Friday, October 23, 2026
22 stories · 268 sources49호
release · Hacker News · 2건Korean coverage · GeekNews

OpenAI releases official CLI for its REST API

OpenAI CLI

The OpenAI CLI is the official command-line tool for the OpenAI REST API, with resource-based commands for generating responses, managing files, creating and editing images, and using admin APIs. Its help works without an API key or network access, and it supports task-based command discovery, shell completion, and filling example commands onto the command line without running them.

Developers who call the OpenAI API with curl or hand-written scripts can now run response, file and image operations straight from the terminal with an official tool, which changes whether they need to maintain their own wrappers.

Today's 72–7

  1. product · Hacker News

    OpenShell: A Safe, Private Runtime for Fleets of Autonomous AI Agents

    OpenShell – A safe, private runtime for fleets of autonomous AI agents

    OpenShell was posted to Hacker News. It is described as a safe, private runtime for fleets of autonomous AI agents.

    For teams running many agents at once, it's a candidate for handling execution isolation and data privacy at the runtime layer instead of in each agent's code.

  2. release · OpenAI

    OpenAI introduces Decisions API for fast, low-cost app decisions

    Build more responsive apps with the Decisions API

    OpenAI introduced the Decisions API as a way to build more responsive apps. It is pitched as offering fast, focused, low-cost decisions for any application.

    Developers can evaluate whether lightweight decision steps like routing or classification could move off larger model calls onto this API, which would change latency and per-call cost in their designs.

  3. product · TechCrunch AI · 2건

    1Password lets AI agents log in without seeing your passwords

    1Password launches a way for AI agents to log in on your behalf without seeing your password

    1Password introduced 1Password for Agents, a feature that lets AI agents sign into a customer's accounts without seeing or handling their passwords. Agents don't get free rein over connected accounts, since users still have to approve sign-in requests.

    Developers building agents that must log into services can keep credentials out of the agent and rely on an approval step for each sign-in, which reduces the risk of password leaks.

  4. industry · Simon Willison

    Anthropic donates Agent Skills to the Agentic AI Foundation

    Donating Agent Skills to the Agentic AI Foundation

    Anthropic is donating the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation, saying it wants to keep skills open, community-driven, and vendor-neutral. Anthropic co-founded AAIF with Block and OpenAI last year and has already donated the Model Context Protocol (MCP) to it.

    If you build tooling around Agent Skills, the format now sits under a vendor-neutral foundation alongside MCP, which lowers lock-in risk when you pick it as a standard.

  5. industry · OpenAI · 2건

    Uber rolls out ChatGPT Work to all 30,000+ employees

    Uber brings ChatGPT Work to all employees with OpenAI

    Uber is rolling out ChatGPT Work to more than 30,000 employees to help teams research, analyze data, create presentations, and complete complex workflows. The company previously ran a much smaller pilot and is now taking the enterprise agent company-wide.

    For teams weighing enterprise agents, a pilot-to-company-wide move at this scale is a concrete reference point for tool selection, per-seat cost planning, and rollout scope.

  6. industry · Hacker News

    Google staff resign in backlash over military AI contracts

    Google staff resign in backlash over its AI contracts with military

    Google employees have resigned in protest of the company's AI contracts with the military. The story was shared on Hacker News.

    If you build on a vendor's models or APIs, internal backlash and staff departures over military contracts are a vendor-dependency risk to weigh for product roadmap and support stability.

Releases today2

  1. claude-code v2.1.308릴리스 노트 없음트래커·GitHub
  2. claude-code v2.1.307Claude API 코드를 API 버전 2026-10-21로 옮기는 /claude-api api-version이 추가됐다트래커·GitHub

Rising on GitHubstars gained in a day

  1. cloudflare/security-audit-skillCoding-agent skill for multi-phase security audits with independently verified, machine-readable findings.+2,979
  2. DietrichGebert/ponytailA tool that makes AI agents think like the laziest senior dev, favoring code never written.+1,949
  3. Graphify-Labs/graphifyA skill for coding agents that turns codebases, docs, schemas, and PDFs into a queryable knowledge graph.+1,407
This week's repositories →

Also13

  1. OlmoEarth v1.3: State-of-the-art embeddings for the whole planet
    Hugging Face·01:11
  2. Beyond tree cover: mapping the world’s forest types
    Google DeepMind·10/22 23:30
  3. Ensemble forecasting with IBM Granite time-series models
    Hugging Face·10/22 16:51
  4. Cloudflare links vulnerability discovery to defense with OpenAI
    OpenAI·10/21 09:00
  5. Today's PHP Security Release, and How We're Strengthening PHP Security
    Simon Willison·10/22 03:41
  6. Meta’s AI could still scrape news sites, legal experts say
    The Verge AI·05:32
  7. What FDE means at OpenAI
    OpenAI·10/22 15:00
  8. Broad access through consistent standards
    OpenAI·10/22 06:00
  9. Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO
    Latent Space·02:40
  10. Claude Mods: Flipping the Harness Inside Out
    Latent Space·10/21 22:04
  11. AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
    Latent Space·10/21 14:01
  12. Meta turns Muse sharing back on
    The Verge AI·05:24
  13. You are the Team — Anekha Sokhal, Moshi
    Latent Space·10/22 23:03

Korea2

  1. KT클라우드, 50MW AI 데이터센터 착공...2029년 준공 목표
    THE AI·10/22 22:44
  2. 모빌린트, 휴머노이드용 AI 플랫폼 개발...NPU부터 SW까지 통합
    THE AI·10/22 22:26