Basic Memory

Basic Memory lets your AI write, read, and reuse what matters: your notes, prompts, and instruction | Cross-LLM| Local-first & open source OR Basic Memory Cloud

[Basic Memory @basic_memory]

The "AI Memory Layer" is definitely the LLM vibe-coding equivalent of the TodoMVC. The thing is though that it is actually a tough and open ended problem.

My new marketing line idea for Basic Memory (AI slop pun intended) is:

>We've been quietly building the AI Memory you wanted when you started vibe-coding your own.

[Gergely Orosz @GergelyOrosz]

One interesting trend:

I’m seeing *so many* VC-funded, internally built infrastructure and bootstrapped solutions around “building a context layer for engineering teams.” Aka trying to solve the problem of “if only eg Claude Code had the context from all your other systems”

[Basic Memory @basic_memory]

Did we get you interested? Read on...
https://t.co/y8Z7Ix1elP

[Basic Memory @basic_memory]

Google's New Open Knowledge Format Is Basically Basic Memory's Thesis

Did Basic Memory just become philosophically aligned with Google? Kind of.

[Basic Memory @basic_memory]

@pkm_tk111 Thanks for the excellent PR @pkm_tk111 !

[Basic Memory @basic_memory]

My take on Fable 5: I think this thing works by the hour. It is soooo slow to do simple things then gets defensive.

[antirez @antirez]

I'm so tired of all the AI bots here. Note, however, how at the base of everything wrong there is human stupidity and greed.

[Basic Memory @basic_memory]

I think @samuelcolvin is right.

[Samuel Colvin @samuelcolvin]

What do I get if I'm right?

> OpenAI and Anthropic's next lock-in play: Databases of coding intent

https://t.co/52qecl3rE9

[Basic Memory @basic_memory]

Holy Sh**t. Now my feed is full of rando dudes' faces all of a sudden. All reaction videos! The absolute worst.

I hope this terrible idea is a fad that fades quickly. So terrible. Its so disappointing that this is where time and energy was spent instead of doing something actually worthwhile. For everyone who was forced to implement this, I'm sorry. Quit this job and do something better. You are wasting your talent.

[Nikita Bier @nikitabier]

Commentary is one of the most important pillars of X. And sometimes the best way to share your thoughts is with video.

Today we're launching a whole new way to make them:
React with Video

Tap the repost button and start recording with green screen, split screen, or picture-in-picture.

Now available on iOS

[Basic Memory @basic_memory]

@t_blom Maybe the "missing piece", per your analogy was the employees who knew how the company operated? The ones you replaced with "AI geniuses".

[Basic Memory @basic_memory]

Am I alone that in thinking that my X feed is mostly AI talking to itself now? I can't stop seeing slopapalooza everywhere.

It makes me think of the Ray Bradbury story "There Will Come Soft Rains", where the machines continue doing their thing after all the people are gone.

[Basic Memory @basic_memory]

I like this. This is now we try to work building Basic Memory. Every time I've had multiple agents doing everything at once I feel like I need to nap for days.

[Ronan Berder @hunvreus]

I've been complaining about AI-bros and their "20 agents working overnight" for a couple of weeks now.

That's getting old. So I thought I'd do the more productive thing and propose what I think is the Right Way™ to develop software with AI.

And I did that by stealing from the iconic agile manifesto.

https://t.co/Ktf1eZfp8T

[Basic Memory @basic_memory]

@pkm_tk111 We'll get it out asap. If you can help us test the embeddings with Japanese characters that would be a big help!

[tk | Obsidian×AIエージェントで無双する大学生 @pkm_tk111]

Basic Memoryの長期記憶の仕組み、かなり理解が深まってきたぞ、う〜〜むよくできてるね

めちゃObsidianライクな構造しとるわ

特にbuild_contextっていうMCPツールが知識グラフの肝となってるぽいね

参考：
https://t.co/31QJUOqdN6

=====

検索システムが面白くて

人間が質問する
↓
AIが質問意図を解釈する
↓
起点ノートが分からなければ `search_notes` で探す
↓
候補ノートの `memory://...` という内部住所を取得する
↓
そのノートを起点に `build_context` を実行する
↓
起点ノートの本文・Observations・Relationsを見る
↓
Relationsを辿って関連ノートへ広げる
↓
関連ノートの本文・Observations・Relationsも見る
↓
`depth` で何ホップ先まで辿るかを制御する
↓
`timeframe` で��の期間の情報を見るかを絞る
↓
`max_related` で各階層の広がりを制御する
↓
`page_size/page` で返す結果の範囲を制御する
↓
AIが文脈を持った状態で回答する

つまり、Basic Memoryの本質は「検索」だけではない

`search_notes` は入口を探すためのもの
`build_context` は、その入口から知識グラフを辿って、文脈を構築するためのもの

Observationsは、そのノートに書かれている事実や要点。
Relationsは、ノート同士のつながり。
そしてAIは、Relationsを辿りながら関連ノートを広げ、Observationsを読んで意味を理解していく

ここを理解すると、これからのメモリ設計やAIエージェント設計の見え方がかなり変わるな

AIが辿れる形で構造化しておくことが重要だね

Basic Memory、かなり面白いぞ^^

[Basic Memory @basic_memory]

@pkm_tk111 Check out our docs site too. Always adding new features! https://t.co/mBvlevcrZs

[Basic Memory @basic_memory]

@pkm_tk111 We are in the middle of a few things, but I'll look at this asap. Thanks!

[Basic Memory @basic_memory]

Your Notes, Your Bucket: The Storage Layer Behind Basic Memory Cloud | Basic Memory Blog https://t.co/6PWNigHwXd

[Tigris Data @TigrisData]

@basic_memory just shipped per-file diff and merge for AI-edited Markdown notes. Pick any prior version, see what changed, apply the lines you want.

It's built on Tigris bucket versioning. Every edit becomes a new object version. One ListObjectVersions call returns the full revision history. CodeMirror handles the merge UI.

Object versioning has everything git would have (full history, content-addressable IDs) without any of the infrastructure: no commits, no branches, no merge engine on the server.

Read how they used Tigris to make it happen👇

[Tigris Data @TigrisData]

Most AI memory locks your context to one platform. Switch tools, lose everything.

@basic_memory flips it: your knowledge graph as plain Markdown in your own bucket that you can edit directly. Any MCP tool — Claude, ChatGPT, Cursor — reads and writes the same files.

New case study on how a four-person team built it on Tigris. 👇

[Basic Memory @basic_memory]

@TigrisData did a super cool case study on Basic Memory. https://t.co/9ezRJX5CXd. Thanks!

One thing I was really excited about after talking to them was being able to create file level versions automatically using their built in bucket versioning.

[Basic Memory @basic_memory]

Hey @thsottiaux, curious how you’re thinking about MCP Apps support across clients. We have the same FastMCP/Prefab components rendering for @basic_memory in ChatGPT web, ChatGPT Desktop, and Claude Desktop, while Codex sees/calls the tool but shows raw JSON. Is Codex inline app rendering on the near-term path?

Related Codex issue: https://t.co/wFyN1Fh3kK