Software Engineering Radio - The Podcast for Professional Software Developers · SE-Radio Team

SE Radio 666: Eran Yahav on the Tabnine AI Coding Assistant

·1 hr 2 min·5 clips

Tabnine's context engine ensures AI generates code tailored to your organization's specific practices and environment.

The frame stays concrete. Gregory Kapphammer introduces Tab9 as an AI coding assistant that uses large language models for code explanation, test case generation, and related work. The guest starts with how the tool shows up across the SDLC: completions, chat, code creation, review, refactoring, translation, tests, documentation, and explanation. That keeps the conversation close to daily engineering work instead of floating around as generic AI talk. Then the interview tightens. Kapphammer asks why Rust was chosen for Tab9's implementation, and the answer is practical, not doctrinal. The choice is described as historical. Rust comes across as useful for cross-platform and cross-architecture deployment without dragging a large runtime across operating systems and machines. Local inference is part of the story. Earlier versions did clever work on the user's machine, which meant dealing with CPU intrinsics and other low-level concerns. The tradeoff is familiar: for this kind of systems work, the comparison becomes C++ or Rust. The guest says plainly that he prefers Rust and does not turn it into a language fight. Kapphammer then resets the vocabulary. He asks for definitions of an AST and a dependence graph, which keeps the interview usable for listeners who know code but may not have the guest's compiler background. The AST explanation is careful enough without getting stiff. Programs are parsed against a language grammar so the tool can work with something better than raw text. Some details fall away. What remains is a tree-shaped view of code, with functions, blocks, conditionals, loops, and other structure available for analysis. That matters because many program operations are easier on structured representations than on plain text. The ending returns to practice. The guest pushes listeners to think about building codebases that future AI systems can use, and the episode lands as a technical conversation about mechanisms, not product mist.

As heard by us

A practical look at how an AI coding assistant is built, not just sold.

SE Radio treats Tab9 as a practical look at what an AI coding assistant actually does across the software lifecycle. Aaron Yehove walks through code completion, chat-based creation, code review, refactoring, translation between languages, and test generation, then gets into the…

Read the full review in PlayNext →

Why you'd press play

You want the engineering case for an AI coding assistant, not a sales pitch.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Software Engineering Radio - The Podcast for Professional Software Developers