Nok

23 • Creative and still learning • Building aura-ctx

[Nok @thisisenoch]

I’ve been frustrated for a while with how every AI tool has its own incomplete version of “me”.
Claude doesn’t know what I told Cursor yesterday.
ChatGPT has no idea about my coding preferences.
Gemini forgets my writing style.
So I started building something simple to fix that.

[Nok @thisisenoch]

I genuinely love using Claude, but things are getting tough this time. I thought I was the only who noticed it. Might be time to start looking at open-source alternatives

[Scott Breitenother @s_breitenother]

Did anyone catch what Anthropic quietly changed? Starting with enterprise, businesses can no longer buy Claude subscriptions. API pricing only. This is bigger than it sounds.

[Nok @thisisenoch]

Everyone here is talking about AI memory layers, harnesses, tools, and plugins.
Almost no one is talking about Identity. The layer that permanently connects your AI to your personal context, values, history, and style. identity and memory are different

[Nok @thisisenoch]

@metedata @Fried_rice @Scobleizer That's amazing

[Nok @thisisenoch]

Absolutely right. We can't complain 😂

[Mete Polat @metedata]

@Fried_rice @Scobleizer Anthropic is now officially more open than OpenAI

[Nok @thisisenoch]

MCP is so back

[Peter Steinberger 🦞 @steipete]

The next version of OpenClaw is also an MCP, you can use it instead of Anthropic's message channel MCP to connect to a much wider range of message providers.

(I know, this is awkward)

[Nok @thisisenoch]

Google shipped “portable memory” for Gemini. but it’s still one-way. You can import your chats and preferences into Gemini.
But there’s no way to programmatically read, export, or reuse that context elsewhere

That’s not portability. That’s ingestion.

[Google Gemini @GeminiApp]

Switching to Gemini from other AI apps just got easier.

Starting to roll out today on desktop, you can now bring your preferences and chat history into Gemini, so you can pick up right where you left off in just a few clicks. 🧵

[Nok @thisisenoch]

@lucasmaes_ Really love how the spatial structure emerges naturally in the t-SNE with no hand-crafted bias. It just comes from the objective. Closing the gap with DINO-WM while staying fully pixels-only feels like the right direction for world models. Congrats to the team!

[Nok @thisisenoch]

Finally! Stable end-to-end JEPA without any tricks or heuristics. LeWorldModel learns world models directly from pixels with just 15M params on 1 GPU and full planning in <1 second. This changes everything

[Lucas Maes @lucasmaes_]

JEPA are finally easy to train end-to-end without any tricks!

Excited to introduce LeWorldModel: a stable, end-to-end JEPA that learns world models directly from pixels, no heuristics.

15M params, 1 GPU, and full planning <1 second.

📑: https://t.co/cpTzgvbTS0

[Nok @thisisenoch]

@rezoundous Claude 'cuz it helps you think better before you build.

[Nok @thisisenoch]

@md_kasif_uddin Claude, perplexity

[Nok @thisisenoch]

Sunday, 3AM.

I should be asleep,
but I just finished reading all the AI articles I bookmarked…

Interesting stuff.

[Nok @thisisenoch]

@DecentCloud_org Fair point. Device sync and team sharing are on the roadmap (opt-in, client-side encrypted). For now it’s git-friendly YAML — push your packs to a private repo and you’ve got sync across machines in 30 seconds.

[Nok @thisisenoch]

I’ve been frustrated for a while with how every AI tool has its own incomplete version of “me”.
Claude doesn’t know what I told Cursor yesterday.
ChatGPT has no idea about my coding preferences.
Gemini forgets my writing style.
So I started building something simple to fix that.

[Nok @thisisenoch]

It’s still very early. If you often switch between multiple AI tools and hate having to re-explain who you are every time, I’d really appreciate your feedback.
https://t.co/uJQHkiOcPw

[Nok @thisisenoch]

aura-ctx lets you define your identity once — stack, rules, preferences, style — in clean YAML files.
Then it serves that identity to all your tools via MCP (Claude, Cursor, ChatGPT, Gemini ... )
Everything stays local. No cloud. No lock-in. Just one source of truth you control

[Nok @thisisenoch]

@98CRuE6 @WesRoth Everything stays local. https://t.co/uJQHkiOcPw

[Nok @thisisenoch]

@98CRuE6 @WesRoth That’s exactly what aura-ctx is trying to solve. It acts as a personal central identity layer:
You define who you are (stack, style, rules, preferences) once in clean YAML files, then serve it automatically to all your tools via MCP (Claude, Cursor, ChatGPT Desktop, Gemini, etc.)

[Nok @thisisenoch]

@AI_Nate_SA @joshwoodward This is exactly what aura-ctx does — define your identity once in YAML, serve it to Claude, ChatGPT, Cursor, and Gemini via MCP. Bidirectional, not just one-way import. Open source: https://t.co/1aUb1assP3

[Elon Musk @elonmusk]

Optimus

[Nok @thisisenoch]

@ninedol @qwerty_ytrevvq Exactly. And if a good prompt is a spec, then a good context is the project brief the spec is built on. The model needs to know who it’s working for before it knows what to do.