- Related ids: RSH-20260409-006, RSH-20260409-007, RSH-20260409-009, DEC-20260409-007, LOG-20260409-012, LOG-20260409-013, IBX-20260412-002

- Treat `microsoft/vscode` as a newly strengthened contender after the reported Insiders `Agent Mode` shift, but only if direct verification shows a workable seam between the IDE substrate and a Superhuman-owned workspace/runtime layer.

- Provenance: `IBX-20260412-002`

- Source: `https://x.com/laogui/status/2041497562574926216`

- Status: reference signal only; not direct verification of a fork candidate

What the signal says:

- A referenced 2026-04-07 post claims VS Code Insiders is developing a new `Agent Mode` with a simplified agent-oriented layout rather than staying purely editor-first.

- The reported shape converges on the same multi-pane agent-manager pattern already appearing elsewhere: task list on the left, execution surface in the middle, and change preview plus file management on the right.

- The reported runtime mix matters because it combines local interaction, background agent execution, and cloud-connected modes while keeping Git, terminal, and multi-workspace affordances nearby.

Why this matters to Superhuman:

- It strengthens the "build down from a mature IDE substrate" side of the substrate debate. If VS Code itself is moving toward a simplified agent cockpit, the line between IDE and agent manager may keep collapsing.

- More importantly, it changes the competitive question. The issue is not just whether an IDE substrate can be reduced into an agent cockpit, but how mature VS Code's own agent-mode evolution already is compared with agent-native frontends that started earlier.

- The key product question is no longer only "IDE fork or not." It is whether VS Code's agent-mode movement is already mature enough to rival or surpass the newer agent-focused frontends on the specific surfaces Superhuman cares about.

Current read:

- Upgrade VS Code's agent-mode evolution from background context to an active contender in the build-up versus build-down question.

- Treat the Insiders shift as a real change in landscape, not just a curiosity: it materially improves the case for forking or heavily borrowing from VS Code compared with the earlier read.

- The next evaluation pressure is maturity, not prejudice. Compare VS Code's agent-mode completeness, coherence, and likely velocity against the agent-first frontends that had a head start on this archetype.

- Seam quality still matters, but mainly as a second-order question after maturity: if VS Code is catching up quickly enough on the agent-manager shape, then the remaining issue is whether Superhuman can adapt that surface without losing ownership of runtime state and product direction.

- Treat this as pressure on the evaluation model: a candidate no longer wins merely by being thinner than VS Code. It also has to explain why it beats an increasingly agent-native VS Code path on work focus, surface maturity, replaceable state boundaries, and Superhuman-owned runtime seams.

- How mature is VS Code's own agent-mode movement relative to the agent-first frontends that had a head start, and is it already mature enough to justify treating VS Code as a top-tier fork or borrowing candidate?

- If VS Code is now a strong contender, what is the smallest acceptable fork or adaptation seam: custom workbench surface, extension-host-first product, narrowed distribution, or a deeper product fork?