All articles

The complete WAKIB archive: every published story, across every section, newest first.

LatestLLMs

Alibaba Previews 2.4-Trillion-Parameter Qwen3.8-Max Days After Moonshot's Kimi K3

Alibaba previewed Qwen3.8-Max, claimed at 2.4 trillion parameters, at WAIC in Shanghai on July 19, 2026, days after Moonshot AI's 2.8T-parameter Kimi K3.

AI AgentsJuly 22, 2026

CNBC AI Forum 2026 Set for October 1

CNBC AI Forum 2026 is scheduled for October 1, 2026, running 1:00 PM to 11:00 PM ET

AI AgentsJuly 22, 2026

A Water-Filling Rule for Multi-Agent AI: New Paper Tightens the Math Behind Learned Communication Graphs

HIBCG (Heterogeneous Information-Bottleneck Coordination Graphs) is a new method for learning sparse communication graphs in cooperative multi-agent reinforcement learning, posted to arXiv on May 17, 2026 and revised July 10, 2026.

ResearchJuly 22, 2026

New Multi-Agent Framework ARCANA Tackles ARC-AGI-2 with Symbolic Reasoning and a Shared Blackboard

ARCANA is a new multi-agent framework published July 10, 2026, built to solve ARC-AGI-2 tasks under strict time and hardware limits

LLMsJuly 22, 2026

New Framework Tells Researchers Exactly Where to Spend Their Human Survey Budget When LLMs Do the Rest

A new arXiv paper introduces 'rectification difficulty,' a metric quantifying how hard it is to correct LLM errors on specific survey questions using human labels.

HardwareJuly 22, 2026

Behind the Hype: The Guardian's On-the-Ground Investigation Into AI's Datacentre Boom

The Guardian's investigation found a proposed £8.2bn AI datacentre complex in rural Scotland misrepresented plans to run entirely on on-site renewable energy.

ResearchJuly 22, 2026

A Shoestring Budget and a Printer-Sized Quantum Computer Just Improved Drug Discovery AI

DTU researchers paired a generative AI model with a compact quantum computer from ORCA Computing to design peptides that bind therapeutic protein targets

LLMsJuly 22, 2026

New Benchmark Exposes a Blind Spot in AI Doctors: They Still Can't Read the Room (or the X-Ray)

MedRealMM is a new benchmark built from 5,620 real multimodal patient-doctor cases across 64 clinical departments at a Chinese internet hospital

ResearchJuly 22, 2026

ALICE: A Single Foundation Model Absorbs Eight Pathology AI Systems Into One Backbone

ALICE is a pathology foundation model built by distilling eight teacher models (vision-only, vision-language, and slide-level) into one unified backbone via multi-stage agglomerative distillation

AI AgentsJuly 22, 2026

FoneClaw Lets You Run Your Android Phone by Just Talking to It

FoneClaw is a voice-controlled application for Android that supports more than 120 spoken commands

ResearchJuly 22, 2026

A Unified Geometry for Batch Normalization: LieBN Extends Deep Learning to Nine Curved Data Spaces

LieBN generalizes batch normalization to nine Lie group geometries, including SPD manifolds, rotation matrices, and full-rank correlation matrices.

ResearchJuly 22, 2026

No Retraining Required: TCLA Sharpens Medical Vision-Language Models on the Fly

TCLA is a training-free few-shot adaptation method for Medical Vision-Language Models, detailed in an arXiv paper submitted July 10, 2026

Computer VisionJuly 22, 2026

CtrlVTON Turns Virtual Try-On Into a Precision Editing Task, Not a Guessing Game

CtrlVTON reframes virtual try-on as a controllable image-editing task, using segmentation masks for pixel-level control over garment size, style, and placement.

ResearchJuly 22, 2026

When Medical AI Can't See Who It's Failing, CAPRA Fills the Blind Spot

CAPRA is a new calibrated proxy-axis framework for detecting hidden subgroup performance gaps in medical imaging AI when demographic metadata is missing.

AI Video GenerationJuly 22, 2026

PixVerse Tops $2 Billion Valuation as Series C Swells to $439 Million

PixVerse's Series C extension brings total funding to $439 million, pushing its valuation above $2 billion

AI AgentsJuly 22, 2026

Helsing Raises $1.8B, Hits $18B Valuation, and Plants Its Flag in West Virginia

Helsing closed a $1.8 billion Series E round, valuing the Munich-based defense AI firm at $18 billion.

ResearchJuly 22, 2026

QAgent Tackles One of Quantum Computing's Quiet Bottlenecks: Writing Correct QASM Code

QAgent, detailed in an arXiv paper from August 26, 2025, is billed as the first autonomous multi-agent framework for end-to-end OpenQASM (quantum assembly) code generation.

AI AgentsJuly 22, 2026

An AI Curriculum Engine Cuts 9-1-1 Dispatcher Training Reviews to Seconds, Not Minutes

PACE, a Personalized Adaptive Curriculum Engine built with Metro Nashville's emergency communications department, was accepted at IJCAI-ECAI 2026 after arXiv posting in March 2026.

All articles