LLMs & Models
Large language models: GPT, Claude, Gemini, Mistral and open weights.
28 published articles
Remote Sensing AI
Smarter tokens just cut satellite AI costs by threefold without losing accuracy
Ai4earth's OlmoEarth v1.1 cuts compute costs by up to 3x over v1 for satellite image analysis, using a smarter token merging technique that maintains performance. The updated models enable cheaper planet-scale map refreshes for partner organizations.
2026-07-13
Lean 4
Leanstral 1.5 proves the old rules of AI pricing don't apply to math
Leanstral 1.5, a 6B active-parameter model, saturates miniF2F, solves 587 PutnamBench problems, and uncovers 5 previously unreported bugs in open-source repositories. At roughly $4 per problem, it undercuts Seed-Prover by 75x and Aleph Prover by 15x, challenging the assumption that formal verification requires massive compute budgets.
2026-07-12
Systems optimization
DSpark shows why fast AI inference is a scheduling problem, not a model trick
DeepSeek's DSpark paper reveals that naive speculative decoding degrades throughput under high concurrency. Its solution, confidence-scheduled verification, adapts block length per request and shifts the Pareto frontier of serving performance.
2026-07-12
Synthetic Data Strategy
Nvidia's data atlas shows why synthetic data matters more than model weights
Nvidia's Nemotron Post-Training v3 Prompt Atlas provides an interactive map of billions of synthetic data samples, highlighting how open synthetic data is the missing layer for building reliable AI agents. The company argues that agent behavior must be inspectable and that synthetic data, released openly, is the only way to preserve proprietary signals without exposing trade secrets.
2026-07-12
Long-context LLMs
Jet-Long's bifocal attention just killed the fixed-scaling trade-off for long-context LLMs
Jet-Long adapts RoPE rescaling dynamically per sequence length, using a local window and a long-range window merged via inclusion-exclusion. On Qwen3 models up to 128K context, it beats existing zero-shot baselines by over 2 percentage points on RULER and achieves the lowest perplexity on PG-19, all while adding less than 4% generation overhead.
2026-07-12
Open Source Infrastructure
Your AI search pipeline is broken. This open-source framework fixes the plumbing.
Teams building AI search infrastructure still spend too much time on plumbing. Search Toolkit unifies ingestion, retrieval, and evaluation into a single open-source framework, eliminating the weeks of integration work needed to stitch together separate tools. It's designed for enterprise use cases like RAG quality, domain-specific retrieval, and agentic search.
2026-07-11
Financial AI
The world's first AI-native credit card just rewired spending into compute
Moonshot AI issues an AI-native credit card with Agricultural Bank of China and American Express, linking membership tiers to card levels and offering token-based rewards. A look at how this embeds AI token economics into daily payments.
2026-07-11
Neuroscience
The neuroscience AI that finally explains what brain parts actually do
Generative causal testing (GCT) distills LLM-based brain-prediction models into concise verbal explanations, then uses an LLM to write stories that causally test those claims in fMRI. In experiments, GCT confirmed known selectivity, teased apart neighboring place-processing regions, and discovered new prefrontal micro-regions tuned to concepts like dialogue and measurements.
2026-07-11
Open Weights Analysis
Gemma 4 is not a chatbot, and that's the point
Google DeepMind's Gemma 4 is an open-weight model designed for self-hosting and customization, not consumer chat. This analysis compares it to ChatGPT, Claude, and Qwen-3.5 across licensing, privacy, and deployment flexibility, revealing why it matters for regulated industries and on-device AI.
2026-07-10
AI Inference Acceleration
JetSpec Breaks the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting
JetSpec trains a causal parallel draft head over fused hidden states from the target model, producing candidate trees aligned with autoregressive factorization. It consistently outperforms baselines across dense and MoE Qwen3 models on math, coding, and chat tasks.
2026-07-10
Artificial Intelligence
Nvidia's new audio model does five jobs at once and beats the specialists at their own game
Nvidia's Audex unifies audio understanding, generation, and text reasoning in a single model, matching or beating task-specific systems on speech and audio benchmarks without sacrificing text performance.
2026-07-09
Alibaba Cloud
Alibaba Cloud's EMR Serverless Spark now processes images and video in plain SQL, no Python needed
Alibaba Cloud’s EMR Serverless Spark now supports images and video frames directly in SQL, letting data engineers skip Python overhead. A case study on autonomous driving data preprocessing shows automated ETL pipelines powered by Qwen vision models replacing manual annotation.
2026-07-09