Source: raw/Use_Claude_2.0_better_than_99%_of_people_11_BIG_Updates.md; second creator: raw/Anthropic_Just_Dropped_Claude_2.0_Full_Breakdown.md (Brock Mesarich, see below); third creator: raw/Claude_2.0_is_Insane_All_Updates_Explained.md (Ben AI, see below); fourth creator: raw/Claude_2.0_Is_Here._EVERYTHING_in_15_Minutes.md (Futurepedia, see below)
Creator: Jay (Robo Nuggets) URL: https://www.youtube.com/watch?v=yMnFHFf43Qg Duration: 17:06 Platform: YouTube Published: 2026-10-28 (YouTube upload date, UTC)
Jay walks through 11 changes to the Claude desktop app’s Code tab after the October 2026 revamp, which he says one Anthropic lead engineer called “Claude 2.0” (the engineer is not named). The useful part is how the session features fit together: a session rail that shows what each chat is using, sessions that start their own sub-sessions, one pinned “master session” that reads and steers the others, and side chats for questions that would otherwise interrupt a long task. The rest is a tour of artifact types, slash commands, mods and keyboard shortcuts, most of which the wiki already covers and links below.
Single creator source
Everything here is Jay’s demo and description. Plan limits, the 20-session default and the shortcuts were not checked against Anthropic documentation. Where the wiki has a first-party source for the same feature, it is linked.
Key Takeaways
-
The desktop app now shows only Chat and Code. Cowork was folded into Chat “as of last month”, which matches the September 2026 merge in Claude Cowork. Jay’s advice is to work in the Code tab because “that’s where Anthropic is shipping the best and the most updates”, and that non-developers can use it since you direct it in plain English.
-
The session rail is a panel of cards about the current chat. Open it with “Show session rail” at the top right if it is hidden.
- Environment card: which folder the chat works in, local or cloud, and a Remote Control toggle to continue the same session from your phone (see Remote Control).
- Context card: how full the context window is and what fills it: your messages, memory files, skills and MCP connectors. Memory, skills and connectors load on every new chat, so they cost usage before your first prompt. Use the card to spot ones that have grown too large.
- Connectors and outputs cards: the apps plugged into this chat, and every file the chat has made.
-
Two settings to switch on (Settings, Claude Code): “connect new sessions to Remote Control”, so every chat is phone-ready without the per-chat toggle, and “keep computer awake while Claude works”, so a long task does not stop when the machine sleeps.
-
Sidebar sections work like folders. Use the three-dot menu on any chat, then “Move to new section”, name it, and drag other sessions in. Jay suggests one section per project, client or part of the business.
-
Nested sessions: a session can start its own sessions, which sit under it in the sidebar and report back when they finish. To try it, ask Claude to do several tasks at once and to start a nested session for each. His example is a campaign launch with one sub-session each for the sales page, the emails and the ads.
-
Nested sessions versus subagents. A subagent is “more like a one-way order”: the main session sends a prompt and gets a result back, and you can watch its progress but not talk to it. Each nested session is a real chat in the sidebar that you can open, read in full and keep working in directly.
-
A master session reads and steers your other sessions. One Claude Code session can see your other sessions. Pin one chat to the top, tell Claude it is your master session, then ask it, for example, which decisions your other chats are waiting on. It reads them and summarises only the ones waiting on you. It can also start new sessions, control them and sort chats into sidebar sections. By default it sees your 20 most recently active sessions; ask it to look further back if you need to.
-
Nested sessions or Claude Code projects. Projects (see Claude Projects) are the same main-session-runs-a-team idea, but:
Nested sessions Claude Code projects Runs where Your computer; stops when it sleeps Anthropic’s cloud; keeps going when you close the laptop Your skills, files, tools Available directly Must be added to the project first; Jay asks Claude to gather them from the computer while it is awake Data location Local Uploaded business files go to Anthropic’s servers Shared rules Per session Shared project memory: set a rule once and every session in the project follows it Overview Sidebar Overview page of what is done, in progress and waiting on you Plans Any plan Beta, Pro and Max; look under “More” in the sidebar or join the waitlist His rule of thumb: if the work needs your computer or you want it local, use nested sessions; for a large, ongoing job that should run while you are away, use projects.
-
Side chats answer questions without interrupting the main task. Open one from the three-dot menu at the top right or with
/btw. It can read the whole main session, so ask it how far along the build is or what the session has done so far. Asking in the main chat instead interrupts the work and spends more tokens. Side chats used to be temporary; now each one is saved and you can open as many as you want. Futurepedia says the opposite; see the contradiction callout in his section below. -
Split view puts several chats on screen. Alt/Option-click a chat in the sidebar, or right-click and choose “Open in split view”. Zooming out (Ctrl/Cmd and minus) fit up to 10 chats for him, but he suggests starting with two.
-
Artifacts are shared canvases with five types: Docs, Slides, Design, Dashboards and Motion. You and Claude work on the same canvas, leave comments for Claude to fix, and publish a live link or share with a team or client. Slides download as PowerPoint or present from Claude. Design settings let you upload your own design systems. The standalone
claude.ai/designsite closes on 14 December. All of this matches Claude Dashboards and Motion, including that Pro and Max do not get Motion. -
AI-powered artifacts put Claude inside the artifact. Turn it on in Settings, Capabilities, “AI-powered artifacts”. Jay’s use: add a chat box or search bar to a shared team dashboard, so teammates who don’t use Claude Code can open the link and ask questions about its data.
-
Slash commands he recommends (details in the CLI reference):
/output-stylefor shorter or plainer answers;/doctorto check settings and CLAUDE.md and offer fixes;/skill-doctorfor each skill’s context cost and how often you use it;/advisorto run a cheaper model and call a stronger one only for big decisions (see the advisor tool);/insightsfor a usage report;/recapfor a one-line summary of where a chat is;/rewindto go back to an earlier point. -
Mods change how the app looks or works, and you can make one by describing it (see Claude Mods). He uses two daily: Ribbon, his own free, open-source bar above the chat box that shows remaining weekly usage and whether the prompt cache is warm or cold, and Savvy Progress, a progress bar plus a live panel of each subagent. His reason for watching the cache: after an hour it goes cold, and the next message re-reads the whole conversation at full cost.
-
Keyboard shortcuts he uses (as shown on his machine; not checked on other platforms):
- Ctrl+Alt+Space opens Claude from anywhere, in a new chat.
- Esc twice triggers
/rewind. - Ctrl+D starts voice dictation.
- Ctrl+/ (Cmd+/ on Mac) shows the full shortcut list.
-
Skip: the Robo Nuggets community pitch and the free PDF guide offer.
A second “Claude 2.0” video: Brock Mesarich, September 2026
Source: raw/Anthropic_Just_Dropped_Claude_2.0_Full_Breakdown.md · Creator: Brock Mesarich · URL: https://youtu.be/lrsW8k-dHSk · Duration: 8:48 · Platform: YouTube · Uploaded: 2026-09-17. Added 2026-10-27.
- The nickname is older than Jay’s video, and it first described the September merge. Brock titled his video “Claude 2.0” on 2026-09-17, the day after Anthropic’s “One Claude” post (see Claude Cowork). He used it for the Chat and Cowork merge plus the new Docs, Slides and in-chat Design outputs, not for the Code-tab features above. He does not attribute the name to anyone at Anthropic. “Claude 2.0” is a creator label for the autumn 2026 app changes, not a release name.
- On launch day the merge had not reached his account. He still saw the chat/Cowork toggle and expected it to go “over the next couple of weeks”. By Jay’s late-October video it was gone, which matches the staged rollout recorded in the Cowork article.
- The Cowork composer, as he showed it on the desktop app: a project selector; approvals set to manual, automatic, or skip all approvals; and an Output picker for Docs, Slides or Design.
- Docs. From a demo client CSV and “turn this into a document that I can edit”, Claude built the doc live in a side pane: title, date and an @-mention of him as author, a summary, and a table with an editable status column (he changed one to “past due”). A list converts to a checklist with one click, and a doc can have several tabs. He compares it to Notion mixed with Excel.
- Slides with your own design system. Choose Slides, pick a design style made earlier in Claude Design (his is called “Bramble”), and attach the data. His verdict on the result: “solid” but it “definitely looks like AI generated this”. You can then change text colour, edit or remove text, and drag in images, “as if we were using something like Google Docs”.
- Present and share. Present mode shows speaker notes. Share copies a link to the artifact, or Export sends it to Canva, Gamma, Miro or other AI tools, or downloads it as an HTML page, a PDF or a PowerPoint file. This agrees with Jay’s “Slides download as PowerPoint or present from Claude” and adds the other targets.
- A connector gap, closed with Zapier MCP. Drip, his email platform, is not in the Claude connector directory. In Zapier MCP he chose Add MCP server, added Drip and selected all its tools, then pasted Zapier’s connect prompt into Claude and clicked “always allow”. His server now reaches about 40 apps. He then picked Output → Docs and asked Claude to analyse his Drip campaigns through the Zapier MCP. A few minutes later he had an editable doc with subscriber and unsubscribe counts, each campaign, and what is and isn’t working. See Claude Connectors and Zapier.
A third “Claude 2.0” video: Ben AI, October 2026
Source: raw/Claude_2.0_is_Insane_All_Updates_Explained.md · Creator: Ben AI · URL: https://www.youtube.com/watch?v=J5EbL1UaqG8 · Duration: 13:40 · Platform: YouTube · Uploaded: 2026-10-25. Added 2026-10-28.
Ben covers the same Code-tab features as Jay, three days before Jay’s video, and they agree on how they work. Like Brock, he credits the “Claude 2.0” name to no one. He had used Cowork for its layout until now and argues non-coders should switch to the Code tab: it is “essentially the same tool as Cowork in the back,” but can work with any file on your computer, use CLIs and get new features first. What he adds:
- Which plan sees which. Pro and Max get Claude Code projects; on a Team plan you see nested sessions instead. As of his recording, project threads share one project memory and can run in the cloud with the laptop closed, while nested sessions get a brief from the main chat and run only locally. He expects projects on all plans “in a couple of weeks.” Jay’s table above says nested sessions work on any plan; the two accounts agree on the differences.
- Getting threads without projects. Open a blank chat and ask it to work on a bigger project: Claude offers to spin up threads, or you tell it to. Ask something unrelated mid-chat and it offers a new thread. Threads appear in a drop-down under the main chat, where you can open each one and keep working in it.
- Why the extra layer helps. His YouTube prep used to be three disconnected chats (titles and thumbnails, outline, visuals and editing) that knew nothing of each other. Now one coordinator runs a thread for each, knows where every thread stands, and each thread can still start its own subagents.
- Side chats can message the main chat. Beyond reading it, a side chat can send messages to the main session. He uses one to review the main chat’s output without adding to its context.
- Split view by drag. Drag a chat from the sidebar into the current chat, or use the chat’s three-dot menu, Open in, Split view.
- Dashboards stay current. A dashboard artifact pulls fresh data from your connectors each time it is opened. See Claude Dashboards and Motion.
- Slash-command results on his setup:
/doctorread 404 files and flagged 312 problems, mostly instructions not needed for the 5.5 models;/skill-doctorfreed about 20,000 tokens of unused skills;/advisorlets his Sonnet inbox-cleaning chat call Opus 5.5 only when drafting an important reply. He also uses/output-stylewith a style that keeps answers short. - A brain-dump mod. His own mod detects a long voice brain dump and rewrites it into a structured prompt before Claude acts on it. See Claude Mods.
- Skip: the free PDF guide and his AI Accelerator pitch.
A fourth “Claude 2.0” video: Futurepedia, November 2026
Source: raw/Claude_2.0_Is_Here._EVERYTHING_in_15_Minutes.md · Creator: Futurepedia · URL: https://www.youtube.com/watch?v=tM0dwXhOv2o · Duration: 14:45 · Platform: YouTube · Uploaded: 2026-11-03. Added 2026-11-09. Like Jay, he attributes the “Claude 2.0” name to an unnamed engineer who worked on Claude Code.
He covers the same Code-tab features and agrees with the accounts above on the two tabs (Home, which holds Chat and Cowork, and Code), sections, split view, Remote Control, “keep computer awake”, nested sessions, the master session and its 20-session default, and Claude Code projects. What he adds:
Topic: are side chats saved?
Existing claim: (from claude-desktop-app-revamp-robo-nuggets.md, Jay, video published 2026-10-28) — “Side chats used to be temporary; now each one is saved and you can open as many as you want.” New source says: (from raw/Claude_2.0_Is_Here._EVERYTHING_in_15_Minutes.md, Futurepedia, uploaded 2026-11-03) — “These side chats aren’t saved permanently, so if you do quit, it won’t be there when you come back.” Status: unresolved
-
Side chat versus fork. A side chat (three-dot menu, or a Cmd/Ctrl shortcut the captions render as “command or control plus”) reads the whole main chat and answers without adding the reply to it. A fork copies the chat from a chosen point into a new chat, so you can try another direction without disturbing the original.
-
The context window is 1 million tokens, and a context ring next to the model picker shows the same usage as the rail’s context card at a glance.
-
More rail panels: a terminal, a changes panel with the history of what Claude changed, a browser for searching and viewing sites, and a files panel listing every file the chat built or used, for people used to an IDE such as Cursor or VS Code.
-
Split view: right-click a chat, Open in, Split view. He opened “a ton” without hitting a limit but usually keeps a few.
-
Cloud sessions. When starting a session you can pick the cloud instead of your computer, so it keeps running with the laptop closed. A cloud session cannot see files on your computer, but it can work from a GitHub repo. See Claude Code in the Cloud.
-
/goalfor long runs. Give Claude a finish line and it keeps working until it can verify it is done. Write a finish line Claude can actually check, not “make it great.” With a goal set and Remote Control on, he walks away and checks in from his phone. See goal walkthrough. -
Task chips. Claude notices related work and shows a card; clicking Start opens it as a nested session under the current chat.
-
The master session sees only local sessions, not ones running in the cloud.
-
Projects, as he describes them: threads run in the cloud by default (you can ask for one to run locally); each thread starts with the project’s instructions in memory, where a nested session only gets what the main chat tells it; and the project has an overview of what is waiting on you plus a library of every file. Beta at his recording.
-
His summary of the three: nested sessions are for side jobs, a master session is for managing what is already running, and projects are for ongoing work that shares one memory and can keep going in the cloud.
-
Mods he uses:
- A prompt-cache keep-warm mod he credits to Nate Herk, which sends a tiny message just before the one-hour cache expires.
- Blast Radius, which shows what a destructive command would delete and offers a cancel button.
- Recording mode, which hides API keys and email addresses on screen.
- Jared Watts’s mod that connects you to a multiplayer Doom server while Claude works.
All four match mods in Claude Mods, where other creators credit the cache mod differently.
-
Artifacts: Dashboard and Motion are still in beta, and his plan, which he calls “Ultra”, did not show Motion yet. See Claude Dashboards and Motion.
-
Skip: his free setup kit and mod starter pack.
Try It
- Set up a master session. Pin a new chat to the top of the sidebar and send: “You are my master session. Look through my other active sessions and list only the decisions they are waiting on me for, one line each, with the session name.” Then: “Group my chats into sidebar sections by client.”
- Fan out a campaign with nested sessions. “Start three nested sessions: one for the sales page, one for the launch emails, one for the ads. Each should report back here when done.” Open the ads session directly when you want to steer it.
- Audit what loads before your first prompt. Open the session rail’s context card on a fresh chat. If memory files, skills or connectors take a large share, run
/skill-doctorand switch off the skills you never use. - Use
/btwduring long tasks. Ask a side chat “how far along is the main session, and what is left?” instead of typing into the running chat. - Make a shared dashboard answerable. With AI-powered artifacts on, ask Claude to add a question box to a team dashboard artifact, then share the link with someone who doesn’t use Claude Code.
Open Questions
- Who called the release “Claude 2.0”? Jay says “one of their lead engineers” but gives no name or link. Partly answered: Brock Mesarich used the name on 2026-09-17 for the September merge and also credits no one, so it was in circulation among creators a month before Jay’s video. An Anthropic source is still missing.
- Is the 20-session default for a master session documented by Anthropic, and is there a setting to change it?
- Which plan gets which artifact type? He showed a table on screen that the transcript does not capture; only the Motion exclusion for Pro and Max is spoken.
- Do nested sessions count against usage limits the same way as separate chats? Not covered.
- Are side chats kept after you quit the app? Jay (late October) says yes, Futurepedia (early November) says no; see the callout in his section. A first-party source, or a quick test of closing and reopening the app, would settle it.
Related
- Claude Cowork (Product Overview) — the September 2026 Chat and Cowork merge.
- Claude Projects — the cloud project redesign that nested sessions are compared with.
- Remote Control — continuing a desktop session from your phone, and the always-on setting.
- Subagents — the one-way helpers that nested sessions differ from.
- Claude Dashboards and Motion — first-party detail on the artifact types and plan access.
- Claude.ai Artifacts — artifact basics and sharing.
- Claude Mods — Savvy Progress and Jay’s usage-bar mod.
- Claude Code CLI Reference — the slash commands above.