blooop · blooop · Sep 30, 2026 · Sep 28, 2026 · Sep 28, 2026 · Sep 30, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,6 +7,34 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.59.0] - 2026-09-30
+
+### Added
+
+- **`aid --model` and `aid --effort`** (#662). Each agent is told them in its own spelling:
+  `claude --model/--effort`, `codex --model` and `-c model_reasoning_effort=`, and
+  `gemini --model`. The values are passed on as typed, since the agents add models
+  faster than `aid` is released. `--effort` beside `--gemini` is refused, as gemini has
+  no such setting. Both work with `aid resume` too.
+- **`aid` asks for the agent, the model and the effort before the prompt** (#662). Each is a
+  picker that lists the recent choices first, so one Enter repeats the last launch. The
+  agent picker has one row per Claude login that `dl --claude-profiles` lists, then
+  `codex` and `gemini`. A name that is not listed can be typed. A flag on the line
+  skips its picker, and Esc stops the background boot.
+- **A bare `aid` picks a workspace** (#662), as a bare `dl` does. So does a line of flags with
+  no workspace, such as `aid --codex`.
+
+### Fixed
+
+- **A pasted prompt arrives whole** (#662). The prompt editor read the terminal in line mode,
+  which cut a paste off at 4096 bytes, submitted at its first line break, and left
+  lines that arrived late for the agent to read as keystrokes. It is now a raw-mode
+  editor with bracketed paste: a paste keeps its line breaks, Alt-Enter or Ctrl-J adds
+  a line, and Enter submits.
+- **The agent, model and effort pickers no longer panic** (#662) on a terminal with a size of
+  zero, and take the workspace picker's `TERM` fallback. With no terminal they can draw
+  on, they are skipped, and the prompt editor still opens.
+
 ## [0.58.0] - 2026-09-30
 
 ### Added

diff --git a/README.md b/README.md
@@ -19,7 +19,7 @@ one argument instead of a clone, a config file and a build command.
 [![GitHub pull-requests merged](https://badgen.net/github/merged-prs/blooop/devlaunch)](https://github.com/blooop/devlaunch/pulls?q=is%3Amerged)
 [![GitHub release](https://img.shields.io/github/release/blooop/devlaunch.svg)](https://GitHub.com/blooop/devlaunch/releases/)
 [![PyPI](https://img.shields.io/pypi/v/devlaunch)](https://pypi.org/project/devlaunch/)
-[![Conda](https://img.shields.io/badge/conda-v0.58.0-brightgreen?logo=anaconda)](https://prefix.dev/channels/blooop/packages/devlaunch)
+[![Conda](https://img.shields.io/badge/conda-v0.59.0-brightgreen?logo=anaconda)](https://prefix.dev/channels/blooop/packages/devlaunch)
 [![License](https://img.shields.io/github/license/blooop/devlaunch)](https://opensource.org/license/mit/)
 [![Platform](https://img.shields.io/badge/platform-linux--64-blue)](https://github.com/blooop/devlaunch/releases)
 [![Pixi Badge](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/prefix-dev/pixi/main/assets/badge/v0.json)](https://pixi.sh)
@@ -341,7 +341,7 @@ clone, and [docs/cleanup.md](docs/cleanup.md) says what it carries one past and
 
 ```bash
 $ dl --version
-dl 0.58.0
+dl 0.59.0
 ```
 
 `--devcontainer <variant|path>` picks a non-default `devcontainer.json`. A bare name means
@@ -407,15 +407,23 @@ once for the whole run:
 aid https://github.com/blooop/devlaunch/pull/579 address the review comments
 ```
 
-With no prompt on the line, `aid` starts the container booting and asks for the prompt while it
-does. Type it free of shell quoting, with no escaping and no history expansion eating a `!`. An
-empty Enter starts the agent's plain session. Piping stdin or setting `DEVLAUNCH_NO_TTY=1` skips
-the question and launches one-shot, so scripts behave as they always have.
+With no workspace on the line, `aid` on a terminal lets you pick one of your workspaces, as `dl`
+does. With no prompt on the line, `aid` starts the container booting and asks for the prompt while it
+does. Type it free of shell quoting, with no escaping and no history expansion eating a `!`. A
+paste keeps its line breaks and can be any length, Alt-Enter or Ctrl-J adds a line, and an empty
+Enter starts the agent's plain session. Before the prompt it asks for the agent (one row per Claude
+login, then `codex` and `gemini`), the model and the effort, each in a picker that lists your
+recent choices first, so one Enter repeats the last launch. A flag on the line skips its picker,
+and Esc stops the boot. See [docs/cli.md](docs/cli.md#the-pickers-ahead-of-the-prompt). Piping
+stdin or setting `DEVLAUNCH_NO_TTY=1` skips the question and launches one-shot, so scripts behave
+as they always have.
 
 | Option | What it does |
 |---|---|
 | `--claude`, `--codex`, `--gemini` | Pick the agent. Default `claude` |
 | `--rm` | Delete the workspace when the agent is done. Appendable to a recalled line |
+| `--model <model>` | The model the agent runs, in that agent's own spelling. Passed on as typed and not checked. See [docs/cli.md](docs/cli.md#model-and-effort-in-each-agents-spelling) |
+| `--effort <level>` | How hard the agent thinks: claude's `--effort`, codex's `model_reasoning_effort`. gemini has none, so beside `--gemini` it stops |
 | `--no-remote-control`, `--no-remote` | Start a plain local session. Remote Control is on by default for `claude`: the session is named after the workspace and can be read and steered from claude.ai/code or the Claude app. It needs a claude.ai login in the container |
 | `--remote-control`, `--remote` | Ask for Remote Control by name. `claude` has it already; beside `--codex` or `--gemini` this says they have not got it and stops |
 | `--devcontainer <variant\|path>` | Passed through to `dl` |

diff --git a/docs/cli.md b/docs/cli.md
@@ -730,6 +730,113 @@ claude all take it offline immediately. The entry can sit in the claude.ai list
 roughly 4 hours after that before it clears, which is the web side timing out rather
 than anything still running on your machine.
 
+## Model and effort, in each agent's spelling
+
+`--model` and `--effort` are `aid`'s own words. Each agent is told them in the
+spelling its own CLI takes:
+
+```bash
+aid --model opus --effort max blooop/devlaunch      # claude --model opus --effort max
+aid --codex --model gpt-5.5 --effort high blooop/devlaunch
+                                                   # codex --model gpt-5.5 -c model_reasoning_effort=high
+aid --gemini --model gemini-3-pro blooop/devlaunch  # gemini --model gemini-3-pro
+```
+
+Both take their value as the next word or joined with `=`, and both go ahead of the
+workspace like the agent flags do. After the workspace they are prompt text.
+
+**`aid` does not check the values.** Each agent adds models and effort levels on its
+own release schedule, and codex already takes effort values it does not know by
+name. A list in `aid` would be wrong a few weeks after each release. So `aid` holds
+only the spelling of each flag, in its agent table, and passes the value on as you
+typed it. A value the agent does not know is the agent's to answer for, in its own words: `claude` warns about an unknown effort and uses its default.
+Leave a flag off and no flag is passed, so the agent starts on its own default.
+
+**gemini has no effort setting.** `--effort` beside `--gemini` is refused before
+anything boots, the same way `--remote-control` is refused beside an agent without
+Remote Control. A value that is missing, empty or starts with `-` is refused too:
+`aid --model --codex <ws>` is a typo, not a model called `--codex`.
+
+## The pickers ahead of the prompt
+
+`aid <workspace>` with no prompt on a terminal asks for up to three settings before
+it opens the prompt editor. The workspace boots in the background the whole time.
+
+1. **The agent**, with one row per Claude login and one row for each other agent:
+
+   ```
+   AGENT   NAME      STATE          ACCOUNT
+   claude  default   authed         me@example.com
+   claude  work      authed         me@acme.example · team
+   codex
+   gemini
+   ```
+
+   The Claude rows are the ones `dl --claude-profiles` lists, less the named
+   profiles with no credential, since a launch naming one of those refuses. A
+   named row becomes `--claude --claude-profile <name>`, and `default` passes no
+   profile. codex and gemini have one row each, because `dl` forwards one login
+   for each of them.
+
+   Choosing a row puts its flags in front of your line and parses the line again,
+   so every rule of the line still holds. A row the line would refuse is not
+   shown: `--remote-control` on the line leaves only claude, and `--effort` leaves
+   out gemini. A typed `--codex` or `--claude` shows only that agent's rows, and a
+   typed `--claude-profile` skips this picker. So does a list of one row.
+2. **The model.**
+3. **The effort**, for the agents that have one.
+
+Each picker lists your recent choices first, newest at the top. For the model and
+the effort that is per agent, followed by `default`, then a few suggestions for a
+first run. The cursor starts on the
+first row, so one Enter repeats the last launch. Type to filter the rows. Type a
+name that no row holds and Enter uses it as typed, which is how a model that came
+out yesterday is chosen. It is listed from then on. Matching is exact rather than
+fuzzy for that reason: a fuzzy match finds a listed row for almost any query.
+Where the name is part of a listed row, such as `gpt-5.5` beside `gpt-5.5-codex`,
+Alt-Enter uses the text as typed instead of the row.
+
+A setting that a flag on the line already gave is not asked for, so
+`aid --model opus --effort max <ws>` goes straight to the prompt. A line with a
+prompt on it asks for nothing, as before.
+
+**Esc cancels the launch.** In a picker Esc and Ctrl-C are keys, not signals, so
+`aid` stops the background boot itself, the same way a Ctrl-C at the prompt editor
+does: its `devpod up` is killed and its staged token file is removed. The exit
+status is 130.
+
+The recent choices live in `aid-recent.tsv` in the devlaunch cache, so
+`XDG_CACHE_HOME` scopes them with everything else. A file that is missing or
+broken reads as no history, and never stops a launch.
+
+With no `TERM`, with `TERM=dumb`, or on a terminal whose size reads as zero,
+there are no pickers. The prompt editor still opens and every setting stays at
+its default.
+
+## The prompt editor
+
+After the pickers, `aid` draws a `> ` prompt under one line that says what is
+booting. It is a small editor of its own, not the terminal's line mode, because
+the line mode could not take a paste: it holds 4096 bytes of a line, it
+submitted at the first line break of a paste, and the lines of a paste that came
+a moment late went on to the agent as keystrokes.
+
+- **A paste is text, line breaks included.** The editor turns on the terminal's
+  bracketed paste, so it knows where a paste starts and ends, and a line break
+  inside one never submits. A terminal without bracketed paste still sends a
+  paste faster than anybody types, so an Enter with more input right behind it
+  is read as a line break too. Windows line ends are one break.
+- **Enter submits. Alt-Enter or Ctrl-J adds a line.**
+- **Backspace, Ctrl-U and Ctrl-W** delete a character, the line and a word.
+  There is no cursor to move: the arrow keys do nothing, rather than printing
+  `^[[D`.
+- **An empty Enter, or Ctrl-D on nothing,** starts the agent's plain session.
+- **Ctrl-C stops the boot** and exits with 130, as Esc does in a picker.
+
+A prompt taller than the screen shows its last lines under a line that counts the
+ones not shown. All of it is sent. Keys typed after the Enter are not read, so
+they reach the agent.
+
 ## `aid resume`: back into a session after a restart
 
 ```bash

diff --git a/rust/Cargo.lock b/rust/Cargo.lock
diff --git a/rust/Cargo.toml b/rust/Cargo.toml
@@ -11,7 +11,7 @@ members = [
 # The single source of the version (docs/rust-rewrite-plan.md: cutover ships
 # 0.1.0, version read from Cargo.toml).
 [workspace.package]
-version = "0.58.0"
+version = "0.59.0"
 edition = "2024"
 license = "MIT"
 repository = "https://github.com/blooop/devlaunch"
@@ -36,6 +36,8 @@ toml = "0.9"
 toml_edit = "0.23"
 clap = { version = "4", features = ["derive"] }
 skim = "0.20"
+# The prompt editor's column widths. skim already brings this version in.
+unicode-width = "0.2"
 # The terminfo lookup skim's terminal setup makes and unwraps, made first so a
 # `TERM` it would panic on, or draw nothing with, can be replaced (dl's
 # `select.rs`). The version skim resolves, so the answer is the one skim would