Topic
All digests tagged Natural Language Processing
Suggested route
Move from earlier context through the available material to the newest signals. This route is a reading aid, not a difficulty ranking.
Louis-François Bouchard introduces 'slop' in AI writing and details the creation of Tone-Bench, a benchmark using 10 scripts, 5 drafts, and 4 LLM judges to measure creative writing quality.
Read article
Discover how Gemma 4 enables powerful, private AI builds that run entirely on-device. Learn about deploying open models for robotics, remote data analysis, and offline coding.
Read article
Deep dive into evaluating OCR and AI agents on complex documents. Learn how to perform root cause analysis when an agent fails, tracing the error back to upstream data issues like OCR inaccuracies or missing watermarks.
Read article
Spotify details how it turned its recommendation system LLM-native using the Large Taste Model. Learn about the NEO four-stage training recipe and Semantic IDs.
Read article
The talk details building the document context layer for AI agents, arguing that RAG requires a three-layer platform: parsing, semantic storage, and repeatable workflows. It addresses the complexity of parsing PDFs and introduces the ParseBench benchmark.
Read article
Designing AI agents for knowledge work requires mimicking human orchestration. Agents must break down complex, ambiguous problems using specialized sub-agents and multiple search primitives to overcome performance ceilings.
Read article
Learn how to build real-time voice agents with Frontier Intelligence. We detail a cascaded architecture using speculative transcription, background tool calling, and prefix caching to eliminate latency.
Read article
Building a voice-first AI companion requires designing for 'conversational volatility.' The team uses tiered model routing based on emotional stakes and rebuilds context per turn, treating memory as a retrieval system.
Read article
Learn how to evaluate AI agents at scale using an LLM-as-a-judge in LangSmith. Automate quality checks for user frustration and manage costs by adjusting the sampling rate.
Read article
Learn how to build low-latency dictation features using the Sync API, connection warming, and LLM cleanup passes. The process is optimized to deliver results in under one second.
Read article
This demonstration shows how an agentic layer, built on AWS's open-source strands framework, allows a rover to perform complex tasks in natural language without explicit training.
Read article
Harmonic quadrupled user retention by shifting from brittle query graphs to Deep Agents. Learn how a model-plus-tools loop and file system abstractions solve context overload.
Read article
Building production voice agents requires a modular, cascading architecture (ASR $\rightarrow$ LLM $\rightarrow$ TTS). Key focus areas include managing latency within a 1-1.5 second budget and implementing robust fallback systems across all components.
Read article
Video AI systems lack memory; this talk details building a durable 'memory layer' by treating video as a spatial temporal volume. The solution uses a Context Graph to enable deep reasoning over vast corpora.
Read article
Universal 3.5 Pro enhances STT with contextual awareness, allowing developers to pass domain-specific prompts and maintain conversation context for highly accurate transcription in voice agents.
Read article