Struggling with AI that can't handle long, multi-page documents? Researchers from HUST and Xiaomi introduce Doc-V*, an OCR-free agent that starts with a thumbnail overview, then actively retrieves and reads only the relevant pages, storing evidence in a working memory. It
MACHINE LEARNING
-
Technical Requirements for Deploying Long-Running AI Agents
By
–
What does it actually take to deploy long-running agents and the runtime capabilities that make it possible? Join us for an evening in New York with
– Robert Xu, Deployed Engineering @LangChain – Austin Berke, Lead AI Product Engineer @harmonic_ai RSVP: Robert Xu, Deployed -
User switches from Grok 4.1 Fast to Grok 4.3 without reasoning
By
–
I just use whatever is available, yday that was Grok 4.1 Fast Non Reasoning, now I try Grok 4.3 via grok-latest with reasoning set to none
-
Anthropics co-founder predicts autonomous self-improving AI by 2028
By
–
Anthropics co-founder Jack Clark:
— Chubby♨️ (@kimmonismus) 8 mai 2026
"My prediction is that by the end of 2028, it is more likely than not that we will have an AI system where you could say to it:
"Make a better version of yourself."
And it would simply go off and do that completely autonomously."
Its coming. pic.twitter.com/OEM7zwcQzcAnthropics co-founder Jack Clark: "My prediction is that by the end of 2028, it is more likely than not that we will have an AI system where you could say to it: "Make a better version of yourself." And it would simply go off and do that completely autonomously." Its coming.
-
RAG is old way; future AI memory is compilation, not retrieval
By
–
RAG is already becoming the “old way” The future of AI memory is not retrieval.
It’s compilation. Here’s the shift in one sentence: From searching information To structuring knowledge The new model? LLM Wiki Instead of: Chunking documents Running similarity -
Secretary MEITY calls for indigenous AI at hackathon
By
–
In his keynote address at the AB PM-JAY Auto-Adjudication Hackathon Showcase 2026, @SecretaryMEITY made the case for building AI that is rooted in India's linguistic and cultural context — and affirmed the IndiaAI Mission's commitment to supporting indigenous foundation model
-
Faster Inference for Gemma 4 on LLaMA.cpp with Multi-Token Prediction
By
–
🚨 STOP WHAT YOU ARE DOING AND LOOK AT THESE BENCHMARKS@atomic_chat_hq just unlocked 1.5x faster inference for Gemma 4 on LLaMA.cpp using Multi-Token Prediction.
— Charly Wargnier (@DataChaz) 8 mai 2026
138 tokens per second on a local 26B model is pure sorcery 👀
Get the code and GGUFs below ↓ https://t.co/o4aF64B5eeSTOP WHAT YOU ARE DOING AND LOOK AT THESE BENCHMARKS @atomic_chat_hq just unlocked 1.5x faster inference for Gemma 4 on LLaMA.cpp using Multi-Token Prediction. 138 tokens per second on a local 26B model is pure sorcery Get the code and GGUFs below ↓
-
GPT 5.5 Impresses in Deep Learning Research
By
–
After another week of intensive work with GPT 5.5, I’m once again confirming my initial conclusions: it’s an impressive model. For deep learning auto-research tasks, the improvement over versions 5.2 and 5.4 is highly noticeable! It’s not just about execution and implementation—
-
AI compute crunch and its impact on chatbots
By
–
What is the #AI compute crunch—and how will it affect #Chatbots?
by @denibechard @sciam Learn more: https://
bit.ly/4ukDNhE #ArtificialIntelligence #MachineLearning #ML #MI -
Secretary MEITY keynote at AB PM-JAY AI Hackathon 2026
By
–
Shri S. Krishnan, @SecretaryMEITY
, delivers the Keynote Address at the AB PM-JAY Auto-Adjudication Hackathon Showcase 2026, a two-day national platform bringing together policymakers, technologists & innovators to advance AI-driven health claims management at @iiscbangalore .