35 stories
Tool output was 84% of a coding agent's context in a September 2026 study. How to tell context is the problem, and where to route the work instead.
OpenSpec is the pick for most: MIT, 38 agents, half Spec Kit's tokens in one test. Eight spec-driven tools, and what each adds over a plain AGENTS.md.
Researchers from Meta, MIT, and the University of Washington introduce Context Language Models (CLMs), a new approach that enables language models to manage and edit their own context rather than relying on predefined mechanisms for summarization, compression, and information retrieval, reporting…
Beyond basic ReAct loops: six production-oriented architectures for building autonomous, reliable, and scalable LLM agents. The post Agentic Systems: A Practitioner's Guide to 6 Advanced Architectural Patterns appeared first on Towards Data Science.
Design skills load 316 to 35,150 tokens when they trigger. What a non-code skill is, where it sits next to AGENTS.md and hooks, and when one makes output worse.
AI is taking shape, moving from models that answer questions to models that can reason. Now models are powering the agentic era, and Dell Technologies Inc. has announced several extensions to its AI Data Platform designed to manage this new paradigm.
Tail Rent: What Actually Happens to Your Agent’s Memory After a Week of Continuous Runtime I had a coding agent running against a real backlog for the better part of four days last month. Not one …
Memory is what an agent noted and a plugin replays. Documentation is what you wrote, reviewed and committed. How each loads, and what to write down.
From Retrieval to Governance — The Architecture Shift Replacing Naive RAG Stop asking “How do we retrieve more chunks?” Start asking “How do we govern and structure what the model …
Claude Code 2.1.285 with 65 MCP tools: tool search cut the first request from 27,184 to 9,722 tokens and a full task by 46%. Off by default behind a proxy.
DeepSeek has released official macOS and Windows desktop apps for DeepSeek Harness v0.2, its MIT-licensed agent harness. The preview adds a plugin manager, file and code-change review, and scheduled Automation Tasks. It also supports non-DeepSeek models through OpenAI-compatible endpoints.
fast-jev-compaction prunes in a second, not 30. At its default threshold it keeps almost no tool results, and a scorer that always says no cuts as much.
Perplexity Research and turbopuffer have released pplx-embed-v2-context-9b-preview, a contextual embedding model for RAG pipelines. Each chunk is embedded with the full document in view. The real change is the training signal.
A Context Language Model is an existing model that rewrites its own context like a file. It manages working context inside a task better than summarisation, with less compute, and throws that context away when the task ends.
Facts every session needs go in AGENTS.md, procedures go in skills, and rules that must hold go in hooks. How to share one handbook across six agents.
A mandated clarifying-question rule fired on 0 of 4 arms. Removing the broader rule it lost to restored it on 4 of 4. Why structure beats instruction, measured.
The enterprise artificial intelligence conversation is moving past models and compute. As companies push projects into production, the harder problem is increasingly what sits underneath them: getting distributed, messy and often sensitive enterprise data ready for systems that need reliable…
Every LLM call is stateless, and a model on its own cannot act; the software that turns it into an agent is the harness.
Showing 18 of 35 stories
Load more storiesGet the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
