Przejdź do treści Przejdź do indeksu witryny

Zarób do 2000 USD, kupując 50 USD w kryptowalutach¹

Zarejestruj się

Coinbase Transparency Report — H2 2020

Do May 5, 2021 • Czytanie na 2 min

Coinbase is proud to publish our second Transparency Report, covering the period from July 1, 2020 through December 31, 2020. The goal of this report is to keep our customers informed about requests for customer information we receive from government agencies and law enforcement, along with where they come from, and how we respond to them. We intend to provide additional insights and observations in future reports as we gather more data.

Coinbase Transparency Report

Period: July 1, 2020 — December 31, 2020

Total Requests Received: 2,313

Key Takeaways:

There were a total of ~4,227 requests in 2020, 54% of those requests made during the time period covered by this report.
Similar to the first half of 2020:
- ~90% of requests are from the U.S., U.K., and Germany
- The FBI (Federal Bureau of Investigation) and HSI (Homeland Security Investigations) comprise the majority of U.S. law enforcement information requests at ~50.9%
- ~51% of total requests during this period came from U.S. agencies, down 7% from the prior period
There was a ~20% increase in requests from France

Law Enforcement and Agency Information Requests

Law Enforcement Information Requests Ranking

US Law Enforcement and Agency Information Requests

For additional background and context, see our inaugural report here.

Najnowsze historie

Part 3/3 Owning Intelligence Series: How Coinbase Built, Evaluated, and Post-Trained a Fraud Agent

Yao Ma, Xuwei Tan, Akshit Trehan, Aman Choudhary, Anyscale Team: Eric Tang, Sumanth Hegde, Seiji Eicher, Jess Kong

Oct 8, 2026

TL;DR: Using SkyRL on Anyscale, we post-trained Qwen3.5-9B for Onramp fraud detection and served it with Ray Serve LLM and vLLM. On our benchmark, it exceeded Opus 4.5 across all four detection metrics, including 9.6% higher F1. In online inference, we reduced end to end latency by 55%.

Part 2/3 Owning Intelligence Series: How Coinbase Built, Evaluated, and Post-Trained a Fraud Agent

Yao Ma, Xuwei Tan

Oct 7, 2026

TL;DR: Newer LLMs are not necessarily better at domain-specific tasks. On our fraud benchmark, newer Opus, Sonnet, and GPT versions all had lower recall, F1, and dollar-weighted recall—even when precision improved. The lesson: choose models using high-quality, domain-specific benchmarks, with reproducible evaluation and ideally deterministic scoring against verified outcomes. These findings also motivated our next step: post-training open-weight models for greater control over quality, cost, and deployment.

Migrating 300+ namespaces from self-hosted Temporal to Temporal Cloud

Ian Yap

Oct 1, 2026

TL;DR: We moved more than 300 Temporal namespaces, across more than 300 services, from a cluster we ran ourselves onto Temporal Cloud. You cannot finish the move by repointing services at Temporal Cloud while workflows are still running on the old one. Where teams could pause new work, we waited for in-flight runs to finish and then switched; for trading, payments, and long-running work, we ran the old cluster and Temporal Cloud, and moved new workflows over in steps, with rollback as a config change. Private connectivity, per-namespace service allowlists, and encrypted payloads were in place before the first production namespace moved.

Zastrzeżenia: Handel instrumentami pochodnymi za pośrednictwem platformy Coinbase Advanced jest oferowany uprawnionym klientom z EOG przez Coinbase Financial Services Europe Ltd. (licencja CySEC 374/19). Aby uzyskać dostęp do instrumentów pochodnych, klient musi przejść nasze standardowe kontrole, aby ustalić spełnienie kryteriów kwalifikacyjnych i odpowiedniość tego produktu.