Tag
The article argues that duplication is cheaper than the wrong abstraction, and advises developers to avoid forcing abstractions that later become complicated with conditionals.
This paper introduces WorkflowView, a framework that uses large language models to abstract low-level, noisy user action sequences into interpretable high-level activities, demonstrating effectiveness across browser logs, MOOC dropout prediction, and privacy-preserving document workflow analysis.
Introduces an abstract visual programming language for composing animated clocks using vectors, scalars, glyphs, and habitats, with a focus on artistic expression and time representation.
Anthropic's Boris Cherny points out that programming is moving towards a higher level of abstraction, with workflows shifting from manually writing code to letting Claude make autonomous decisions, and predicts that the next paradigm shift will arrive this year.
A reflective thread arguing that success in frontier AI labs requires the ability to operate without a map, where research is about navigating uncertainty and engineering is about compressing complex systems into useful abstractions.
The author uses personal experience to introduce a tutorial on architect thinking in the AI era, emphasizing that the ability to understand the underlying essence when abstraction leaks is more critical than tool usage, and shares two modes: assembly thinking and object-oriented thinking.
Introduces context-dependent argumentation frameworks (CDAFs) that model how an agent can strategically influence which attacks succeed by choosing a context, enabling manipulation scenarios not possible in value-based argumentation. Defines the ACTIVATION-MANIPULATION decision problem and provides baseline complexity bounds.
This paper presents an Answer Set Programming (ASP) based implementation of the CARCASS framework for constructing abstractions in reinforcement learning, demonstrating its effectiveness on Blocks World and Minigrid domains.
An AI researcher asserts that MCP (Model Context Protocol) will be recognized as fundamentally important for agents, not just for tool connection but for enabling new abstractions, and mentions a new self-improving system built on it.
The article proposes treating AI prompts as executable logic using visual nodes and logic gates, similar to visual programming languages, and introduces a prototype called Prompt Logic Gates (PLG).
The article draws parallels between how AI is deskilling programming today and how JavaScript frameworks deskilled frontend development in the past, arguing that both represent a shift to a higher level of abstraction that reduces barriers to entry but weakens workers' bargaining power.
CODA introduces a GPU kernel abstraction that rewrites transformer computations as GEMM-plus-epilogue programs, reducing memory-bound operations and improving efficiency in training.
This paper presents the first theoretical model for out-of-distribution generalization in reinforcement learning, showing that smaller abstract state spaces enable cross-scale generalization in POMDPs.
Martin Fowler reflects on AI’s impact on code quality, emphasizing that human laziness drives crisp abstractions while LLMs risk bloating systems with unnecessary complexity.
This article argues that a comprehensive specification is not equivalent to code, because a spec defines a set of possible implementations while code is one concrete instance. It discusses the role of abstraction and why programmers are still needed to write specs even with automated code generation.
OpenAI discovers multimodal neurons in CLIP that respond to the same concept across different modalities (visual, symbolic, textual), mirroring biological neurons and explaining the model's robustness on challenging vision tasks. This interpretability research provides insights into how vision-language models organize and represent abstract concepts.