Is this the Future of Smart Homes?

Harper 5.2 helps architecture and platform leaders improve performance, control infrastructure costs, secure production workloads, and reduce operational complexity by bringing faster data access, agentic capabilities, scheduling, backup, routing, and protection into one unified runtime.

News

Harper 5.2 helps architecture and platform leaders improve performance, control infrastructure costs, secure production workloads, and reduce operational complexity by bringing faster data access, agentic capabilities, scheduling, backup, routing, and protection into one unified runtime.

Aleks Haugom

Senior Manager of GTM

News

Harper 5.2: More Throughput per Node, Fewer Systems Around It

Harper 5.2 helps architecture and platform leaders improve performance, control infrastructure costs, secure production workloads, and reduce operational complexity by bringing faster data access, agentic capabilities, scheduling, backup, routing, and protection into one unified runtime.

Aug 2026

News

Harper 5.2: More Throughput per Node, Fewer Systems Around It

Harper 5.2 helps architecture and platform leaders improve performance, control infrastructure costs, secure production workloads, and reduce operational complexity by bringing faster data access, agentic capabilities, scheduling, backup, routing, and protection into one unified runtime.

News

Harper 5.2: More Throughput per Node, Fewer Systems Around It

Personalization is a data-delivery problem. Every architectural choice reduces to two distances: compute to user, and compute to fresh data. This piece maps five real architectures against both axes, scored on a concrete retailer workload where stale or slow data breaks the business.

5 Architectures for Web Personalization

Blog

Personalization is a data-delivery problem. Every architectural choice reduces to two distances: compute to user, and compute to fresh data. This piece maps five real architectures against both axes, scored on a concrete retailer workload where stale or slow data breaks the business.

Aleks Haugom

Senior Manager of GTM

Personalization is a data-delivery problem. Every architectural choice reduces to two distances: compute to user, and compute to fresh data. This piece maps five real architectures against both axes, scored on a concrete retailer workload where stale or slow data breaks the business.

5 Architectures for Web Personalization

Jul 2026

Personalization is a data-delivery problem. Every architectural choice reduces to two distances: compute to user, and compute to fresh data. This piece maps five real architectures against both axes, scored on a concrete retailer workload where stale or slow data breaks the business.

5 Architectures for Web Personalization

Personalization is a data-delivery problem. Every architectural choice reduces to two distances: compute to user, and compute to fresh data. This piece maps five real architectures against both axes, scored on a concrete retailer workload where stale or slow data breaks the business.

5 Architectures for Web Personalization

Harper 5.1 adds int8 quantization, dropping index size by roughly 3x, search throughput improves roughly 5x, and p99 latency falls from ~9s to ~0.5s under load. The tradeoff is approximately 1% recall degradation before reranking.

5x Faster HNSW Vector Search with int8 Quantization

JavaScript

Tutorial

Harper 5.1 adds int8 quantization, dropping index size by roughly 3x, search throughput improves roughly 5x, and p99 latency falls from ~9s to ~0.5s under load. The tradeoff is approximately 1% recall degradation before reranking.

Kris Zyp

SVP of Engineering

Harper 5.1 adds int8 quantization, dropping index size by roughly 3x, search throughput improves roughly 5x, and p99 latency falls from ~9s to ~0.5s under load. The tradeoff is approximately 1% recall degradation before reranking.

5x Faster HNSW Vector Search with int8 Quantization

Kris Zyp

Jun 2026

Harper 5.1 adds int8 quantization, dropping index size by roughly 3x, search throughput improves roughly 5x, and p99 latency falls from ~9s to ~0.5s under load. The tradeoff is approximately 1% recall degradation before reranking.

5x Faster HNSW Vector Search with int8 Quantization

Kris Zyp

Harper 5.1 adds int8 quantization, dropping index size by roughly 3x, search throughput improves roughly 5x, and p99 latency falls from ~9s to ~0.5s under load. The tradeoff is approximately 1% recall degradation before reranking.

5x Faster HNSW Vector Search with int8 Quantization

Kris Zyp

Harper runs live, personalized apps 4–14x faster than Vercel by unifying data, logic, cache, and messaging in one runtime.

Vercel vs. Harper Benchmark Test

Comparison

Harper runs live, personalized apps 4–14x faster than Vercel by unifying data, logic, cache, and messaging in one runtime.

Harper

Harper runs live, personalized apps 4–14x faster than Vercel by unifying data, logic, cache, and messaging in one runtime.

Vercel vs. Harper Benchmark Test

Harper

Jun 2026

Harper runs live, personalized apps 4–14x faster than Vercel by unifying data, logic, cache, and messaging in one runtime.

Vercel vs. Harper Benchmark Test

Harper

Harper runs live, personalized apps 4–14x faster than Vercel by unifying data, logic, cache, and messaging in one runtime.

Vercel vs. Harper Benchmark Test

Harper