Bilingual

AI Is a Bad Tool
8 days ago
- AI can be useful as a data distiller, condensing information retrieval and refinement steps that were previously manual.
- Using AI to generate code is inefficient; it often produces unverified, opaque outputs that lack proper abstraction and increase repetition.
- AI's opacity makes 'prompt engineering' questionable, as manipulating a black-box system without understanding its source code is unstable and unreliable.
- AI exposes a lack of proper abstraction in software development, revealing that many programming tasks could be automated by better tools or frameworks.
- The fear of AI making humans irrelevant stems from its ability to handle trivial, unabstracted work, highlighting that many software jobs were already inefficient.
- Dealing with AI's impact may require focusing on passion-driven work, full understanding, and proper abstraction to avoid reliance on AI for non-problems.
Code Was Our Medium for Thought
6 days ago
- The author created a beeswarm chart showing AI model releases over time, with each mark representing a model and its horizontal position indicating the publication date.
- They discuss the initial excitement of AI-generated code feeling like magic, but overuse leads to producing low-quality work and overwhelming PR reviews, forcing a choice between being a bottleneck or relying on agents with half-hearted efforts.
- Reflecting on the past, the author notes that coding involved problem exploration, decision-making, and team collaboration, with code evolving from fuzzy ideas to clear implementations as a medium for thought.
- They acknowledge that while reverting to old ways isn't desirable, achieving a flow state requires new workflows and tooling, similar to how IDEs evolved over decades to support coding practices.
- The history of programming is traced from punch cards to assembly, compilers, scripting languages, and libraries, each step abstracting details to allow faster and more ambitious creation.
- Critiquing current AI use, the author argues that relying on prompts and raw code reviews leads to unreadable code and agent overuse, whereas modern AI models, rooted in language and translation research, can robustly translate code across mediums.
- They propose that AI agents could create custom whiteboards or playgrounds for exploration, starting with images or prototypes instead of code, to maintain intuition, mental mapping, and team alignment.
- The conclusion emphasizes that if AI is only used to edit code faster, it risks losing the thinking process; instead, tools should enable human-centric thinking to make coding feel grounded and engaging.
Wolfram LLM Benchmarking Project
2 days ago
- The Wolfram LLM Benchmarking Project focuses on evaluating large language models (LLMs) using Wolfram Language for a code generation task.
- The task converts English-language specifications into Wolfram Language code, based on exercises from 'An Elementary Introduction to the Wolfram Language'.
- The project utilizes tools to assess functional correctness, with test cases previously completed by millions of humans online.
- Results and previous versions are accessible in computable form via the Wolfram Data Repository.
- For LLM developers, Wolfram offers access to datasets, tools, and opportunities for LLM inclusion in benchmarking.

Hasty Briefsbeta

#code generation

AI Is a Bad Tool

Code Was Our Medium for Thought

Wolfram LLM Benchmarking Project