knowledge-work

LlamaIndex launches granular bounding boxes in LlamaParse, enabling visual citations for every word in a document to allow human audit of exact numbers and figures.

Anthropic releases Fable 5, claiming it is state-of-the-art on key benchmarks in software engineering, science, knowledge work, and vision, exceeding all previously available models.

This article presents findings from a study by Perplexity and Harvard Business School on how AI agents like Perplexity's Computer reshape knowledge work, showing increased autonomy, efficiency, and scope at lower cost.

This study uses Perplexity production data to analyze how AI agents reshape knowledge work, finding that agents reduce time and cost by over 87%, improve quality, and expand the scope of automated tasks.

Kimi Work is a desktop AI agent that can run 300 agents in parallel locally, with browser automation and scheduled tasks, aiming to boost productivity for solo knowledge workers and agencies.

This study uses production data from Perplexity to compare AI agents versus conversational assistants, finding that agents reduce completion time by 87% and costs by 94% while expanding the scope and quality of knowledge work.

An opinion piece argues that AI has not taken meaningful jobs but rather exposed the pointlessness of many knowledge-worker roles, while hands-on jobs like nursing and plumbing remain AI-proof.

OpenAI introduces new role-specific plugins for Codex, along with annotations and the ability to create interactive apps, expanding the tool's use to non-developers in various fields.

OpenAI announces new role-specific plugins for Codex, expanding its use beyond developers to analysts, marketers, and other knowledge workers, along with features like annotations and app creation.

OpenAI's new report shows Codex expanding beyond coding to become a productivity tool for knowledge workers, with over 5M weekly active users and rapid growth in non-developer tasks like data analysis and workflow automation.

A detailed guide on using OpenAI's Codex as an operating system for knowledge work, including setup, workflows, and a seven-day starter plan, written using Codex itself.

This paper proposes a three-step framework for designing and reporting benchmarks for knowledge work AI, emphasizing alignment between benchmark tasks and real-world work activities. It derives 18 work activities from the O*NET database and analyzes three existing benchmarks (GDPval, OfficeQA Pro, APEX-SWE) to demonstrate gaps between benchmark scores and actual work capability.

Every's CEO Dan Shipper points out with real data that after fully embracing AI agents, the company has actually increased hiring, challenging the common expectation that AI will massively replace jobs, and wrote an article to explain this paradox.

This paper presents QuestBench, a benchmark built by students to evaluate deep research systems across humanities and social science domains. Results show that even advanced systems like GPT-5.5 pass only 57.58% of questions, highlighting failures in trustworthiness.

Andrej Karpathy discusses the limitations of current AI models, the importance of human skill-building over outsourcing thinking, and his vision for a new educational platform inspired by Starfleet Academy.

Jason Liu shares how he uses OpenAI's Codex for knowledge work beyond coding, leveraging durable threads, voice input, and steering to integrate coding agents into his broader workflow.

Jason Liu shared how to use Codex as a central hub to handle knowledge work across tools, such as slides, PDFs, spreadsheets, to maximize its value.

Jason Liu shares his workflow primitives for using Codex effectively, including durable threads, voice input, and steering to extend AI agents beyond coding into knowledge work.

A discussion on how knowledge work revolves around persuasion, featuring a quote from Dwarkesh Patel about the conflation of intelligence and power.

Jerry Liu discusses how AI agents in finance rely on high-quality context engineering from documents, covering use cases like invoice processing and equity research, and shares workshop slides and a repository for building document parsing pipelines with human-in-the-loop review.