PLUS: Hollywood unites against AI copyright proposals
Good morning, AI enthusiasts. The AI capability curve just found its "Moore's Law" moment — with new research showing task completion abilities doubling every 7 months since 2019.
With systems tackling hour-long human tasks today and potentially month-long projects by 2030, is the world prepared for the automation tsunami quickly approaching?
-
AI capabilities following ‘Moore's Law’
-
Hollywood against AI copyright proposals
-
Improving non-reasoning AI responses
-
Nvidia’s open-source reasoning models
-
4 new AI tools & 4 job opportunities
AI RESEARCH
The Rundown: Researchers at METR just published new data showing that the length of tasks AI agents can complete autonomously has been doubling approximately every 7 months since 2019, revealing a "Moore's Law" for AI capabilities.
-
The study tracked human and AI performance across 170 software tasks ranging from 2-second decisions to 8-hour engineering challenges.
-
Top models like 3.7 Sonnet have a "time horizon" of 59 minutes — completing tasks that take skilled humans this long with at least 50% reliability.
-
Older models like GPT-4 can handle tasks requiring about 8-15 minutes of human time, while 2019 systems struggle with anything beyond a few seconds.
-
If the exponential trend continues, AI systems will be capable of completing month-long human-equivalent projects with reasonable reliability by 2030.
-
Moore's Law predicts that computing power doubles roughly every two years — explaining why devices get faster and cheaper over time.
Why it matters: The discovery of a predictable growth pattern in AI capabilities provides an important forecasting tool for the industry. Systems that can handle much longer (months-long tasks for humans) and more complex tasks independently will completely reshape how businesses across the world approach AI and automation.
TOGETHER WITH MONGODB
The Rundown: Step into MongoDB’s Generative AI Use Cases Repository to discover how MongoDB powers GenAI applications— from text and vector search to robust AI Retrieval methods.
-
Notebooks on real-world scenarios like advanced RAG techniques and AI agents
-
Integration with top LLM providers and frameworks
-
Pre-built datasets and embedding tools to accelerate your GenAI development
-
A central resource and community for developers using MongoDB
CELEBRITIES VS. AI
Image source: Ideogram / The Rundown
The Rundown: More than 400 Hollywood creatives signed an open letter urging the Trump administration to reject OpenAI and Google’s proposals to expand AI training on copyrighted works—arguing that it would let them "freely exploit" creative industries.
-
The letter is a direct response to OpenAI and Google's AI Action Plan submissions, which argued for expanded fair use protections for AI training.
-
OpenAI framed AI copyright exemptions as a "matter of national security," while Google said the current fair use framework already supports AI innovation.
-
Ben Stiller, Mark Ruffalo, Cate Blanchett, Paul McCartney, Taika Waititi, and Aubrey Plaza are among the high-profile creatives who have signed the letter.
-
They have emphasized that AI companies could simply "negotiate appropriate licenses with copyright holders — just as every other industry does."
Why it matters: Hollywood vs. AI represents a values collision — the tech industry’s "move fast and iterate" mindset vs. Hollywood's centuries-old IP frameworks. But with AI giants across the globe already ingesting the world’s data even without copyright protections, this fight, in reality, may be more symbolic than action-oriented.
AI TRAINING
The Rundown: In this tutorial, you will learn how to dramatically improve the intelligence of non-reasoning AI models by implementing a structured reasoning approach with XML tags—forcing the model to think step-by-step before answering.
-
Structure your prompt with XML tags like
and to separate the reasoning process from the final output. -
Provide specific context and task details, including examples.
-
Force step-by-step reasoning by explicitly instructing the model to “think” first, then answer.
-
Compare results with and without your reasoning framework to see the dramatic improvements in quality.
Pro tip: You can use this technique especially when asking AI to match writing styles or analyze complex information before generating content.
PRESENTED BY DAGSTER
The Rundown: Dagster consolidates your AI capabilities into one powerful orchestrator that developers love — helping reduce costs, eliminate complexity, and ensure reliable pipelines from prototype to production.
-
Consolidate all AI capabilities under one intuitive interface
-
Save 40%+ on infrastructure costs by optimizing AI workloads
-
Ship AI features 3x faster with standardized development practices
Schedule a demo to learn more about how Dagster can simplify your AI platform.
NVIDIA
The Rundown: Nvidia released its Llama Nemotron family of open-source reasoning models, designed to accelerate enterprise adoption of agentic AI capable of complex problem-solving and decision-making.
-
The new model family comes in three sizes: Nano (8B), Super (49B), and Ultra (249B) — each optimized for different deployment scenarios.
-
Early benchmarks show impressive performance, with the Super version outperforming both Llama 3.3 and DeepSeek V1 across STEM and tool testing.
-
The models feature a toggle that allows AI systems to switch between intensive reasoning and direct responses based on the task.
-
Post-training resulted in 20% better accuracy than base Llama models and 5x faster speed than rival open reasoners.
-
Nvidia is also releasing an "AI-Q Blueprint" framework in April to help businesses connect AI agents with their existing systems and data sources.
Why it matters: Nvidia’s reasoning models may be overshadowed by the insane amount of releases over the past 48 hours, but the chipmaking giant has seemingly built every block necessary to be a force across the entire AI stack — from the most advanced hardware to high-quality reasoning models ready for the agentic era.
-
Diamond - Graphite’s agentic AI-powered code review companion -
Canvas - Gemini’s new collaborative space for document editing and coding -
Stable Virtual Camera - Images into 3D videos with dynamic camera paths -
Hunyuan 3D 2.0 MV - Open model for high-quality 3D shape generations
Google AI and UC Berkeley researchers proposed "inference-time search" as a new AI scaling method, producing several answers in parallel and selecting the best option.
LG released EXAONE Deep, a reasoning AI that achieves comparable performance to models like DeepSeek V1 in math, science, and coding with just 32B parameters.
Muse released Muse S Athena, a headband wearable combining EEG and sensors to measure both brain activity and oxygen levels for AI-powered cognitive fitness training.
Nvidia and xAI are joining Microsoft, BlackRock, and MGX in the AI Infrastructure Partnership, aiming to raise $30B initially and potentially $100B for AI data centers.
xAI debuted its first image generation API featuring the ‘grok-2-image-1212’ model, allowing developers to create multiple JPG images per request at $0.07 each.
Microsoft is partnering with neuroscience AI startup Inait to develop brain-inspired AI that learns from real-world experiences rather than data patterns.
Join our next workshop this Friday at 4 PM EST to learn how to use AI to transform and enhance images for content creation and creative projects with Dr. Alvaro Cintas, The Rundown’s AI professor.
We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.
Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team
[ad_2]
Source link