Search the archive

CODE WITH CLAUDE DEVELOPER CONFERENCE

Anthropic has announced its first developer event focused on hands-on sessions and best practices for building with Claude - it is set to take place in May in San Francisco.

THE LLM COURSE FROM HUGGING FACE

Hugging Face has revamped its popular NLP course into a broader LLM curriculum with expanded chapters on fine-tuning, reasoning models, and modern AI agent workflows.

DEVIN, THE VIRAL CODING AI AGENT, GETS A NEW PAY-AS-YOU-GO PLAN

Cognition launched a $20 entry-level plan for its AI programming tool, Devin, which has been improved to handle complex coding tasks better.

ANTHROPIC'S LLM FOR EDUCATION

Anthropic has introduced Claude for Education, which has features like Learning Mode to promote critical thinking and broad university access through partnerships with major institutions and education platforms.

CLAUDE AVAILABLE FOR U.S. GOVERNMENT USE

Claude has achieved FedRAMP High and IL-2 compliance through Google Cloud's Vertex AI, making it usable by federal agencies and defense organizations with strict security needs.

META AND UFC PARTNER TO ENHANCE FAN ENGAGEMENT WITH AI AND VR

Meta Platforms and the Ultimate Fighting Championship (UFC) have entered a multiyear partnership to integrate Meta's AI and VR technologies into UFC events. This collaboration aims to provide fans with immersive experiences using platforms like Meta Quest and AI-powered glasses. Meta's branding will feature in the UFC's Octagon during events and Threads will serve as the official social media partner. Financial terms were not disclosed.

PAPER BENCH

OpenAI has released a new benchmark for academic paper creation. It requires full replication of certain papers, including understanding their experiments, results, and creation of novel ideas.

GEOMETRY CRAFTER

A powerful model that uses video diffusion as a prior for consistent geometry estimation over time. It runs about 1.5 FPS for full point cloud estimation and can also do proper camera pose estimation.

ENHANCED CELL SEGMENTATION

CellVTA improves vision transformer-based models for cell instance segmentation by injecting high-resolution spatial features through a CNN-based adapter, achieving state-of-the-art performance on multiple datasets.

DSO: ALIGNING 3D GENERATORS WITH SIMULATION FEEDBACK FOR PHYSICAL SOUNDNESS

Most 3D synthesized data is generated for aesthetic quality. However, that means that when they are put into physics environments, they cannot stand up or be self supported. This work fine-tunes a model slightly to make it more physically plausible.

MEDICAL REASONING DATASET

A large-scale medical reasoning dataset built to support faithful and explainable problem-solving in LLMs that aims to push forward research in medical AI.

DEEPMIND'S STUDY FOR KERNEL FUZZING

Snowplow is a kernel fuzzer with a learned white-box mutator that improves test mutation efficiency, significantly increasing code coverage and bug discovery in Linux kernel fuzzing.

SATYA NADELLA: DEEPSEEK IS THE NEW BAR FOR MICROSOFT'S AI SUCCESS

Microsoft CEO Satya Nadella emphasized the importance of turning AI research into successful products following an $80 billion AI investment. The company is focused on improving its AI offerings, like Copilot and Muse, while ensuring investments align with sustainability goals. Microsoft aims to be carbon-negative by 2030 despite challenges posed by increased AI workloads.

ALPHABET'S AI DRUG DISCOVERY PLATFORM ISOMORPHIC LABS RAISES $600M FROM THRIVE

Isomorphic Labs, spun out of DeepMind, raised $600 million from Thrive Capital to further develop its AI drug design platform. The funding will also support hiring top researchers and advancing discovered drugs to clinical trials. Notably, Isomorphic Labs has partnerships with Eli Lilly and Novartis for its AI model.

THE HOTTEST AI MODELS, WHAT THEY DO, AND HOW TO USE THEM

This article outlines the top AI models released since 2024, highlighting their uses and strengths. Notable models include OpenAI's GPT-4.5 Orion, which has strong world knowledge; Google's Gemini 2.5 Pro, focused on coding; and Cohere's Aya Vision, which excels in image tasks. The overview aims to simplify navigating the rapidly evolving AI landscape.

DEEPSITE OPEN SOURCE CANVAS (HUGGING FACE HUB)

DeepSite is a DeepSeek powered open source canvas for “vibe coding” that updates apps in real time while the system writes the code.

NOTEBOOKLM ADDS WEB-BASED SOURCE DISCOVERY

Google's NotebookLM now includes a Discover feature that pulls in curated web sources based on a user-defined topic, streamlining research and information gathering.

AI-POWERED CONVERSATIONAL VIDEOS

Tavus uses Llama 3.3 to power realistic AI-generated video conversations by combining visual question-answering and multi-image reasoning for lifelike digital interactions.

price is changing on April 10th @ 11:59p.

OPENAI ANNOUNCES $40B IN NEW FUNDING

OpenAI has raised $40B at a $300B valuation to support AI research, infrastructure scaling, and its growing user base and partnered with SoftBank to expand AGI development.

GEMINI ROBOTICS FROM GOOGLE DEEPMIND

Google DeepMind has revealed its Gemini Robotics models, which extend Gemini 2.0 with fine-tuning for physical actions.

OPENAI LAUNCHES OPENAI ACADEMY, A FREE AI LEARNING PLATFORM FOR EVERYONE

OpenAI has launched OpenAI Academy, a free platform that offers a range of AI courses, from beginner lessons to advanced topics like AI safety and governance. Catering to various audiences, the platform aims to broaden access to AI education and encourage critical thinking about AI's societal impacts. Initial reactions highlight the platform's accessibility and comprehensive approach to demystifying AI for a global audience.

VIDEO MOTION SEGMENTATION

Following the recent trend in tracking systems, this work adds dense pixel tracking to improve long term segmentation with Dino and SAM2.

OPEN HANDS CODING MODEL

A strong 32B model trained with RL on top of Qwen that surpasses many substantially larger models on agentic coding tasks.

OPEN-REASONER-ZERO: AN OPEN SOURCE APPROACH TO SCALING UP REINFORCEMENT LEARNING ON THE BASE MODEL

There is a huge open question in RL reasoning efforts whether or not a strong enough base model is required and necessary for emergent reasoning. This work explored many cases of scaling RL on a base mode. It found that strong base models are very helpful for reasoning convergence.

EASI3R: ESTIMATING DISENTANGLED MOTION FROM DUST3R WITHOUT TRAINING

Easi3R is a 3D vision system that can better estimate 3D scenes with high motion. It improves substantially over previous methods in full scene reconstruction. This was achieved by masking moving objects and learning those and the background separately.

BENCHMARK FOR RL-BASED VIDEO UNDERSTANDING

SEED-Bench-R1 is a benchmark that evaluates post-training methods like RL and SFT for multimodal LLMs in complex video-based tasks, showing RL's strengths in perception and data efficiency but challenges in logical coherence.

FLOW PREDICTION FOR AUTONOMOUS DRIVING

UniOcc is a unified framework for forecasting and flow prediction in driving scenarios built for multi-dataset training and cross-domain evaluation in both real and synthetic environments.

NEXTHOP AI LOCKS UP $110M LED BY LIGHTSPEED

Nexthop AI secured $110 million led by Lightspeed Venture Partners to enhance networking solutions for hyperscalers, focusing on cost and power efficiency. The round also saw participation from Kleiner Perkins, WestBridge Capital, Battery Ventures, and Emergent Ventures. CEO Anshul Sadana emphasized Nexthop AI's role in innovating alongside cloud companies.

ALIBABA HEAD WARNS AI INDUSTRY IS SHOWING SIGNS OF BUBBLE

Alibaba chairman Joe Tsai warns of a potential AI bubble due to massive investments in data centers without clear demand. Despite $52 billion committed to AI development, concerns are rising about overinvestment. Recent disruptions like Chinese startup DeepSeek highlight investor fears of overpaying in the AI sector.

INSIGHTS FROM CLAUDE 3.7 SONNET

Anthropic has introduced the Economic Index, which uses Claude Sonnet 3.7 to assess AI's influence on different job sectors. The model evaluates how AI affects productivity, automation risks, and labor market trends. This initiative aims to provide data-driven insights for policymakers and businesses navigating AI-driven economic shifts.

AMAZON'S ALEXA FUND IS NOW BACKING AI STARTUPS

Amazon's Alexa Fund is expanding its investment focus to include AI startups, investing in companies like NinjaTech AI, Hedra, Ario, and HeyBoss.

AI-POWERED THERAPY SHOWS SHOCKING RESULTS IN MENTAL HEALTH STUDY

A Dartmouth study found that AI therapy chatbot Therabot significantly reduced symptoms in participants with depression, anxiety, and eating disorders.

DATABRICKS HAS A TRICK THAT LETS AI MODELS IMPROVE THEMSELVES

Databricks has developed Test-time Adaptive Optimization (TAO), a method that leverages reinforcement learning and synthetic data to improve AI models without the need for clean labeled data.

Read the full report for more insights and takeaways.

AMAZON NOVA ACT

Amazon has introduced Nova Act, an AI model that allows agents to perform tasks within web browsers. A research preview SDK is available. The SDK helps developers create agents to execute complex, multi-step tasks by breaking them down into atomic commands and enabling browser manipulation for improved reliability. Nova Act aims to expand agent capabilities beyond simple tasks covered by APIs, enhancing business productivity and task automation.

RUNWAY'S NEW VIDEO MODEL

Runway has released its next-generation video model, which excels at prompt adherence and cinematic motion generation.

OPENAI TO RELEASE AN OPEN WEIGHT MODEL

OpenAI is soliciting feedback for an open weight model that has reasoning.

VIDEO GENERATION FAITHFULNESS BENCHMARK

A benchmark to determine how faithful video generation is to the prompt given. It also proposes methods to encourage model generation quality with respect to the user's input prompt.

UNLEARNING SENSITIVE CONTENT FROM LLMS

This paper proposes a model merging technique to selectively forget sensitive content in LLMs while preserving general knowledge.

OPTIMAL STEPSIZE IN DIFFUSION MODELS

Optimal Stepsize for Diffusion Sampling (OSS) optimizes sampling in diffusion models by learning efficient stepsize schedules through dynamic programming. It achieves 10x speedup while preserving nearly full generation quality.

🔹Limited capacity. Learn how to get in early →

SAMWISE VIDEO SEGMENTATION

This work gives SAM 2 open vocabulary segmentation and more precise semantic tracking over long videos.

ORPHEUS

Orpheus is a text-to-speech system. It is easy to install and runs without a GPU, similar to Llama cpp.

VIDEO-R1

Video-R1 introduces a rule-based RL approach for video reasoning using a temporal variant of GRPO and new datasets. It can be trained efficiently on 4 H20 or 5 A100 GPUs.

FAST TEXT-TO-3D

Using Progressive Rendering Distillation to train 3D generators from text prompts without ground-truth meshes generates high-quality 3D meshes in 1.2 seconds and outperforms prior methods.