tags": [
|
|
"Benchmarks",
|
|
"Memory",
|
|
"Retrieval",
|
|
"LoCoMo"
|
|
],
|
|
"pdf": "/pdfs/hanzo-context-memory.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-ai-chain",
|
|
"title": "Hanzo AI Chain (AIVM)",
|
|
"subtitle": "Useful-Work Mining, the Native AI Coin, and Post-Quantum Omnichain Settlement",
|
|
"date": "Aug 2026",
|
|
"tags": [
|
|
"Proof of AI",
|
|
"A-Chain",
|
|
"P3Q",
|
|
"AI Mining"
|
|
],
|
|
"pdf": "/pdfs/hanzo-ai-chain.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-cloud-economics",
|
|
"title": "Cloud Economics of a Sovereign OSS Stack",
|
|
"subtitle": "A Regulated Capital-Markets Migration Case Study",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Sovereign AI",
|
|
"Cloud Economics",
|
|
"Sovereign Stack",
|
|
"OSS"
|
|
],
|
|
"pdf": "/pdfs/hanzo-cloud-economics.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-native-stack-thesis",
|
|
"title": "One Native Stack for Private, Continuously-Learning AI",
|
|
"subtitle": "Inference at the Bandwidth Wall, On-Device QLoRA, One Engine",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Inference",
|
|
"On-Device",
|
|
"Privacy"
|
|
],
|
|
"pdf": "/pdfs/hanzo-native-stack-thesis.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-rocm-inference",
|
|
"title": "Native ROCm Inference on a Consumer RDNA3.5 APU",
|
|
"subtitle": "Reaching llama.cpp Decode Parity via a Unified 1-bit-to-Full Quant Core",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"ROCm",
|
|
"Inference",
|
|
"Benchmarks"
|
|
],
|
|
"pdf": "/pdfs/hanzo-rocm-inference.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-cross-backend-inference",
|
|
"title": "Edge LLM Inference Across Commodity Accelerators",
|
|
"subtitle": "Measured AMD, NVIDIA, and Apple Performance",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Inference",
|
|
"Edge",
|
|
"Benchmarks"
|
|
],
|
|
"pdf": "/pdfs/hanzo-cross-backend-inference.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-native-training",
|
|
"title": "Native Training in the Hanzo Engine",
|
|
"subtitle": "One Engine, One Quant Format, One Backend",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Training",
|
|
"QLoRA",
|
|
"On-Device"
|
|
],
|
|
"pdf": "/pdfs/hanzo-native-training.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-continuous-learning-privacy",
|
|
"title": "Continuously-Learning Private AI",
|
|
"subtitle": "Self-Improvement and Privacy in the Hanzo Native Stack",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Privacy",
|
|
"Continual Learning",
|
|
"On-Device"
|
|
],
|
|
"pdf": "/pdfs/hanzo-continuous-learning-privacy.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-platform-infra-desktop",
|
|
"title": "Economical Serving and the hanzo.ai Platform",
|
|
"subtitle": "A Bandwidth-Optimal Unified Train-and-Serve Runtime",
|
|
"date": "Jun 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Economics",
|
|
"Platform",
|
|
"On-Device"
|
|
],
|
|
"pdf": "/pdfs/hanzo-platform-infra-desktop.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-unified-tenant-cloud",
|
|
"title": "The Unified Sovereign-Tenant Cloud",
|
|
"subtitle": "Linear Shared-Nothing Scaling via Per-Tenant SQLite and In-Process Composition",
|
|
"date": "Jul 2026",
|
|
"tags": [
|
|
"Cloud",
|
|
"Multi-Tenancy",
|
|
"SQLite",
|
|
"Benchmarks"
|
|
],
|
|
"pdf": "/pdfs/hanzo-unified-tenant-cloud.pdf"
|
|
},
|
|
{
|
|
"slug": "refutation-driven-perf",
|
|
"title": "Refutation-Driven Performance Engineering",
|
|
"subtitle": "An Empirical Study of a Multi-Week GPU Kernel Campaign",
|
|
"date": "Jul 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"GPU Kernels",
|
|
"Benchmarks",
|
|
"Methodology"
|
|
],
|
|
"pdf": "/pdfs/refutation-driven-perf.pdf"
|
|
},
|
|
{
|
|
"slug": "evolutionary-schedule-search",
|
|
"title": "Evolutionary Schedule Search in a One-Source Kernel DSL",
|
|
"subtitle": "And the Continual-Improvement Loop It Anchors",
|
|
"date": "Jul 2026",
|
|
"tags": [
|
|
"Engine",
|
|
"Autotuning",
|
|
"Evolutionary Search",
|
|
"Kernels"
|
|
],
|
|
"pdf": "/pdfs/evolutionary-schedule-search.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-router",
|
|
"title": "Hanzo Router",
|
|
"subtitle": "Memory-Aware, Local-First LLM Routing with a Learned SLO-Constrained Policy",
|
|
"date": "Jul 2026",
|
|
"tags": [
|
|
"Router",
|
|
"Routing",
|
|
"Economics",
|
|
"On-Device"
|
|
],
|
|
"pdf": "/pdfs/hanzo-router.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-dormant-agents",
|
|
"title": "What a Dormant Agent Costs",
|
|
"subtitle": "State, Execution and Isolation Measured Separately at a Fleet of One Million",
|
|
"date": "Sep 2026",
|
|
"tags": [
|
|
"Agents",
|
|
"Cost",
|
|
"Isolation",
|
|
"Benchmarks"
|
|
],
|
|
"pdf": "/pdfs/hanzo-dormant-agents.pdf"
|
|
},
|
|
{
|
|
"slug": "hanzo-multi-hop-retrieval",
|
|
"title": "Hops Do Not Fix Multi-Hop Retrieval",
|
|
"subtitle": "A Negative Result on Long-Horizon Conversational Memory, and What the Failure Actually Is",
|
|
"date": "Sep 2026",
|
|
"tags": [
|
|
"Memory",
|
|
"Retrieval",
|
|
"LoCoMo",
|
|
"Benchmarks"
|
|
],
|
|
"pdf": "/pdfs/hanzo-multi-hop-retrieval.pdf"
|
|
},
|
|
{
|
|
"slug": "locomo-subject-scope",
|
|
"title": "Retrieval Scoped by Subject",
|
|
"subtitle": "A factorial measurement of conversational memory on LoCoMo, and the negative controls that locate the effect",
|
|
"date": "Sep 2026",
|
|
"tags": [
|
|
"Memory",
|
|
"Retrieval",
|
|
"LoCoMo",
|
|
"Negative Controls"
|
|
],
|
|
"pdf": "/pdfs/locomo-subject-scope.pdf"
|
|
},
|
|
{
|
|
"slug": "network-whitepaper",
|
|
"title": "Hanzo Network Whitepaper",
|
|
"subtitle": "L1 Blockchain for Decentralized AI Compute",
|
|
"date": "Oct 2024",
|
|
"tags": [
|
|
"Blockchain",
|
|
"Infrastructure",
|
|
"Network",
|
|
"Architecture"
|
|
],
|
|
"pdf": ""
|
|
}
|
|
]
|
|
}
|