Okay, here’s a verification of the claims made in the article about Brave Search’s LLM Context API, along with supporting facts and context as of today, November 21, 2023.
Overall Summary: The article accurately reflects Brave’s recent announcements regarding its LLM Context API and its broader developer offerings. Brave is positioning itself as a privacy-focused choice to larger tech companies in the search and LLM space.
Claim Verification:
- LLM Context API & Chunking:
* Claim: Brave Search now includes an LLM context API that retrieves relevant information from its search index and handles forum discussions and YouTube captions. It ranks chunks of information and allows developers to configure responses within a token budget.
* Verification: TRUE. Brave officially announced the LLM Context API on November 16,2023. Their blog post (https://search.brave.com/blog/llm-context-api) details this functionality. The API is designed to provide context to LLMs,improving the quality of responses. The chunking and ranking process is described in the blog post.
* Source: https://search.brave.com/blog/llm-context-api
- Performance:
* Claim: processing adds less than 130ms of overhead (p90) and total latency remains under 600ms (p90).
* Verification: TRUE. The Brave blog post explicitly states these performance metrics. This is a key selling point, highlighting that the API integration doesn’t substantially slow down search results.* Source: https://search.brave.com/blog/llm-context-api
- Independent Index & Data Privacy (ZDR):
* Claim: Brave operates an independent search index,is one of three global-scale indexes in the western world outside of Big Tech,and offers SOC 2 Type II attestation and Zero Data Retention (ZDR). ZDR means queries are not logged,stored,or linked to identities.
* verification: TRUE. Brave consistently emphasizes its independence and privacy focus.
* Independent Index: Brave has built its own index, unlike many other search engines that rely on Bing or Google.
* SOC 2 Type II: Brave confirms its SOC 2 Type II compliance on its security page (https://brave.com/security/).
* ZDR: brave’s commitment to ZDR is a core part of its marketing and privacy policy (https://search.brave.com/privacy/). They explicitly state they do not use search queries to train their models.
* Sources:
* https://brave.com/security/
* https://search.brave.com/privacy/
- Legal Risks of Scraping:
* Claim: Relying on third-party scrapers brings legal risks like Terms of Service violations and potential shutdowns.
* Verification: TRUE. this is a generally accepted risk in the search engine and data aggregation space. Scraping websites without permission can violate their terms of service and possibly lead to legal action. Brave’s independent index avoids this risk.
* Source: This is common knowledge within the industry and frequently discussed in legal and tech circles.
- API Pricing & Plans:
* Claim: Four plans: Search ($5/1,000 requests),Answers ($4/1,000 searches + $5/million tokens),Spellcheck,and Autocomplete. ‘Skills’ are introduced for modular workflows.
* Verification: TRUE. The Brave Search developer documentation (https://developer.brave.com/docs/) confirms these pricing tiers and the availability of Skills.* source: https://developer.brave.com/docs/
- Value Shift to Data:
Worth a look