Yoav Farhi
@yoavf.bsky.social
KL Divergance? Gradient Norm? Clip Ratio? I keep forgetting what these ML metrics mean and what to look for in my training runs.
Since I couldn't find a central resource with simple explantations, I created one.
Since I couldn't find a central resource with simple explantations, I created one.
November 26, 2025 at 6:41 AM
KL Divergance? Gradient Norm? Clip Ratio? I keep forgetting what these ML metrics mean and what to look for in my training runs.
Since I couldn't find a central resource with simple explantations, I created one.
Since I couldn't find a central resource with simple explantations, I created one.
Can an AI model teach itself to draw?
I was experimenting with reinforcement learning on small(ish) models, and had a fun idea: If a small vision-language model can generate SVG illustrations, can it also "look" at the illustrations and rate their quality?
I was experimenting with reinforcement learning on small(ish) models, and had a fun idea: If a small vision-language model can generate SVG illustrations, can it also "look" at the illustrations and rate their quality?
November 20, 2025 at 3:09 PM
Can an AI model teach itself to draw?
I was experimenting with reinforcement learning on small(ish) models, and had a fun idea: If a small vision-language model can generate SVG illustrations, can it also "look" at the illustrations and rate their quality?
I was experimenting with reinforcement learning on small(ish) models, and had a fun idea: If a small vision-language model can generate SVG illustrations, can it also "look" at the illustrations and rate their quality?
Six years off stage
I realized today it's been six years since my last conference talk at WordCamp Kathmandu 2019. I used to make a point of speaking at least once a year – it took a lot of work to prepare, but the energy in the room and the conversations afterward always made it worthwhile. Covid…
I realized today it's been six years since my last conference talk at WordCamp Kathmandu 2019. I used to make a point of speaking at least once a year – it took a lot of work to prepare, but the energy in the room and the conversations afterward always made it worthwhile. Covid…
Six years off stage
I realized today it's been six years since my last conference talk at WordCamp Kathmandu 2019. I used to make a point of speaking at least once a year – it took a lot of work to prepare, but the energy in the room and the conversations afterward always made it worthwhile. Covid interrupted that streak and I somehow never picked it back up. I really do miss it.
yoav.blog
July 2, 2025 at 6:33 PM
Six years off stage
I realized today it's been six years since my last conference talk at WordCamp Kathmandu 2019. I used to make a point of speaking at least once a year – it took a lot of work to prepare, but the energy in the room and the conversations afterward always made it worthwhile. Covid…
I realized today it's been six years since my last conference talk at WordCamp Kathmandu 2019. I used to make a point of speaking at least once a year – it took a lot of work to prepare, but the energy in the room and the conversations afterward always made it worthwhile. Covid…
Comparing AI coding agents with a real task
I threw an issue from a side project at 6 AI tools: Claude Code, OpenAI Codex (the CLI version), Factory.ai, Devin, Jules, and All Hands. The task was a combined bug fix + feature…
https://yoav.blog/2025/06/10/comparing-ai-coding-agents-with-a-real-task/
I threw an issue from a side project at 6 AI tools: Claude Code, OpenAI Codex (the CLI version), Factory.ai, Devin, Jules, and All Hands. The task was a combined bug fix + feature…
https://yoav.blog/2025/06/10/comparing-ai-coding-agents-with-a-real-task/
June 10, 2025 at 8:30 PM
Comparing AI coding agents with a real task
I threw an issue from a side project at 6 AI tools: Claude Code, OpenAI Codex (the CLI version), Factory.ai, Devin, Jules, and All Hands. The task was a combined bug fix + feature…
https://yoav.blog/2025/06/10/comparing-ai-coding-agents-with-a-real-task/
I threw an issue from a side project at 6 AI tools: Claude Code, OpenAI Codex (the CLI version), Factory.ai, Devin, Jules, and All Hands. The task was a combined bug fix + feature…
https://yoav.blog/2025/06/10/comparing-ai-coding-agents-with-a-real-task/