| • Introduction to static checkers
|
| • Rubber duck debugging technique and its uses
|
| • Importance of taking breaks in problem-solving (including going for a walk or engaging in meditation)
|
| • Matan Peled's background: Ph.D. candidate, research on meta programming and static analysis, experience working in industry and academia
|
| • Discussion of programming language design and creating animations using a custom language
|
| • The speaker's past project didn't meet expectations due to lack of usability and tests.
|
| • The importance of being practical and useful vs. exploring new ideas without immediate utility.
|
| • Startup culture and its emphasis on flexibility and taking risks.
|
| • Static analysis, including its purpose and challenges in dynamic languages vs. typed languages.
|
| • The theoretical basis for static analysis, citing the halting problem and Rice's theorem.
|
| • Meta-programming using static analysis as a focus of the speaker's research.
|
| • Static analysis techniques for code
|
| • Machine learning (ML) applications to code, such as GitHub Copilot
|
| • Formal methods for static analysis, including type checking
|
| • Limitations of ML approaches, such as lack of explicit knowledge about code structure and syntax
|
| • Potential next steps in AI-generated code development, including creation of static and dynamic checkers
|
| • Discussion of taint analysis as an important tool in recent years
|
| • Mention of static checking and its use to prevent bugs early on in the development process
|
| • Examples of types of bugs that can be caught through static checking (e.g. memory allocation issues, multi-threading problems)
|
| • Overview of static checkers for Go, including Staticcheck and Errorcheck
|
| • Goal of creating a custom static analysis tool with user-definable rule sets
|
| • Challenges in performing points-to analysis, including aliasing and dynamic type determination
|
| • Comparison of static analysis at compile-time vs. runtime debugging
|
| • Reverse debuggers allow stepping back and going back in time to see what happened before
|
| • They keep snapshots of operations, but only at certain points (e.g. before input/output)
|
| • Dynamic analysis involves using information from compile-time plus real-time values
|
| • Instrumenting is a form of dynamic analysis that involves adding code to track program behavior
|
| • Structured logging can be seen as a form of dynamic debugging or tracing
|
| • Go's open-source nature allows for understanding and use of its toolchain packages
|
| • Self-hosting compilers, where compilers are written in the language they compile
|
| • Concept of "Reflections on trusting trust" and its implications for compiler security
|
| • Difficulty of detecting backdoors in compiled code
|
| • Gödel, Escher, Bach book recommendation for exploring self-referential concepts
|
| • Unpopular opinion: static analysis doesn't work beyond a certain complexity
|
| • Software engineer job security due to limitations of AI replacing human tasks
|
| • Next-level abstraction with AI guiding programming and automation of tasks
|
| • Research on synthesis and machine learning-based program generation
|
| • Limitations of current AI capabilities in optimizing code and replacing human programmers
|
| • No code/no bugs as a future goal
|
| • Difficulty of writing comprehensive tests and ensuring they don't contradict each other
|
| • Using static analysis to check for test contradictions and identify potential issues
|
| • Pure functions and their benefits in languages like Rust compared to Go's ability to have side effects in methods and functions
|
| • Limitations of static analysis when dealing with input or unknown values |
