Trần Ngọc Sơn

[Superman @thesupermanmx]

China open-sourced a fully local memory system that gives any AI agent human-like long-term recall.

No cloud. No API bills. No flat vector piles.

- 61.38% fewer tokens per session
- 76% persona accuracy (up from 48%)
- 51.52% higher task success rate
- works with any agent

100% Open Source.

[0xMarioNawfal @RoundtableSpace]

AGENT-REACH GIVES ANY AI AGENT FREE ACCESS TO X, REDDIT, YOUTUBE, GITHUB, AND 10+ MORE PLATFORMS WITHOUT PAID APIS.

[Trần Ngọc Sơn @nomiss194]

@XFreeze @yunta_tsai If the price goes down, will the cursor limit increase even more?

[Trần Ngọc Sơn @nomiss194]

@OfficialLoganK @jun_song We just need a good model comparable to Opus 4.8.

[Cohere @cohere]

Cohere has released open-source models Transcribe, Command A+, and North Mini Code so far this year, all available under Apache 2.0. With more to come...

Own your AI.

[0xMarioNawfal @RoundtableSpace]

Andrej Karpathy's modular knowledge vault strategy turns static Obsidian notes into a persistent context engine for AI workers.

[0xMarioNawfal @RoundtableSpace]

THIS AR APP MAKES INVISIBLE WIFI SIGNALS VISIBLE SO YOU CAN FIND DEAD ZONES AND FIX YOUR CONNECTION.

[LlamaIndex 🦙 @llama_index]

As of LiteParse v2.8.0, image → PDF conversion no longer depends on ImageMagick. Instead, it’s now handled natively in Rust 🦀

The result? Conversion is now 1.2× to 7.2× faster, depending on the image format⚡️

But the bigger win is what this enables: we’re one step closer to making LiteParse fully self-contained, with zero external system dependencies 🚀

If you’re using LiteParse, make sure to upgrade and to check out the project on GitHub 👉️ https://t.co/jgcEJgICQ0

[🚨 AI News | TestingCatalog @testingcatalog]

Airtap has launched an AI agent that users can talk to via iMessage or RCS.

Airtap AI agent focuses on tasks that normally run entirely within mobile apps, allowing users to run them on a recurring basis.

For example, Airtap integrates with Google Maps, allowing users to preview and create routes on the fly!

[Avi Chawla @_avichawla]

RAG vs. Graph RAG, explained visually!

RAG has many issues.

For instance, imagine you want to summarize a biography, and each chapter of the document covers a specific accomplishment of a person (P).

This is difficult with naive RAG since it only retrieves the top-k relevant chunks, but this task needs the full context.

Graph RAG solves this.

The following visual depicts how it differs from naive RAG.

The core idea is to:
- Create a graph (entities & relationships) from documents.
- Traverse the graph during retrieval to fetch context.
- Pass the context to the LLM to get a response.

Let's see how Graph RAG solves the above problem.

First, a system (typically an LLM) will create a graph from documents.

This graph will have a subgraph for the person (P) where each accomplishment is one hop away from the entity node of P.

During summarization, the system can do a graph traversal to fetch all the relevant context related to P's accomplishments.

The entire context will help the LLM produce a complete answer, while naive RAG won't.

Graph RAG systems are also better than naive RAG systems because LLMs are inherently adept at reasoning with structured data.

Once the right architecture is in place, the next leverage point is efficiency.

Most RAG architectures rely heavily on vector search, and that layer can be made 32x more memory efficient using binary quantization.

I covered the full implementation in the article below.

👉 Over to you: Which RAG architecture are you running in production?

[Trần Ngọc Sơn @nomiss194]

@OfficialLoganK Please release Gemini 3.5 Pro as soon as possible; it just needs to be better than 3.1 Pro. Best wishes!

[X Freeze @XFreeze]

“No one has ever built a fully reusable orbital heat shield”

When Starship reenters at orbital velocity, it comes in like a flaming meteor….a raging ball of fire. Most heat shields partially melt or get destroyed during reentry

SpaceX is solving this incredibly hard problem right now with the most advanced heat-shield tiles ever flown

Starship V3 upgrades are improving tile retention and durability with every flight. The progress is insane

[Trần Ngọc Sơn @nomiss194]

@atomic_chat_hq Compare it to the cursor.

[Mayank Vora @aiwithmayank]

GPT-5.6 Sol is ridiculously good at building beautiful websites.

But if your prompt says “make me a landing page,” you will get the same generic AI design as everyone else.

Here are 10 copy-paste prompts that make almost any website look expensive:

[Andrew Ng @AndrewYNg]

Announcing OpenWorker! An open-source agent that doesn't just chat with you, but delivers finished work -- like hand you a polished document, send a slack message, or update a calendar entry.

Ask it to prepare a customer brief, untangle your calendar, draft a report, or triage a Slack alert. It works across your files and everyday tools, produces the deliverable, and checks in before doing anything consequential.

OpenWorker runs on your Mac, with Windows support coming soon. It does not lock you into any one model. Bring your own API key and run it with GPT 5.6 Sol, Claude Fable, Gemini 3.6, an open weight model (like Kimi, GLM, DeepSeek, Inkling), or Ollama to keep your data local. Your data does not leave your machine except through an LLM provider and integrations that you choose.

@rohitcprasad and I are building OpenWorker because AI coworkers are an important way to get work done, and we want there to be an open, privacy-preserving, model-independent option. Check it out and let us know what you think!

Try it out: https://t.co/P0mGnI1o31 (requires your own API key)
Source code: https://t.co/NYCiTD6hSq

[Trần Ngọc Sơn @nomiss194]

@N_and_ni Grok

[Musfara Ai @MusfaraAi]

🚨 BREAKING: Claude can now design logos that look like they cost $10,000.

I tested it myself.

The final result cost me exactly $0.

Here are the 7 prompts I used:

Save this for later

[Trần Ngọc Sơn @nomiss194]

@patloeber It works quite well in AIstudio, but in AG, I'm constantly lazy, forgetful, and make mistakes.

[Trần Ngọc Sơn @nomiss194]

@patloeber I use Gemini in AIstudio and it works quite well, but it's poor in AG. Why is that?

[Trần Ngọc Sơn @nomiss194]

@OfficialLoganK I use Gemini in AIstudio and it works quite well, but it's poor in AG. Why is that?