Chat is not going away.
Your work is leaving it.
The claim doing the rounds is that the chat window is finished. It is not, and anyone telling you otherwise is selling something. What is true is narrower and more useful: the work that produces a deliverable has started moving out of chat, and there is a clean test for what should follow it.
Anthropic still ships chat on every plan and shows no sign of removing it. What changed is that Claude Cowork exists alongside it, described by Anthropic in flat terms as a system that executes multi-step work and is not a chat assistant. Two tools, two jobs. The interesting question is which of your tasks sit in the wrong one right now.
The copy-paste tax is the real reason work leaves chat
In a chat window the output arrives as text in a box. You then transplant it: into a document, a spreadsheet, a ticket, a slide. That transplanting is unpaid work, it happens every single time, and it is where formatting breaks and details get dropped. Claude Cowork writes to the file instead. Nothing about the intelligence changed; the last mile did.
Long threads cost more than you think
Anthropic’s own guidance is direct about this: longer conversations that trigger automatic context management consume more of your usage limit, and tools and connectors are token-intensive. A thread you have been feeding all afternoon is re-reading itself on every turn. A task-scoped run starts clean. If you are hitting caps sooner than colleagues doing similar work, thread length is the first thing to look at.
The test: are you asking for a decision or a deliverable?
If what you want back is an opinion, an explanation, a second read on something you have written, or a way of thinking about a problem, that is chat and it always will be. If what you want back is a file that did not exist before, or twenty files that now agree with each other, that is a task. Most people know which they wanted, and put it in the wrong place out of habit.
What genuinely stays in chat
Thinking out loud. Learning something new, where the back and forth is the point. Sanity-checking a decision before you commit to it. Any question where you would not know what finished looks like until you had seen an answer. Handing that to an autonomous task wastes it, because you cannot define done for a conversation you have not had yet.
The habit worth changing this month
When you next catch yourself pasting a Claude response into a document for the third time in a session, stop and write the whole thing down as a task instead: the input folder, the output file, the format, the exception rule. That rewrite takes about five minutes and it usually replaces an hour of shuttling text back and forth. Do it once for a job you repeat weekly and the payback is immediate; do it for a job you repeat once a quarter and it probably is not worth the effort.
Nobody at Anthropic has announced the end of the chat interface, and this piece is not claiming they have. The shift is in where the centre of gravity sits for work with an output. If your day is mostly producing things rather than discussing them, Claude Cowork training will change more about your week than any amount of prompt refinement in a chat window.
Things people
usually ask.
No. Chat is available on every plan, including the free one, and there has been no announcement of it being withdrawn. The change is that work with a file-shaped output now has a better home in Claude Cowork.
That is the normal pattern. Think in chat, produce in Claude Cowork. The two share the same usage allowance, so a long unattended task and a long chat thread compete with each other inside the same window.
Then chat is probably still the right home for most of it, and the useful investment is in briefing and context rather than switching tools. Claude Cowork for non-specialists covers where the line falls for roles that are mostly conversation and judgement.