PIMP MY IDE
Coding-first IDEs / context retention / 14 OCT 2026

Let weak context fall away.

Every lookup does not deserve a tab. Keep the method, test, object, or diff only when it still explains the change.

Garage note 361 / a working scene needs gravity

An investigation is not a collection

Pulsar is a new coding-first IDE for Pharo. Its author describes programming as a graph of questions. A working session moves through methods, tests, senders, objects, experiments, repositories, files, and terminals.[1]

The useful design move is not the longer tool list. Pulsar treats most exploration as temporary. A sender lookup or inspected object can disappear after it answers the question. The programmer can pin, dock, or detach the few results that still explain the work.[1]

Persistence should be a decision, not a side effect of opening a tool.Keep context that explains the active change. Release context that already answered its question.

Agents raise the cost of stale context

A coding agent can open files, run tests, inspect failures, and generate several candidate edits in one turn. A flat tab strip preserves the order of those actions. It does not show which result still matters.

A better working scene keeps the current method near its protecting test, relevant runtime state, and recovery point. The scene should also show whether each item is live, stale, or only a note. This turns context into an explicit review aid rather than a browser history.

JetBrains takes a related route in Rider 2026.2. It exposes coverage data, profiler snapshots, refactoring support, and framework guidance to coding agents instead of making them reconstruct project knowledge from files and terminal output.[2] Pulsar comes from a different environment, but both projects treat IDE knowledge as structured working material.

Live state needs labels

Pharo is a live environment. Its official site describes editing methods while the program is running, inspecting live objects, and changing code without a compile and restart cycle.[3] That proximity is powerful. It also makes freshness part of the interface contract.

An inspected object can outlive the state that made it interesting. A test result can predate the latest edit. A Git diff can point at a different recovery boundary. The IDE should keep the item close and mark its age or revision. Proximity without provenance can mislead.

The artifact is early

The Pulsar repository says the project has nightly builds but no stable release. It targets Pharo 14 and lists project browsing, Git history, tests, debugging, inspection, files, terminals, and early command-line agent integration.[4] The repository was inspected at revision 9845df492fec6e80e72f5144b45c59fe0fae3382. Pulsar was not installed or run for this note.

The useful lesson does not depend on a runtime verdict. IDEs should make temporary context cheap and retained context deliberate. That rule matters more when an agent can create context faster than a person can review it.

Interactive makeover / context gravity bench

Pin what earns a place.

A tab strip remembers what opened. This native pinboard asks what should stay near the active change, shows the retained scene, and drafts a copyable review brief. It does not inspect an IDE.

Choose retained context

Investigation tray

Retained working scene

Review pinboard

2 ITEMS RETAINED / BRIEF STRUCTURE READYMethod and Test stay near the active change. Evidence still needs review.
This control records your retention choices. It does not read code, inspect runtime state, run tests, verify a revision, or save an IDE workspace.
Source ledger

What was read

Four sources and one discovery route
  1. Esteban Lorenzano, "Pulsar: Use It. Break It. Improve It.", read October 14, 2026. This supports the project-scope, working-scene, temporary-exploration, and alpha-status claims. These are the author's design claims.
  2. JetBrains, Rider 2026.2 release post, read October 14, 2026. This supports the Rider coverage, profiler, refactoring, framework guidance, and agent-integration claims.
  3. Pharo features, read October 14, 2026. This supports the live-object, running-program, and edit-without-restart description.
  4. Pulsar README at revision 9845df492fec6e80e72f5144b45c59fe0fae3382, inspected October 14, 2026. This supports the nightly, stability, target-version, implemented-tool, and early agent-integration claims. The application was inspected, not installed or run.
  5. Hacker News item 50085306, resolved through the official HN API and read October 14, 2026. This was the discovery route. Its score is not evidence about product quality.