Nvwa Lj

[Nvwa Lj @nvwalj]

Most engineers don't know their AI coding agent writes 5–20 MB of memory + session telemetry into ~ per day.

Built a native macOS viewer for it: every CLAUDE.md, AGENTS.md, GEMINI.md, plus the .jsonl sessions VSCode chokes on.

v0.4.6 just shipped 🧠

https://t.co/3Il1UZZzT7

[Nvwa Lj @nvwalj]

@0xPrajwal_ PLAN.md becomes PLAN-V2-FINAL.md by Thursday — the actual fix is one canonical entry that imports the rest in load order, but you learn this only after a few "wait which md did Claude actually read" sessions. Happy to look at anyone's setup free if it's a real mess.

[Nvwa Lj @nvwalj]

v0.4.8 — AIMR now reads ~/.qwen (Alibaba Qwen Code) and ~/.kimi (Moonshot Kimi CLI).

10 agents auto-discovered now. Same 3 MB universal binary, still no network.

https://t.co/3Il1UZZzT7

[Nvwa Lj @nvwalj]

AI Memory Reader v0.4.6 shipped.

New since v0.4:
- Strict file filtering (no JSONL noise)
- Codex session viewer
- Update banner
- iCloud bookmark sync

3 MB universal binary. Free + open source.

https://t.co/mvq7cQ5wBA + open source.

https://t.co/3Il1UZZzT7

[Nvwa Lj @nvwalj]

@BeauJohnson89 Thanks Beau — the "category, not the viewer" framing is exactly the angle I was hoping someone would notice. Appreciate the boost!