P
Home

AI & Research

AI & Research category

12 articles

Tracking token usage across OpenAI, Anthropic, and Gemini: every streaming gotcha I hit
News4 min

Tracking token usage across OpenAI, Anthropic, and Gemini: every streaming gotcha I hit

Developers tracking LLM costs across OpenAI, Anthropic, and Gemini face hidden gotchas in streaming token reporting — from different cache conventions to split-event stream formats.

AI & Research
RAG Pipeline: The Uncle-Nephew Complete Learning Guide
News4 min

RAG Pipeline: The Uncle-Nephew Complete Learning Guide

Learn how Retrieval-Augmented Generation (RAG) pipelines work — from vector retrieval and prompt augmentation to answer generation — and how to implement them effectively in production.

AI & Research
AI Agents in Group Chat: Octo Solves Multi-Agent Coordination
News3 min

AI Agents in Group Chat: Octo Solves Multi-Agent Coordination

Mininglamp Technology's Octo open-source platform reimagines AI agent orchestration by making agents first-class participants in group chat conversations, solving the session island coordination crisis.

AI & Research
After spat with Chinese gov't, Meta cuts AI Manus off from its internal systems and is 'sunsetting' platform, report claims — Beijing-ordered breakup of $2 billion AI deal begins
News4 min

After spat with Chinese gov't, Meta cuts AI Manus off from its internal systems and is 'sunsetting' platform, report claims — Beijing-ordered breakup of $2 billion AI deal begins

Meta has locked Manus AI out of its systems and is sunsetting the agentic AI platform after China's NDRC ordered the $2 billion acquisition unwound. Founders are now racing to raise $1 billion for a buyback.

AI & Research
LUCID: Learning Embodiment-Agnostic Intent Models from Unstructured Human Videos for Scalable Dexterous Robot Skill Acquisition
News7 min

LUCID: Learning Embodiment-Agnostic Intent Models from Unstructured Human Videos for Scalable Dexterous Robot Skill Acquisition

Researchers from Carnegie Mellon University have introduced LUCID, a two-stage framework that enables robots to learn dexterous manipulation skills from unstructured human videos at internet scale. The system separates intent prediction (what should happen next in a scene) from embodiment-specific motor control, allowing the same intent model to work across different robot platforms — from dexterous hands to parallel-jaw grippers. Evaluated on five real-world tasks including stirring, wiping, binning, push-T, and cable routing, LUCID achieved zero-shot transfer to novel scenes and objects using only internet video as training data.

AI & Research
The Algorithm Is Not the Behavior: Learned Priors Override Look-Ahead in a Chess-Playing Neural Network
News5 min

The Algorithm Is Not the Behavior: Learned Priors Override Look-Ahead in a Chess-Playing Neural Network

Researchers at BIFOLD discovered a phenomenon called 'forgotten puzzles' in Leela Chess Zero, where the neural network correctly computes tactical solutions like checkmates in its intermediate layers but systematically overrides them in the final output due to learned safety priors. Using activation steering, they recovered 61.7% of these forgotten solutions, proving that safety biases can override correct algorithmic computations. The findings challenge fundamental assumptions in mechanistic interpretability and AI safety research.

AI & Research
Can AI Agents Synthesize Scientific Conclusions?
Featured
News4 min

Can AI Agents Synthesize Scientific Conclusions?

New research introduces SciConBench — a 9,110-question benchmark — finding even the best AI agent achieves only a 0.337 factual F1 score in clean-room scientific conclusion synthesis.

AI & Research
Accelerating discovery with the AI for Math Initiative
News4 min

Accelerating discovery with the AI for Math Initiative

Google DeepMind and Google.org launch the AI for Math Initiative with five top research institutions to accelerate mathematical discovery using AI systems including AlphaProof, AlphaEvolve, and Gemini Deep Think.

AI & Research
Gemini for Science: AI Experiments and Tools for a New Era of Discovery
News3 min

Gemini for Science: AI Experiments and Tools for a New Era of Discovery

Google DeepMind launches a collection of science tools and experiments to expand the scale and precision of scientific exploration using Gemini AI.

AI & Research
Microsoft and OpenAI broke up — now they're ready to fight
Featured
Opinion4 min

Microsoft and OpenAI broke up — now they're ready to fight

Microsoft and OpenAI have restructured their landmark partnership, ending Azure exclusivity and allowing OpenAI to offer models on AWS and Google Cloud. The two companies transition from close allies to direct competitors while maintaining revenue-sharing ties. The shift reshapes the enterprise AI landscape, giving customers multi-cloud access to frontier AI models.

AI & Research
Nous Research's NousCoder-14B is an open-source coding model in the Claude Code era
News4 min

Nous Research's NousCoder-14B is an open-source coding model in the Claude Code era

Nous Research released NousCoder-14B, an open-source coding model achieving 67.87% Pass@1 on LiveCodeBench v6 (+7.08% over base Qwen3-14B). Trained with DAPO (Dynamic Sampling Policy Optimization) on 48 Nvidia B200 GPUs over 4 days using 24,000 competitive programming problems. Fully open source under Apache 2.0 license, competing in the Claude Code and Codex CLI era.

AI & Research
Alphabet raises record-breaking $85B for Google's AI business
News4 min

Alphabet raises record-breaking $85B for Google's AI business

Alphabet shattered records with an $85 billion stock sale to fuel Google's AI ambitions. The oversubscribed offering raised $45 billion in its first tranche — beating the initial $40B target — with Berkshire Hathaway alone buying $10 billion. A second $40B tranche is planned next quarter. The raise signals enormous public market appetite for AI investment ahead of anticipated IPOs from Anthropic, SpaceX, and OpenAI.

AI & Research