Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors
Rust
Revert "docs(research): disambiguate Liquid AI / LFM as not an agent-memory tool"
AkitaOnRailse7a1b9e28c23
diff --git a/docs/research-2026-landscape.md b/docs/research-2026-landscape.mdindex 4b816177..2a455d43 100644--- a/docs/research-2026-landscape.md+++ b/docs/research-2026-landscape.md@@ -287,22 +287,6 @@ lead, not the whole chart. Two mild flags: Letta ships commits actively but itslast *tagged* release is from May 2026 (formal releases lag the code), andLangMem/Honcho publish no GitHub *releases* (PyPI/tag-only versioning).-**Disambiguation — Liquid AI / LFM is not a memory tool.** A quick note to-preempt a recurring confusion: **"LiquidLM"** does not exist as an agent-memory-product under any spelling. Every search resolves to **Liquid AI's LFM (Liquid-Foundation Models / LFM2 / LFM2.5)** — a family of *device-native foundation-models* (hybrid non-transformer, on-device/edge, e.g. LFM2.5-1.2B-Thinking-fitting in <1 GB; LFM Open License v1.0, arXiv:2511.23404, actively developed by-Liquid AI under `github.com/Liquid4All`). The name collision is that LFM's-headline is *memory efficiency* (small KV-cache/RAM footprint) — memory in the-**inference/hardware** sense, not the **agent-persistence/recall** sense this-report covers. So LFM is **not** in any camp here and is **not** a competitor or-migration target. It is, if anything, **complementary**: a small local model like-LFM2.5 is exactly the kind of provider that could run ai-memory's *opt-in*-consolidation/lint/embedding fully on-device — a candidate backend for the-`ai-memory-llm` provider traits, not a rival to the memory store. No comparison-row is warranted; this note is its record.-## 4. Research developments worth knowing- **"Rethinking How to Remember: Beyond Atomic Facts in Lifelong LLM