Claude Hub

Every Claude workflow I've tested and published: Code workflows, Cowork setups, Skills tutorials, CLAUDE.md files. Real prompts, real failures, nothing theoretical.

Jul 13, 2026

^{Last updated: July 2026. Actively maintained.}

TL;DR: Claude has grown into a small ecosystem, and keeping the whole thing straight is now a job in itself. I built Claude Hub to gather every Product with Attitude guide I’ve written on Claude, including how I choose between Chat, Cowork, Design and Code; how I use Skills, connectors, memory and CLAUDE.md; and what Anthropic’s 2026 releases changed for builders. I’ve also given every release two reads: my PM read and my critical AI literacy read.

I started writing about Claude in September 2025.

Back then, Cowork didn’t exist and Skills weren’t a thing. Claude Code was a niche CLI tool that maybe a few thousands developers used daily.

Many posts and tens of thousands of words later, this page lists all my articles about Claude published Product with Attitude.

I recommend you start with official Anthropic documentation to learn what each feature does and what are the official tips. Once you’ll start using Claude, you’ll find that this is not enough.

Official Anthropic documentation gives you the features and recommended practices. Good. Start there. After a while, though, you’ll likely discover that real usage gives you edge cases and questions the docs cannot cover. That’s when practitioner guides like this one become very handy.

What’s Inside

What is Claude Hub?

I started Claude Hub after one of my readers mentioned that he was having rather a difficult time finding older articles he wished to revisit. It exists because Claude has become too big to navigate from memory. Which is annoying. Also powerful. Today, Claude Hub is Product with Attitude’s independent library for all things Claude.

The library brings together everything I learn and document while conducting my own experiments, from Claude Cowork to Claude Design and Claude Code.

Claude Guide 2026: The Fast Decision Map

In 2026, treating Claude as a single app is magnificently wrong. That said, I quite understand why finding one’s way around its many interfaces can be confusing. Chat, Cowork, Design, Code, Skills, Projects, Connectors, and CLAUDE.md all solve different problems.

Claude Cowork vs Chat vs Design vs Code: Which One Do You Need?

Claude is Anthropic’s AI assistant. Four working surfaces: Chat, Cowork, Design and Code. The right choice depends entirely on what you need Claude to accomplish.

Side-by-side screenshots of Claude's four surfaces, each running Opus 4.8: Chat with a "How can I help you today?" prompt and Write, Strategize, and Career chat shortcuts; Cowork with a "Type / for skills" project workspace; Code with a terminal-style "Describe a task or ask a question" box scoped to a local repo and branch; and Design with a new-project setup using the Product with Attitude Design System. A green footer labeled "The layer beneath all four" highlights Skills and CLAUDE.md as the shared layer that turns a one-time prompt into a permanent default across every mode.

I use Chat for a quick lookup. Cowork comes in when the work spans files and tools or needs a workflow. Design when the result works but doesn’t exactly spark joy. And Code for building, or if my work lives in a repo. Skills and CLAUDE.md make behavior repeatable, saving you from becoming the world’s most exhausted onboarding manager.

Claude surface comparison table from Anthropic breaking down four products: Chat handles questions, conversations, and artifacts with no file access or sub-agents; Cowork runs autonomous knowledge work by planning tasks, spawning parallel sub-agents, and reading and writing real files in a sandbox; Design turns prompts into editable UI with layouts, components, and a machine-readable spec for handoff to code; and Code is Anthropic's CLI tool with direct filesystem access that runs in the terminal, reads your codebase, and executes multi-step development tasks.

Section divider for Claude guides library on Product with Attitude.

Claude Releases 2026

I’m a PM. I track product releases recreationally. And Claude changed materially in 2026. I find product unbundling to be the most significant shift. Yes, Fable 5 and Mythos moved the goalposts and rearranged our work yet again. More significant, however, was the broadening of Claude’s surfaces, which greatly expanded whom the tool could serve.

Claude’s 2026 Release Timeline

Claude 2026 release timeline from January through July 14, summarizing major model, Cowork, Office, agent, and enterprise changes and why each matters to AI builders.

January 2026

In January, Anthropic launched Cowork as a research preview for Max users on Claude Desktop. The goal: take Claude Code-style agency and apply it to knowledge work beyond coding. Claude’s biggest non-technical adoption blocker, the black terminal window, disappeared.

_{My PM read:}_{Cowork moved Claude for non-coders beyond chat. This release gave everyone the ability to execute workflows using files and MCP integrations. From conversations about work, to execution.}

_{My critical AI literacy read:}_{This is also where risk increases. Once Claude can work around files and MCP integrations non-coders need to understand access, permissions, logging, approvals, and failure paths.}

February 2026

February proved rather busy. We gained the ability schedule tasks in Cowork. Skills, plugins and connectors were gathered into a new Customize section in Claude Desktop. Anthropic also launched a Cowork plugin marketplace, Claude for PowerPoint, improved Claude for Excel and released Sonnet 4.6 with advances across coding, computer use, long-context reasoning, agent planning, knowledge work and design. Sonnet 4.6 also received a beta one-million-token context window.

_{My PM read:}_{This release cluster was all about repeatable workflow infrastructure. Scheduled tasks remove some of the manual prompting. Customize improved how we discover and manage capabilities. PowerPoint and Excel moved it into the boring-and-expensive work where companies bleed time. The product strategy is now fairly obvious: Anthropic is shaping Claude into a}_{managed work}_layer_.

_{My critical AI literacy read:}_{Every connector, plugin, admin setting and scheduled task expands Claude’s access. Value goes up. So does the need to understand permission design, logging, failure handling and accountability.}

March 2026

What I appreciated the most in March was Cowork threads persisting and following me from mobile. Memory reached free accounts. Charts moved inline. Excel and PowerPoint got better at carrying context between them. Computer use entered research preview.

_{My PM read:}_{these updates belong in one bucket. Claude’s workspace survives the jump between chats and devices.}

_{My critical AI literacy read:}_{Better continuity = the need for better memory hygiene.}

April 2026

On April 9, 2026 Claude Cowork became generally available on macOS and Windows through Claude Desktop.

Seven days later, Anthropic launched Opus 4.7 , better at long-running coding and vision tasks, and triggered SaaSpocalypse among already-jumpy software investors.

The next day it launched Claude Design, scared Figma, and made Lovable offer users double credits.

_{My PM read:}_{This was a brilliant month for Anthropic. It was no longer adding isolated features. It was swallowing more of the workflow. Each new surface removed another reason to leave Claude.}

_{My critical AI literacy read:}_{A model generating designs doesn’t guarantee good design, nor does a market panic prove that existing tools have become obsolete. Judge the designs by how well they solves the user’s problem. Always.}

May 2026

Opus 4.8 arrived. Better agents, better governance, better work. More honesty about uncertainty. Crazy-good benchmark scores, at the time. And it wasn’t even the main event.

Anthropic was already pointing towards Claude Mythos, making Opus 4.8 both a bridge model and a rehearsal space for higher-autonomy work: effort control, parallel agents, self-verification, uncertainty signalling, and human review loops.

_{My PM read:}_{Capability + control = enterprise AI game.}

_{My critical AI literacy read:}_{We got to practice the workflows and build the habits we need before the next class of models landed.}

June 2026

June probably takes the prize for sheer chaos. Anthropic released Fable 5 for general use and restricted Mythos 5 to approved partners. 72 hours later, both models were suspended following a US government directive.

_{My PM read: The way these models were launched may be a preview of how frontier AI gets packaged from here: one model family, but different access levels. If that pattern holds, tiers will be defined by trust, not only price.}

_{My critical AI literacy read: The best model can disappear three days after launch. We suspected this, now we know for sure. Which means, that model dependency is now product risk for builders. Plan for this.}

July 2026

July is still underway, but the direction seems to be clear. Fable 5 and Mythos 5 are back. Cowork moved to web and mobile, finally, with remote sessions that continue after we close our laptops. Anthropic also launched a reflection dashboard for examining how you use Claude.

_{My PM read: The Claude workspace has slipped another bolt holding it to the desktop.}

_{My critical AI literacy read: Know what Claude can change when you are not looking.}