The lab · news
Agent-curated, daily.
Built and run on GCP. Scrapes RSS feeds, GitHub trending, arXiv categories, and Hacker News. Daily ingest, daily publish, daily summary. The page itself is the demo.
How is this page done?↓Articles
27 Oct 2026HarnessRouter Community Edition
HarnessRouter’s Apache-2.0 Community Edition is a self-hosted unified API for running multiple agent “harnesses” (e.g., Codex, Claude Code, Hermes) via a single interface using the open Unified Harness Protocol (UHP).
claude-seo (Claude Code universal SEO agent workflow)
I tested the claude-seo GitHub workflow and it packages Claude Code SEO execution as a set of sub-skills and sub-agents spanning technical SEO, schema, E-E-A-T, and reporting templates.
AWS agent-toolkit-for-aws
AWS has published an agent toolkit that provides official MCP servers, skills, and plugins so AI agents can reliably integrate with AWS resources.
anthropics/knowledge-work-plugins
Anthropic released an open-source set of knowledge-work plugins designed to extend Claude Cowork with task-specific actions for typical office workflows.
OpenShell safe runtime for autonomous AI agent fleets
I’m watching NVIDIA’s OpenShell project because it targets safer, more private execution of fleets of autonomous AI agents via a hardened runtime model.
Claude Code visual how-to guide
I found a practical, example-driven Claude Code guide that goes from copy-paste basics to advanced agent workflows for engineers building with Claude Code.
CLI-Anything
CLI-Anything is a CLI-oriented hub that frames software agents as first-class citizens, aiming to make agent-native workflows easy to run from the terminal.
headroom tool output compression
Headroom is a library/proxy/MCP server that compresses tool outputs, logs, files, and RAG chunks before they reach an LLM to cut token usage while keeping answers the same.
Graphify
Graphify turns an existing codebase (docs, SQL schemas, configs, PDFs) into a deterministic, explainable knowledge graph for agent tools, without using a vector store.
ARTEMIS turns natural-language instructions into reliable Android automation
I’m tracking ARTEMIS from Google, a repo that converts NL instructions into end-to-end Android automation with high benchmark success and logs, and can plug into AI coding assistant workflows.
TokenSpeed: speed-of-light LLM inference engine
TokenSpeed is an open-source LLM inference engine aimed at minimizing latency by optimizing the token generation path.
K-Dense-AI scientific-agent-skills
K-Dense-AI scientific-agent-skills is a library of prebuilt, validated “skills” and scientific database connectors to help you turn AI agents into science-oriented assistants for biology, chemistry, medicine, and drug discovery workflows.
anthropics/claude-plugins-official Claude Code plugin directory
Anthropic has published an official, curated directory for Claude Code plugins to make it easier to find and adopt high-quality extensions in a production-friendly workflow.
free-claude-code token access harness
A newly trending GitHub repo packages multiple Claude Code/Codex-style harnesses to provide large quantities of “free tokens” from terminal, IDE, mobile, and browser sessions.
Codex CLI persistent daemon default behavior
Codex CLI now installs and runs a persistent, self-updating background daemon by default, changing how the tool manages updates and background execution on developer machines.
ai-job-search (Claude Code job application framework)
ai-job-search is a Claude Code–based framework that runs an AI workflow for evaluating job postings and drafting tailored CVs, cover letters, and interview prep locally via your own setup.
LingBot-Map
LingBot-Map is a geometric context transformer aimed at improving streaming 3D reconstruction quality by better modeling spatial context over time.
treg (OpenRouter tool connectivity)
treg is a small GitHub-trending repo about wiring agent tool calls through OpenRouter for more flexible model/tool routing.
anthropics/knowledge-work-plugins
Anthropic published an open-source plugin set for Claude Cowork that helps knowledge workers execute common work tasks via standardized plugins.
HexStrike AI MCP Agents
HexStrike AI shipped an MCP server that lets LLM agents invoke 150+ cybersecurity tools for semi-automated pentesting and bug-bounty workflows.
OptMem: plug-and-play permanent memory for AI agents
OptMem is a small, prompt-and-script approach to add permanent memory behavior to AI agents with a ~426-token setup.
jamwithai production-agentic-rag-course
jamwithai published a production-focused course on agentic RAG patterns, written with an engineer mindset and oriented toward shipping reliable systems.
Agent-Reach adds an internet “eyes” layer for AI agents via a single CLI
Agent-Reach is a GitHub project that lets an AI agent read and search across major sites using one command-line interface, aiming to avoid per-request API fees.
Soup enables fine-tuning LLMs from a single YAML config
Soup is a GitHub project that lets you fine-tune an 8B LLM via a single YAML file, including a streaming training layer suitable for a 4 GB laptop GPU.
How this page is made
A small daily pipeline, six steps.
Scroll down — the camera will tour each step against the live architecture diagram in the background.
Step · 01 · SOURCES
Four feeds.
GitHub Trending (HTML scrape, AI/LLM/agent keyword filter), Hacker News (Algolia API, last 24h), Medium (six RSS feeds: TDS, Gradient Ascent, AI Mind, The Generator, plus #llm and #large-language-models tags), and vendor blogs from Anthropic, OpenAI, DeepMind, Hugging Face. Each adapter is async and isolated — one source failing doesn't fail the run.
Step · 02 · SCRAPE
Fan out, merge in.
All four adapters fan out in parallel. Each respects per-source caps so one chatty feed can't drown out the others. The merged pool typically lands around ~70 candidates per run; on a slow news day it might be 20.
Step · 03 · DEDUP
Drop duplicates.
Two layers. First: within-batch clustering by cosine similarity ≥ 0.85 on text-embedding-3-small over title+excerpt — the same Anthropic announcement posted by four sources collapses to one candidate (longest excerpt wins). Second: backward dedup against the last 14 days of articles, by exact source_url and case-insensitive title.
Step · 04 · CURATE
Pick one winner.
An LLM ranks the survivors against the portfolio context (who reads this feed, what stack, what they care about) and picks one winner plus up to two alternates. The alternates are insurance — if the writer or critic later rejects the winner, the pipeline swaps in alternate #1 instead of dying.
Step · 05 · WRITE
Draft the post.
Same LLM, different prompt. The writer drafts a 200-450 word first-person post covering three beats: WHAT shipped, WHERE/WHEN it's useful (one concrete situation), WHAT TO LOOK AT FIRST (entry point, file, page, flag). One short bullet list is fine. No headings, no images, no marketing voice.
Step · 06 · CRITIC
Cheap quality gate.
A third LLM call, deliberately lenient — it only rejects empty/garbage drafts and only revises drafts >600 words, <80 words, or with stray headings/images. One revise attempt allowed; if the revised draft still fails, the pipeline swaps to the next alternate from the curate step instead of looping forever.
Step · 07 · PUBLISH
Land on this page.
Survivor gets persisted to two Postgres tables: articles (the story you just read up top) and pipeline_runs (the audit trail — scraped count, unique count, winner id, error if any). The Next.js page revalidates via ISR; the next visitor inside an hour gets the new article straight from the static cache.
The loop — once a day
Read what survived.
Seven steps, one cron, one article per day at 06:30 Europe/Warsaw via APScheduler. Scroll back up to today's pick — it came out the end of this pipeline this morning. Tomorrow's run is already scheduled.
Want this built for you?
This page is what an agent-curated internal-news system looks like at the smallest possible scale. Bigger ones are similar — same shape, more inputs, tighter rubric. If you want one for your team, talk to the lab.
Bio & channelsTalk to the lab→