Kyle Corbitt
Latest posts
Codex, File My Taxes. Make No Mistakes.
A Pocket Guide to Surviving the Robot Apocalypse
OpenPipe is Joining CoreWeave
RULER: Easy Mode for RL Rewards
Reward Hacking 101
ART·E: How We Built an Email Research Agent That Beats o3
ART Trainer: A New RL Trainer for Agents
One Right Answer or Many? A Useful Distinction for Evaluating and Fine-Tuning LLMs
Analyzing OpenAI's Reinforcement Fine-Tuning: Less Data, Better Results
Using Reinforcement Learning and $4.80 of GPU Time to Find the Best HN Post Ever (RLHF Part 1)
A Founder's Guide to AI Fine-Tuning
Introducing Direct Preference Optimization (DPO) Support on OpenPipe
Fine-tuning Best Practices Chapter 2: Models
Fine-tuning Best Practices Series Introduction and Chapter 1: Training Data
Announcing Llama 3.1 and GPT-4o Mini fine-tuning through OpenPipe!
OpenPipe Mixture of Agents: Outperform GPT-4 at 1/25th the Cost
The Ten Commandments of Fine-Tuning in Prod (a Mastering LLMs Conference Talk)
What we've learned in 3 days of Llama 3
We Raised $6.7M to Replace GPT-4 with Your Own Fine-Tuned Models
Mixtral Curious? Comparing Mistral 7B and Mixtral for fine-tuning
S-LoRA: Serving Thousands of Models From One GPU for Fun and Profit
Axis Improves Generation Quality and Lowers Costs With Fine Tuning
Product Updates December 2023
How we built "Mistral 7B Fine-Tune Optimized," the best 7B model for fine-tuning
Announcing Automatic Evals for Fine-tuned Models
Is AI the Next Crypto? Insights from 2M HN comments
Llama 2 vs Mistral: Believe the Hype
Fine-tune your own Llama 2 to replace GPT-3.5/4
From Prompts to Models
Private, Recoverable SSO with FixedID
In FixedID: Building a Better SSO I described a system for building a proof-of-personhood protocol with built-in account recovery. In that post, I mentioned the possibility of using the system for SSO while maintaining user privacy. This post will elaborate a specific system for achieving that.
Crypto Basic Income
We’re at a unique moment in history for universal basic income. Over just the last few years, the technical problem of implementing universal basic income has gone from completely impractical to almost trivially easy.