1. A Philosophy of Software Design — John Ousterhout
2. Clean Architecture — Robert C. Martin
3. Clean Code — Robert C. Martin
4. Code Complete — Steve McConnell
5. Designing Data-Intensive Applications — Martin Kleppmann
6. Domain-Driven Design — Eric Evans
7. Domain-Driven Design Distilled — Vaughn Vernon
8. Implementing Domain-Driven Design — Vaughn Vernon
9. Patterns of Enterprise Application Architecture — Martin Fowler
10. Refactoring — Martin Fowler
11. Release It! — Michael T. Nygard
12. The Pragmatic Programmer — Andrew Hunt and David Thomas
13. Working Effectively with Legacy Code — Michael Feathers

"I wanted to know whether Gemma 4 could replace a cloud model for my day-to-day agentic coding."

cross-posted from: https://mander.xyz/post/50244557

The benchmark is a set of handcrafted 2d puzzle games that are easy to solve by humans, but require features like skill acquisition and long-term planning by agents.___

Dafny is a good intermediate step for LLM generated code.

this is the abstract of the paper:

Using large language models (LLMs) to generate source code from natural language prompts is a popular and promising idea with a wide range of applications. One of its limitations is that the generated code can be faulty at times, often in a subtle way, despite being presented to the user as correct. In this paper, we explore ways in which formal methods can assist with increasing the quality of code generated by an LLM. Instead of emitting code in a target language directly, we propose that the user guides the LLM to first generate an opaque intermediate representation, in the verification-aware language Dafny, that can be automatically validated for correctness against agreed on specifications. The correct Dafny program is then compiled to the target language and returned to the user. All user-system interactions throughout the procedure occur via natural language; Dafny code is never exposed. We describe our current prototype and report on its performance on the HumanEval Python code generation benchmarks.

Some issues can be prevented when vibe-coding, but LLMs find a way of messing up anyway.