12-03-23ChatGPT After One Year
12-01-23Data Quality in LLMs
11-30-23Discord and AI GTM
11-24-23How AI Changes Workflows
11-22-23Strategies for the GPU-Poor
11-18-23The Model is Not the Product
11-17-23The AI-Neid
11-16-23Model Merge - (Frankenmerge)
11-15-23The Cost of Index Everything
11-13-23Copilot is an Incumbent Business Model
11-09-23AI Agents Today
11-08-23Norvig's Agent Definition
11-07-23The Context Length Observation
11-05-23Improving RAG: Strategies
11-03-23Lessons from llama.cpp
11-02-23Why Model Evaluation is Difficult
11-01-23Mechanical Turks
10-30-23What If OpenAI Builds This?
10-28-23Infrastructure as Code Will be Written by AI
10-26-23Between Images and Text, CLIP
10-24-23Tech Invariants
10-23-23Horizontal Tuning: Instruction, Chat, and What Else?
10-22-23Retrieval Augmented Generation
10-19-23Benefits of Small LLMs
10-18-23Can OpenAI Win Consumer and Enterprise?
10-16-23Revision: Generative text-to-UI
10-14-23An Intelligent Wikipedia
10-13-23The Half-Life of the AI Stack
10-09-23Moravec's Paradox
10-02-23Generative Interfaces
09-30-23Compression / Learning Duality
09-29-23Is AI a Platform Shift?
09-28-23Passkeys, Crypto, and Signing AI Content
09-27-23Is Data Still a Moat?
09-26-23Multi-Modal AI is a UX Problem
09-20-23AI Biographers
09-19-23Customized End User Software (with AI)
09-16-23The Age-old Resistance to Generated Code
09-14-23Undetectable AI
09-11-23Fine-tuning Stable Diffusion XL with Personal Photos
09-09-23Beyond Prompt Engineering
09-05-23Type Constraints for LLM Output
09-01-23Capital Intense AI Bets
08-30-23Llama 2 in the Browser
08-29-23The Contrarian Strategy of OpenAI
08-27-23AI and Text-First Interfaces
08-25-23The Free Lunch of Model Distillation
08-21-23A Model API Gateway for 20+ LLMs
08-16-23What is a Prompt Engineer?
08-13-23My Everyday LLM Uses
08-11-23Llama/Unix
08-10-23Deterministic, Structured LLM Output
08-08-23A Fine-Tuning Marketplace
08-01-23Automatic and Universal Adversarial Prompts
07-20-23Robots.txt for LLMs
07-19-23Why Did Meta Open-Source Llama 2?
07-15-23Scale to Zero for AI Workloads
07-07-23The Anti-AI Movement
07-04-23Where AI Fits in Engineering Organizations
06-30-23Personal Lessons From LLMs
06-29-23Overcoming LLM Hallucinations
06-25-23Model Evaluation is (Still) An Art
06-22-23No Feature is Free (Especially AI Ones)
06-21-23Mixture of Experts: Is GPT-4 Just Eight Smaller Models?
06-20-23The LLaMA Ecosystem
06-17-23The Low-Background Steel of AI
06-16-23Why Does Every AI Cycle Start With Chat?
06-15-23A Token Efficient Language for LLMs
06-12-23It’s Too Early To Call Winners in AI
06-09-23The Myth of the AI Infrastructure Phase
06-08-23LLMs For Software Portability
06-07-23ChatGPT Plugins Don't Have PMF
06-06-23Levels of Autonomy in AI Agents
06-05-23The Problem with Tokenization in LLMs
06-04-23What Diffusion Models Can Teach Us About LLMs
06-03-23Sequence and Version Control Models
05-31-23Faster Horses: AI Products That Companies Think They Want
05-29-23AI Means More Developers
05-27-23Prompt Engineering is Configuration Engineering
05-26-23SEO Inside AI
05-24-23The ChatGPT Plugin Specification
05-21-23The New Tax on Engineering Against the Grain
05-19-23On Regulating AI
05-18-23On Device AI?
05-17-23A List of Leaked System Prompts
05-16-23Intercloud Brokers
05-15-23React LLM: Run Models in the Browser with Headless Components
05-14-23Context-Free Grammar Parsing with LLMs
05-12-23StackOverflow/ChatGPT
05-11-23Self-hosted Compilers and Bootstrapped AI
05-10-23Unix Philosophy for AI
05-08-23The New AI Moats
05-05-23ReLLM: Exact Structure for Large Language Model Completions
05-04-23llm.ts
04-30-23Implementing LLMs in the Browser
04-26-23Probabilistic Data Structures and LLMs
04-21-23Autonomous LLM Agents Are At Least 10 Years Out
04-19-23Sandbox Your Prompts
04-17-23Jevons Paradox and LLMs
04-14-23Synthetic Data From Compilers
04-13-23Foundational Models Are Not Enough
04-10-23A List of 1 Billion+ Parameter LLMs
04-09-23No GPUs before Product-Market Fit
04-07-23Buyers in the Foundational Model Stack
04-05-23A High-level LLMOps Architecture
04-03-23The Automation Frontier
04-02-23Why Open-Source a Model?
03-30-23The AI Partnership Race
03-29-23A Hacker's Guide to LLM Optimization
03-27-23Code, not Chat, in Generative AI
03-26-23Distributed Systems and AI
03-23-23Are Incumbents Accruing All The AI Value?
03-22-23Model Arbitrage
03-21-23Modeling Context Length vs. Information Retrieval Cost in LLMs
03-17-23Foundational Models Are Commodities
03-16-23On OpenAI's Kubernetes Cluster
03-15-23Choosing the Right Model
03-13-23On Prompt Injection
03-12-23Local LLaMA on a Mac M1
03-11-23Automatic1111 and AI Aggregators
03-09-23Chain of Thought Paradigms in LLMs
03-05-23Will The AI Stack Be Open Or Closed?
03-02-23ChatML and the ChatGPT API
02-25-23Commoditization of Large Language Models: Part 3
02-19-23Why Python Won't Be the Language of LLMs
02-16-23Why ChatGPT Needs AuthZ
02-12-23LLM Ops, Part 1
02-10-23Multi-Model vs. Multi-Cloud
02-05-23Composable Models
01-28-23Overview of GPT-as-a-Backend
01-23-23Prompt Engineering Shouldn't Exist
01-22-23GPT Lineage
01-14-23Garbage In, Garbage out?
01-12-23Minix and nanoGPT
01-10-23Lessons from the Last AI Cycle
01-08-23Fine-Tuning an OCR Model
01-02-23A New ML Stack
12-30-22Local AI: Part 2
12-29-22Local AI: Part 1
12-26-22Turing Social: Twitter, For Bots
12-21-22ML Developer Experience
12-15-22AI-driven Interfaces
12-14-22LAION, The Pile, and more datasets
12-12-22Lessons from Lensa
12-08-22Spam Filtering AI Content
12-06-22Stack Overflow Bans ChatGPT
12-05-22Will LLMs Disrupt Google Search?
12-03-22A Conversation with ChatGPT
11-27-22Human-in-the-Loop and Other AI Mistakes
11-19-22Generative AI Value Chain
11-13-22LLMs for Code
10-21-22AI Will Write Most Code
10-19-22AI Scaling Laws
09-12-22TensorFlow vs. PyTorch
07-25-22Defensible Machine Learning Model Naming
06-22-22How to Use GitHub Copilot Effectively
11-09-21Open-sourced GPT-J
08-17-21One Month of Using GitHub Copilot
07-10-21GitHub Copilot
06-26-21Machine Learning Operations (MLOps), Convergent or Divergent?
03-14-21ScapeNet: Real-time object detection in RuneScape