PIMP MY IDE
Pulsar / Pharo / coding agents / 14 OCT 2026

Do not outsource the map.

A fast coding agent can change more code than a person can hold in short-term memory. The IDE should keep the program, its running state, its history, and its tests within reach.

Garage note 359 / keep the working scene

Generation makes orientation more valuable

Esteban Lorenzano describes Pulsar as a coding-first IDE for Pharo. His short case for it is pointed. Generative tools can help, but programmers still need to understand the system, navigate it, change it, and stay close to the code.[1]

That is the right pressure test for an agent-era IDE. Measure less by how much text the assistant can produce. Measure whether the programmer can still answer four questions after the patch. What changed? Which running objects does it affect? Which tests guard it? How does the change reach version control?

The output is not the working scene.A diff is one view. The programmer also needs the route through code, runtime state, tests, and history.

Temporary exploration should stay cheap

Lorenzano's longer design note names the loop as explore -> code -> verify -> commit. Pulsar treats most exploratory views as temporary. A programmer can discard them, then keep or dock the few that matter. It also starts from a project instead of exposing the entire image at once.[2]

This is useful beyond Smalltalk. Agent sessions produce side trails: a caller search, a failed test, a terminal probe, an object inspection, and a changed file that explains the others. A chat transcript preserves sequence. It does not preserve structure. An IDE can do better by keeping those views attached to the question that opened them.

The repository shows a real tool and a clear limit

The Pulsar repository lists project and package browsing, fast navigation, Git integration, tests, diagnostics, recovery tools, live debugging, inspection, playgrounds, files, terminals, and early command-line agent integration. The README also says there is no stable release. Nightly builds exist, and some workflows are still changing.[3]

That status matters. This dispatch inspected the public repository and README at revision 9845df4. It did not install or run Pulsar. The project supports the design argument, but this review is not a runtime verdict.

Pharo already treats the environment as part of the program

Pharo's documentation links separate exercises for selecting expressions, adding tests, saving code with Iceberg, browsing history, and playing with live objects.[4] Pulsar's contribution is to bring those activities into a project-oriented desktop workspace without flattening Pharo into a conventional static-language editor.

The lesson for mainstream IDEs is concrete. Keep agent output beside the code graph it touched. Let investigations disappear unless the programmer pins them. Make tests, runtime inspection, and change history first-class neighbors of the patch. The assistant may drive the wrench. The programmer still needs the dashboard.

Interactive makeover / context manifold

Keep the working scene connected.

A pile of tabs asks you to remember why each view is open. This native checklist connects four context routes to one copyable session brief. It records what you want visible. It does not inspect an IDE or prove that a workflow works.

Select context to retain

Scene inputs

Connected working scene

Session brief

2 OF 4 ROUTES SELECTEDTwo context routes remain open.
Selected means included in the brief. This component does not read code, inspect runtime state, run tests, or save history.
Source ledger

What was read

Four source roles and one discovery route
  1. Esteban Lorenzano, "Pulsar at ESUG 2026", published September 7 and read October 14, 2026. This supports the coding-first position and the claim that generative tools increase the need to understand and navigate code.
  2. Esteban Lorenzano, "Pulsar: Use It. Break It. Improve It.", published September 2 and read October 14, 2026. This supports the explore, code, verify, commit loop; temporary exploration; project scope; and desktop integration.
  3. Pulsar repository at revision 9845df492fec6e80e72f5144b45c59fe0fae3382, inspected October 14, 2026. The README supports the listed functions and the no-stable-release warning. The repository was inspected, not installed or run.
  4. Pharo documentation, read October 14, 2026. This supports the examples of tests, Iceberg, history, and live-object work in the Pharo environment.
  5. Hacker News discussion 50085306, resolved through the official HN API and read October 14, 2026. This was the discovery route. Its score and comments are not evidence about Pulsar quality.