TLDR AI 2026-05-19

Are you asking users to put up with search timeouts, empty results, or irrelevant answers? This Algolia whitepaper lays out the full stack of architecture & data foundations for AI search.

Read it to learn how to:

• Combine lexical precision with semantic recall via hybrid retrieval
• Engineer p95 and p99 as product features
• Setup reranking and recommendations grounded in retrieved sources

Complete with code snippets, paste-ready RAG prompt contracts, and prod anti-patterns, this is your technical guide to scaling AI-powered search.

Download the whitepaper

Qwen3.7 Preview is now on Arena for Text and Vision. Qwen3.7 Max Preview ranks 13th overall in Text Arena, while Qwen3.7 Plus Preview ranks 16th overall in Vision Arena.

Cursor introduced Composer 2.5, an updated coding agent trained with targeted reinforcement learning, synthetic data, and new distributed training techniques.

Anthropic acquired developer tools startup Stainless, whose SDK automation platform was widely used by AI companies, including OpenAI, Google, and Cloudflare.

Qwen3.5-9B's political censorship is a small circuit that can be read and turned off. The factual knowledge is already in pretraining. The censorship behavior is layered on top of the facts. The model never loses the knowledge, it just learns to route around it.

LLM evaluation has shifted from static benchmarks to more dynamic, real-world agent systems. Effective evaluation now requires realistic harnesses to test agents over long time horizons in complex environments. This is crucial as agents increasingly adopt high-stakes roles, such as coding and medicine, necessitating rigorous performance measurement and outcome-oriented evaluation.

AI agents are everywhere, and so are the risks. Join AI security experts from Microsoft, Google, and Amazon at Zenity Labs' AI Agent Security Summit on May 27th in SF to see what's next. Want to test your knowledge? Register for free for Foundations of AI Security and earn your 🎓professional certification.

NVIDIA Cosmos Predict 2.5 generates videos from text, adapting for specific tasks like robot manipulation using LoRA/DoRA to inject trainable adapters, minimizing memory use. These methods offer efficient fine-tuning on a single GPU, preventing catastrophic forgetting while generating synthetic trajectories quickly. Fine-tuning with LoRA and DoRA significantly improves video quality, with LoRA more suited for tight memory conditions and DoRA preferred for addressing training instability.

HRM-Text is a 1B text generation model based on the HRM architecture. It can be trained with 130-600x less compute and 150-900x less data than foundation models, making foundation model pretraining accessible. The 0.6B parameter version of the model can be trained on 8 H100s on a single node in about 50 hours for around $800. The 1B parameter model can be trained on 16 H100s on two nodes in about 46 hours for around $1,472.

Language models (LMs) undergo unpredictable switches between parroting patterns and exhibiting adaptive intelligence during pre-training, a phenomenon termed "mode-hopping." This behavior cannot be corrected by standard optimization techniques and presents as a competition for model capacity, influenced by data from each training window. Researchers propose leveraging these dynamics to better select pre-training checkpoints, curate data for stable generalization, and evaluate metrics predicting LM behavior.

The first Nvidia Vera CPUs recently arrived at Anthropic, OpenAI, SpaceXAI, and Oracle. They were hand-delivered by Nvidia Vice President of Hyperscale and High-Performance Computing, Ian Buck. Vera features 88 custom Nvidia-designed Olympus cores, 1.2 TB/s of memory bandwidth, and 50% faster per-core performance. It is the host processor for Vera Rubin NVL72, which pairs via second-generation Nvidia NVLink-C2C to a pair of Rubin GPUs.

Elon Musk's lawsuit against Sam Altman and OpenAI has been dismissed. A jury has decided that Musk waited too long to file his lawsuit. Musk says he plans to appeal.

Production agents need durable execution, the ability to resume from where they left off without starting over. Join LangChain to learn how to make it work in real deployments.Save your seat

xAI launched "Skills" for Grok, allowing users to teach it functions once, which it remembers across interactions.

This post contains a pattern for building personal knowledge bases using LLMs.

Scheduled Tasks 2.0 enhances automation by allowing tasks to run with context, ensuring continuity in workflows across different projects and apps.

Skills in Lovable allow users to create reusable, markdown-based instructions to eliminate repetitive explanations.

TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. Learn more.