Xiaming Chen

Research on AI continual learning, neurosci. inspired AI, neurosymbolic tech. Data & ai guy @alibaba_cloud. Founder of @mirasoth. Ph.d @sjtu1896.

[Xiaming Chen @drxiamingc]

To make this practical, I turned the workflow into agentic skills: init specs, validate constraints, generate history, enforce consistency.

I put them here: https://t.co/7tM5MOA8HG

If you’re building something complex, this method worth trying. It changes how AI coding feels.

[Xiaming Chen @drxiamingc]

What determines the upper limit of an agentic and RSI technology, is the quality of context.

[Xiaming Chen @drxiamingc]

Small models like Jev may be the future. Let’s see how long it takes us to train an advanced world model like Jev from scratch.

The bigger question: can we build a way to produce these models at scale, almost like a gigafactory for intelligence?

[Xiaming Chen @drxiamingc]

A deepagents powered agent to make human OUT of the loop https://t.co/RCedwBmDYT

[Xiaming Chen @drxiamingc]

opensoothe showcase https://t.co/Xl738wqZIw 来自 @YouTube

[George @odysseus0z]

https://t.co/GqhyL4KRJw

[Xiaming Chen @drxiamingc]

#PlatonicCoding deserves a try to accelerate your development. This would make your AI coding more fluent.

https://t.co/7tM5MOA8HG

[Xiaming Chen @drxiamingc]

#PlatonicCoding vs. Prompting

[Xiaming Chen @drxiamingc]

I love coding and found a more efficient method of AI coding. @karpathy maybe you will be interested and let more people master Platonic Coding. 🥳🥳

[Xiaming Chen @drxiamingc]

To make this practical, I turned the workflow into agentic skills: init specs, validate constraints, generate history, enforce consistency.

I put them here: https://t.co/7tM5MOA8HG

If you’re building something complex, this method worth trying. It changes how AI coding feels.

[Xiaming Chen @drxiamingc]

To make this practical, I turned the workflow into agentic skills: init specs, validate constraints, generate history, enforce consistency.

I put them here: https://t.co/7tM5MOA8HG

If you’re building something complex, this method worth trying. It changes how AI coding feels.

[Xiaming Chen @drxiamingc]

I built Grism, an AI-native, neurosymbolic hypergraph database in modern design, from concept to working prototype in THREE days. Honestly, I didn’t expect it to move that fast for something this complex.

https://t.co/bHUl2JCSgG

[Xiaming Chen @drxiamingc]

What changed next was how I worked with agents. I stopped explaining intent and started referencing structure. Conversations became about invariants, definitions, and allowed changes. When specs are explicit, agents don’t guess. They operate inside a shared world.

[Avi Chawla @_avichawla]

8 AI model architectures, visually explained:

Everyone talks about LLMs, but there's a whole family of specialized models doing incredible things.

Here's a quick breakdown:

1. LLM (Large Language Models)

Text goes in, gets tokenized into embeddings, processed through transformers, and text comes out.

↳ GPT, Claude, Gemini, Llama.

2. LCM (Large Concept Models)

Works at the concept level, not tokens. Input is segmented into sentences, passed through SONAR embeddings, and then uses diffusion before output.

↳ Meta's LCM is the pioneer.

3. LAM (Large Action Models)

Turns intent into action. Input flows through perception, intent recognition, task breakdown, then action planning with memory before executing.

↳ Rabbit R1, Microsoft UFO, Claude Computer Use.

4. MoE (Mixture of Experts)

A router decides which specialized "experts" handle your query. Only relevant experts activate. Results go through selection and processing.

↳ Mixtral, GPT-4, DeepSeek.

5. VLM (Vision-Language Models)

Images pass through a vision encoder, text through a text encoder. Both fuse in a multimodal processor, then a language model generates output.

↳ GPT-4V, Gemini Pro Vision, LLaVA.

6. SLM (Small Language Models)

LLMs optimized for edge devices. Compact tokenization, efficient transformers, and quantization for local deployment.

↳ Phi-3, Gemma, Mistral 7B, Llama 3.2 1B.

7. MLM (Masked Language Models)

Tokens get masked, converted to embeddings, then processed bidirectionally to predict hidden words.

↳ BERT, RoBERTa, DeBERTa power search and sentiment analysis.

8. SAM (Segment Anything Models)

Prompts and images go through separate encoders, feed into a mask decoder to produce pixel-perfect segmentation.

↳ Meta's SAM powers photo editing, medical imaging, and autonomous vehicles.

What else would you add?

👉 In my next tweet, I have shared my latest AI Engineering Guidebook (380+ pages) with 150+ core lessons.

Everything is 100% open-source (25k+ stars 🌟)

[Xiaming Chen @drxiamingc]

https://t.co/2NMq2ZLuPq

[Xiaming Chen @drxiamingc]

🚀 **cogents-core v1.0.0 is now stable!**

A comprehensive cognitive agentic framework for building intelligent multi-agent systems.

✨ Features:
• Multi-LLM support (OpenAI, Ollama, LiteLLM, etc.)
• DAG-based goal management (Goalith)
• Tool orchestration with MCP and more

[Xiaming Chen @drxiamingc]

How to crack pitfalls of LLM on continual learning. Here is my proposed solution, Composable Learning. A position post here,
https://t.co/Nu2IUkuTX1

[Quant Science @quantscience_]

343+ Quant and Algorithmic Trading Projects in Python

This is how to get your hands on them for free.

(a thread)

[Pedro Domingos @pmddomingos]

I've found the path to AGI:
https://t.co/NRvZViBVUc

[Philipp Schmid @_philschmid]

What can we learn from Anthropic’s Blog about building tools for Agents? Here are the key principles I extracted for building Tools and MCP Servers for Agents:

- Consolidate workflows: Implement `schedule_event` instead of separate `list_users`, `get_availability`, and `create_event` tools.

- Evaluate with real tasks: Test with goals like "find the project lead and schedule a meeting," not just "get user."

- Automate metrics: Programmatically run evaluations to track success rates, token counts, and tool call frequency.

- Let AI refactor: Feed failed evaluation transcripts to an LLM to get suggestions for improving your tool's code.

- Namespace tools clearly: Use prefixes to differentiate tools, such as `jira_search_issues` vs. `github_search_issues`.

- Return meaningful context: Prefer human-readable names like "Jane Doe" instead of cryptic `usr_7a3b4c9` IDs.

- Enforce token efficiency: Implement pagination and filtering for tools that can return long lists of results.

- Make errors helpful: Instead of `Invalid input`, return `Error: Missing 'user_id'. Example: search(user_id=12345)`.

- Guide with messages: If output is cut, guide with "Showing 10 of 100 results. Use the 'page' parameter for more."

- Use unambiguous parameter names: Use `user_email` or `user_id` instead of a generic `user` parameter.

- Define formats explicitly: In parameter descriptions, specify requirements like "Date must be in YYYY-MM-DD format."

- Provide examples in descriptions: Add an `Example: search_query='status:open'` to the parameter's help text.