The Real Skill in Building With AI Is Writing the Brief. There is a job now where the whole task is describing to a machine what to build, and the person doing it has never written code. The craft did not vanish. It moved. ai-tools clarity context-window product-judgment taste There is a job now where …
The Real Skill in Building With AI Is Writing the Brief · Hanh D. Brown Read More »
Contents# Dealing With Inference Latency# 1. Implementing Model Quantization# 2. Utilizing Key-Value Caching# L3. everaging Speculative Decoding# 4. Transitioning to Continuous Batching# 5. Pruning and Distilling Your Models# 6. Deploying with Optimized Inference Engines# 7. Optimizing Context and Prompt Management# Stacking Optimizations in Practice # Dealing With Inference Latency As large language models (LLMs) move from research prototypes into production, engineering teams run into …
7 Approaches to Reduce Inference Latency in Your LLM Workflows Read More »
One Answer, Fifteen Models: How an AI Agent Works at Scale. Ask a production agent one question and a stopwatch starts. Two seconds. In that window it will run more models than most people think exist in the whole system. ai-agents reliability voice-ai agent-architecture outcome-pricing Ask a production agent one question and a stopwatch starts. …
Our newest music generation model, Lyria 3.5, delivers significant advancements across musicality, lyrics, and vocal quality, empowering you to craft richer tracks. We’re rolling it out today in Google Flow Music, where we want to help you create songs you love, with creative control. Here’s what’s new: Improved musicality: Create richer, more complex melodic structures …
Most of What the Agent Writes Should Be Thrown Away. A team built about a hundred versions of one loading spinner and shipped a dozen. The eighty in the bin were not waste. They were the method. prototyping iteration product-decisions engineering-management ai-tools About eighty versions of a single loading spinner went into the bin. The …
Most of What the Agent Writes Should Be Thrown Away · Hanh D. Brown Read More »
a cleaning company to request a quote for cleaning a couch. They asked about its size and material and requested a picture of it. They also needed my address because travel time varies significantly across the city and therefore affects the price. After receiving the quote for the couch, I asked about apartment cleaning. This …
I Replaced a 15-Minute Booking Process with a LangGraph AI Agent Read More »
Why Your Messy Codebase Is Now a Performance Problem. Tidiness used to be an argument about taste. Now the half finished migration nobody wanted to talk about is quietly setting the speed limit for every tool you bought. technical-debt codebase-hygiene ai-tools migrations engineering-management Watch a new hire on their first week. They open the codebase, …
Why Your Messy Codebase Is Now a Performance Problem · Hanh D. Brown Read More »
Contents# Introduction# Why the Sequential Pattern Doesn’t Work# Streaming Speech-to-Text# Turn Detection: Deciding When the User Is Actually Done# Streaming the Response Into Text-to-Speech# Handling Interruption Without Breaking State# Tool Calling Mid-Conversation# How the Components Actually Connect# Conclusion # Introduction Most people picture building a voice agent as stitching three things together: speech-to-text (STT), a large language model (LLM), and text-to-speech (TTS). Wire them …
How Much AI Autonomy Can Your Team Actually Supervise?. Every AI tool has a dial on it. Most of the argument about agents is really an argument about where that dial should sit, and almost nobody says so out loud. autonomy human-in-the-loop ai-products agent-ready-docs verification Every Artificial Intelligence (AI) tool you use has a dial …
How Much AI Autonomy Can Your Team Actually Supervise? · Hanh D. Brown Read More »
For robots to assist humans in everyday environments, accurate spatial reasoning is not enough. Robots must also think fast, timing their decisions and reasoning with the real-time speed of the physical world. That’s why today we’re launching Gemini Robotics ER 2, our most capable “embodied reasoning” model for robotics. Think of Gemini Robotics ER 2 …
