Tuesday 2025-02-18 Assorted Links
Assorted Links links
Published: 2025-02-18
Tuesday 2025-02-18 Assorted Links

Assorted links for Tuesday, Febuary 18:

  1. A brief and incomplete comparison of memory corruption detection tools

    ASAN detects a lot more types of memory errors, but it requires that you recompile everything. This can be limiting if you suspect that the problem is coming from a component you cannot recompile (say because you aren’t set up to recompile it, or because you don’t have the source code). Valgrind and AppVerifier have the advantage that you can turn them on for a process without requiring a recompilation.

  2. Why Mocks Fail: Real-Environment Testing for Microservices
    • Use mocks for edge cases and scenarios requiring controlled inputs.
    • Leverage real environments to validate integration flows, complex API behaviors and performance characteristics against real dependencies.
  3. Emerging Patterns in Building GenAI Products:
    • Direct Prompting: Send prompts directly from the user to a Foundation LLM
    • Embeddings: Transform large data blocks into numeric vectors so that embeddings near each other represent related concepts
    • Evals: Evaluate the responses of an LLM in the context of a specific task
    • Hybrid Retriever: Combine searches using embeddings with other search techniques
    • Query Rewriting: Use an LLM to create several alternative formulations of a query and search with all the alternatives
    • Reranker: Rank a set of retrieved document fragments according to their usefulness and send the best of them to the LLM.
    • Retrieval Augmented Generation (RAG): Retrieve relevant document fragments and include these when prompting the LLM
  4. How Meta discovers data flows via lineage at scale

    In order to build high-quality data lineage, we developed different techniques to collect data flow signals across different technology stacks: static code analysis for different languages, runtime instrumentation, and input and output data matching, etc.

  5. Sam Altman lays out roadmap for OpenAI’s long-awaited GPT-5 model

    GPT-5 will be a system that brings together features from across OpenAI’s current AI model lineup, including conventional AI models, SR models, and specialized models that do tasks like web search and research.