The name we changed
We shipped under "ADE" — agentic development environment — long enough that our own X handle still says it. Then we changed the words on the front page to Agent IDE. Here is what the two names each get right.
Two names for the same shift#
Both names are trying to describe the same change: the thing you spend the day in is no longer where you type every line, it is where you direct work that gets typed for you. That is a different tool even when it looks like an editor, because the questions it has to answer are different — not "where is this symbol defined" but "which of these four agents is blocked, on what, and is the diff any good".
- ADE — agentic development environment
- Accurate and load-bearing: it says the whole environment is agentic, not that an editor grew a chat panel. It is also the name of a category that does not exist yet, which is precisely its problem — nobody types it into a search box.
- Agent IDE
- Less precise — "IDE" carries three decades of assumptions about a text editor at the centre — but immediately legible. Someone who has never read our site knows within a second roughly what kind of thing this is.
We now lead with Agent IDE on the front page and keep ADE as the category we would like to see named. If you came here from our X account, that is why it still reads @Mirasim_ADE: the handle is older than the decision, and renaming a handle costs every link that points at it.
What the thing has to do to earn either name#
Names are cheap. The test is which problems the tool takes off your hands once more than one agent is running. Three of them are unavoidable, and they are the ones a chat panel bolted onto an editor does not address.
- Keeping two agents out of each other's way
Two agents on one checkout overwrite each other silently. The environment has to give each one a place of its own — for us, a linked git worktree per session — or parallelism is a demo rather than a working mode.
- Not owning the agent
The agents people already trust are third-party CLIs with their own logins. An environment that only works with its own built-in agent is asking you to give up the one you chose. Ours launches Claude Code and Codex as themselves, under your accounts.
- Telling you whether the work is any good
More output per hour is not the win; more work you can accept is. That is why evaluation sits inside the loop here rather than after it — the evidence comes back as the next batch of tasks.