AI Agents Articles
14 articles across AI Catchup's news, guides, tutorials, and comparisons.
All AI Agents articles
Prompting Claude Fable 5: What to Change and What to Delete
Prompt Claude Fable 5 with less, not more: brief instructions now beat enumerated rule lists, and old skills written for prior models can degrade output. Use effort as your main cost control, expect longer turns, ground progress claims in tool results, and configure fallback to Opus 4.8 for safeguard refusals.
Prompting Claude Opus 5: Trim the Verbosity, Delete the Verification
Claude Opus 5 needs opposite prompting from its predecessors: you prompt for conciseness because effort no longer controls visible length, and you delete verification and double-check instructions because the model already does both. Constrain scope on narrow tasks, cap subagent spawning for cost, and keep thinking enabled at low effort rather than disabling it.
Prompting GPT-5.6: Message Roles, Effort, and Agentic Prompts That Work
GPT-5.6 rewards precise, explicit prompts: structure developer messages as identity, instructions, examples, then context, keep stable content first for prompt caching, and pick reasoning effort deliberately (xhigh for complex multi-step work). Move saved prompt objects into code before OpenAI shuts down v1/prompts on November 30, 2026.
OpenAI Codex Adds Record & Replay: Turn a Demonstrated Mac Workflow Into a Reusable Skill
OpenAI shipped Record & Replay in Codex app 26.616, a macOS feature that turns a workflow you demonstrate into a reusable skill. It builds on Computer Use, which you or your administrator must enable, and is unavailable at launch in the European Economic Area, the United Kingdom, and Switzerland.
Cursor Origin Enters Early Beta: Code Hosting, Pull Requests, and GitHub Sync
Cursor's Origin code-hosting platform is now rolling out in early beta to paid plans, with hosted repositories, pull requests, code browsing, GitHub sync, and agents in every repo.
OpenAI Codex adds Locked computer use on Mac (keep Computer Use running after lock)
Codex can now keep using Mac apps after your screen locks. Locked computer use is a narrow, Codex-only unlock path with a short-lived authorization window and safeguards like relocking on local input.
Codex Chrome Extension: How Codex Drives a Signed-In Browser for LinkedIn, Salesforce, Gmail, and Internal Tools
OpenAI's Codex Chrome extension lets the agent use Chrome for browser tasks that need signed-in state -- LinkedIn, Salesforce, Gmail, internal tools. Available in the Codex app in all regions except EU and UK at launch. Setup is Codex > Plugins > add Chrome > install extension > approve permissions. Invoke with @Chrome. By default Codex asks before each new website; allowlist/blocklist and elevated-risk options live in Computer Use settings.
Codex CLI 0.128.0 Lands Persisted `/goal` Workflows: Ralph-Style Agents That Don't Stop Until Done
OpenAI shipped Codex CLI 0.128.0 in April 2026 with a new `/goal` system that keeps a goal alive across turns and runs the agent until it is achieved. The release adds app-server APIs and model tools for goals, runtime continuation, and TUI controls to create, pause, resume, and clear goals. Felipe Coury credits co-worker Eric Traut (the Pyright lead) for the design, and frames it as Codex's take on the Ralph loop pattern.
Perplexity Personal Computer: Complete Guide to the Always-On Mac AI Agent
Perplexity Personal Computer is a Mac app launched April 16, 2026 that lets a Perplexity agent operate your local files, native apps, and browser through a single Cmd+Cmd activation. It runs best on a dedicated Mac mini for 24/7 use, requires the $200/month Perplexity Max plan, and works through a sandboxed action layer with auditable, reversible actions. This guide covers setup, the workflows where it actually saves time, and what to do when it goes sideways.
OpenAI Codex Goes 'For Almost Everything': Mac Computer Use, Browser Comment Mode, and Thread Automations Explained
OpenAI shipped a major Codex update on April 16, 2026 that pushes the product past coding into general work. Three changes matter: Codex can now drive your Mac apps directly, an in-app browser captures both screenshots and DOM elements through 'comment mode', and Codex threads can run continuously to watch Slack, email, and PRs. Here is how each works, the workflows that justify each one, and where Codex now sits relative to Perplexity Personal Computer and Claude Code Routines.
Perplexity Personal Computer vs OpenAI Codex Computer Use vs Claude Computer Use: Which AI Should Run Your Mac?
Perplexity Personal Computer is the right pick if you want an always-on Mac mini agent and you already pay for Perplexity Max. OpenAI Codex Computer Use is the right pick if you live in ChatGPT and need one agent driving your code, your browser, and your desktop apps on either macOS or Windows. Claude Computer Use is the right pick if you want raw computer-use through your own API or want to run it inside Claude Code with the strongest sandboxing model.
Claude Code Routines: Schedule, API, and GitHub-Trigger Your AI Agents
Claude Code Routines is Anthropic's new way to run saved Claude Code configurations automatically -- by schedule, API call, or GitHub event. Routines run on Anthropic's cloud infrastructure with a prompt, repo, and MCP connectors. Available in research preview on Pro, Max, Team, and Enterprise plans.
Scheduled AI Coding Agents in 2026: Claude Code Routines vs Cursor Automations vs Codex vs Warp vs Gemini CLI
Five major AI coding platforms schedule and trigger agents differently. Claude Code Routines and Cursor Automations both ship schedules, API calls, and native source-control events; Cursor's trigger set is by far the widest. OpenAI's product is now ChatGPT scheduled tasks and is still schedule-only. Warp orchestrates other vendors' agents. Gemini CLI ships no scheduler.
AI Tools Landscape: What Changed in Early 2026
Three shifts defined AI tooling in early 2026: MCP settled as the cross-tool standard, coding assistants grew past autocomplete into multi-file workflow partners, and narrowly autonomous agents reached production. MCP's tipping point was earlier than this page first said -- Visual Studio Code made MCP support generally available in June 2025.