Mohammed Daba @m_dabaAI

MS Robotics @UESTC1956 | Enjoying AI & Robotics linkedin.com/in/mohammed-da… Chengdu, China Joined June 2013

Tweets

274
Followers

43
Following

538
Likes

607

Mohammed Daba @m_dabaAI

4 months ago

Google has the most powerful LLM (Gemini 2.5 pro) and the most powerful coding agent (AlphaEvolve). I bet I/O event goona be huge, training next LLM beast on the result of AlphaEvolve on the hardest math, coding tasks and iteratively do that with the huge infastructure google has

0 0 0 98 0

Mohammed Daba @m_dabaAI

4 months ago

If an ASI generates content and the same ASI can tell it's AI-made, the generator isn't truly "super" enough to be indistinguishable. But if the assessor can't tell, then it's not "super" enough. in both cases, it makes me wonder if ASI is just endless arm race, never attainable.

0 0 0 19 0

Mohammed Daba @m_dabaAI

4 months ago

it is strange how LLMs are saturating benchmarks though still not working in real-world scenario, are they training on them? o3 discrpency result on ARC-agi for example. just hype and smoke without fire. Are they gonna staurate all benchmarks without reaching AGI?

0 0 0 29 0

Mohammed Daba @m_dabaAI

6 months ago

GPT-4.5 are ‘embarrassingly sub-optimal.’ That’s not a flaw—it’s a goldmine. cracking this unlock next era, maybe next innovation it’ll emerge from area we’re not expecting. Quantum computing? Diffusion? ...

Taelin @VictorTaelin

6 months ago

229 163 2K 369K 981

0 0 0 39 0

Mohammed Daba @m_dabaAI

6 months ago

After OpenAI's disappointing GPT-4.5 launch, it's clear they stopped chasing benchmarks. AGI/ASI hype is over when everyone can hit SOTA. OpenAI now banks on emotional intelligence while Anthropic codes. USA’s hope is Grok; if China drops another deepseek, the bubble could bursts

0 0 0 123 0

Download Image

Mohammed Daba @m_dabaAI

8 months ago

A crawler that converts any website (static or dynamic) into markdown files for documentation or RAG purposes. Just give it base url and the whole website will be converted as markdown github.com/dabaspark/craw…

0 0 0 39 0

Mohammed Daba @m_dabaAI

8 months ago

I got tired of llms not knowing newer api, also basic rag setups failing on technical docs. so I built something for myself: - crwal any docs - free embeddings (stella_en_400M_v5) - dirt cheap LLM (deepseek) - Free tier DB (Supabase) repo github.com/dabaspark/agen…

0 0 2 135 1

Vaibhav (VB) Srivastav @reach_vb

8 months ago

The DeepSeek Technical Report is out!! 🔥 Trained on 14.8 Trillion Tokens, outperforms all open-source models, comparable to GPT-4o and Claude-Sonnet-3.5 Key contributions: > Load Balancing Strategy: Introduced an auxiliary-loss-free approach to minimize performance…

7 45 402 76K 133

Download Image

Qwen @Alibaba_Qwen

9 months ago

🎄Happy holidays and we wish you enjoy this year. Before moving to 2025, Qwen has the last gift for you, which is QVQ! 🎉 This may be the first open-weight model for visual reasoning. It is called QVQ, where V stands for vision. It just reads an image and an instruction, starts…

62 270 2K 375K 448

Download Image

Akshay 🚀 @akshay_pachaar

9 months ago

Turn any GitHub repository into LLM-ready text! Simply replace "hub" with "ingest" in a GitHub URL and receive a prompt-friendly text ingest for LLMs. Gitingest is 100% open-source and provides: - Directory structure - A brief summary of the project - The entire content as…

55 341 3K 189K 4K

Download Video

Zhou Xian @zhou_xian_

9 months ago

Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics…