Ben AI calls this a "Claude 2.0" desktop overhaul, but the video is about the Claude Code workspace, not a new model named Claude 2.0. His seven demonstrations are most useful when separated into interface changes, coordinated work, artifact creation and local customization. A few access and behavior claims also need checking against Anthropic's own documentation.
For a long task, open the Code tab and use the session details pane, a side chat for a question, and split view for a second session. If you have the new Projects beta, try one goal with two well-scoped threads. Docs, Slides and Design are broadly available, but Dashboards and Motion have narrower plan access. Mods can change Claude Code deeply, so install only code you trust.
Watch Ben's 13-Minute Walkthrough
Credit and disclosure: The examples and preferences are Ben AI's. His course and community promotion appears after the projects segment and again at the end. This article uses the supplied transcript for what he demonstrates and Anthropic's current documentation for access, commands and limitations; we did not test features on every subscription plan.
The Seven Feature Groups
1. Session details and sections find the work
What Ben shows
A plan checklist in the conversation and cards for the environment, context, connectors and outputs. He groups conversations into sidebar sections so separate client or content tasks are easier to find.
Use it well
Check whether the session is local or remote and which folder it can reach before asking it to edit files. Context compaction can free room, but a summary may omit detail; keep important decisions in project files, not just chat history.
2. Projects and coordinated threads divide larger goals
What Ben shows
One coordinator breaking a video-production goal into title, outline and visual threads that can be revisited individually. Anthropic says the redesigned beta shares project memory and a library, while each cloud thread works in its own session.
Check first
The new Projects experience is a staged beta for selected Pro and Max Claude Code users, not a feature every account already has. Ben also shows locally nested work on a Team account; do not assume it has the same cloud continuity or shared memory. Parallel threads draw down usage faster, and edits to the same code still need conflict review.
3. Side chats ask without interrupting
What Ben shows
A question about an in-progress task opens alongside the main session, using its context without diverting its plan. This is useful for checking an assumption or asking what has been done so far.
Important correction
Anthropic says a side chat does not add anything back to the main conversation and is not saved to disk after the app closes. Ben describes persistent side chats and messaging back to the main chat, but that is not the behavior guaranteed by the current desktop reference. Put any decision you need the main task to use into the main session explicitly.
4. Split view compare live sessions
What Ben shows
Two chats visible at once, so he can monitor separate tasks without repeatedly switching the sidebar selection. Anthropic documents opening a second session side by side from the sidebar or its Open in menu.
Use it well
Keep the tasks independent or use separate worktrees if both edit a Git repository. Split view improves visibility; it does not merge session context or prevent conflicting edits.
5. Docs, Slides, Design, Dashboards and Motion create artifacts
What Ben shows
Editable documents, decks and designs beside a conversation; dashboards that query connected data; and code-built motion pieces that can export as video. These outputs can be revisited and shared as artifacts.
Availability and limit
Anthropic lists Docs, Slides and Design on all plans. Dashboards are a paid-plan beta; Motion is a Team and Enterprise beta, subject to admin settings. A live dashboard still needs correct source permissions and query checks, and an MP4 export is not evidence that every animation will be faster or cheaper than a video model.
6. Slash commands control the workflow
What Ben shows
An output style for concise answers, a setup and prompt audit, a skill-usage report, and a stronger-model advisor for selected decisions.
Check first
Use the command shown by /help in your installed version. The transcript's shorthand for the doctor command is easy to mishear; Anthropic documents /doctor and /doctor prompt-audit. Advisor consultations can use additional plan allowance or API tokens.
7. Mods customize Claude Code
What Ben shows
His brain-dump mod restructures a long voice prompt before Claude processes it. Anthropic says mods are small TypeScript functions that can change a prompt, add UI or replace built-in behavior in the CLI or desktop app.
Safety check
Mods run with Claude Code's machine access and are not sandboxed. Review source, permissions and plugin origin before installing one, especially on a work computer. Prompt rewriting also deserves an easy way to inspect the final text before a consequential task runs.
Four Commands, Checked Against the Docs
Ben's command segment is useful, but the exact name and behavior can vary by Claude Code version. This is the current reference to use when trying his examples:
| Command | What it does | Before relying on it |
|---|---|---|
/output-style | Choose a built-in response style or a custom one. Anthropic's output-style guide explains the distinction from project instructions. | A style changes how Claude responds; it is not an enforcement control. Put project facts in CLAUDE.md and guaranteed checks in hooks. |
/doctor | Checks installation, extensions, configuration and usage. /doctor prompt-audit focuses on stale or conflicting instructions, per the official skills guide. | Review proposed fixes; Anthropic says it asks before changing anything. The prompt-audit subcommand needs a recent Claude Code version. |
/skill-doctor | Shows skill context cost and usage so you can identify unused additions, according to Anthropic's guide. | It reports candidates to turn off; do not delete an infrequently used skill solely because of one usage window. Availability requires a supported version. |
/advisor | Lets a compatible main model consult a second model at selected decision points; see the advisor guide. | Experimental and Anthropic-API dependent. Each consultation reads the conversation and uses additional tokens or allowance; it does not make the main model's result automatically correct. |
Should You Move Everything from Cowork to Code?
Ben prefers Code because its session controls and extensions fit his work. That is a workflow preference, not a universal migration requirement. Anthropic's desktop reference describes Code as the place for development sessions and file-aware work, while Cowork still serves Dispatch and longer general tasks. If a task needs a repo, tests, diffs or parallel code sessions, Code is the natural trial. If it is mainly a connected research, document or spreadsheet workflow, keep Cowork in the comparison. In either place, inspect connector access and the files Claude can change.
Three Deeper Videos from Ben AI
These are the follow-ups Ben names in the episode. They sit beside the corresponding artifact and mod features, not as three unrelated recommendations.
Video Chapters
| Time | Segment |
|---|---|
| 00:00 | Intro and the redesigned desktop |
| 00:37 | Why Ben prefers Code to Cowork |
| 01:18 | Feature 1: session details and sections |
| 02:51 | Feature 2: projects and nested work |
| 06:43 | Feature 3: side chats |
| 07:47 | Feature 4: split view |
| 08:15 | Feature 5: Docs, Slides, Design, Dashboards and Motion |
| 10:08 | Feature 6: four slash commands |
| 12:25 | Feature 7: mods |
Sources and Limits
Demo evidence: Ben AI's walkthrough and the supplied transcript describe his screens and personal examples. Product evidence: Anthropic's desktop reference, project announcement, artifact access table, skills and doctor guide, advisor guide, and mods announcement. Rollouts and documentation can change; check your own version and plan before following a click path. Sources checked 25 October 2026.