Don't Worry About the Vase
@thezvi.blogsky.venki.dev
A world made of gears. Doing both speed premium short term updates and long term world model building.
Automated feed of thezvi.substack.com, find more feeds at @blogsky.venki.dev
Automated feed of thezvi.substack.com, find more feeds at @blogsky.venki.dev
DeepSeek v3.2 Is Okay And Cheap But Slow
DeepSeek v3.2 Is Okay And Cheap But Slow
DeepSeek v3.2 is DeepSeek’s latest open model release with strong bencharks.
thezvi.substack.com
December 5, 2025 at 4:25 PM
DeepSeek v3.2 Is Okay And Cheap But Slow
AI #145: You've Got Soul
AI #145: You've Got Soul
The cycle of language model releases is, one at least hopes, now complete.
thezvi.substack.com
December 4, 2025 at 2:56 PM
AI #145: You've Got Soul
On Dwarkesh Patel's Second Interview With Ilya Sutskever
On Dwarkesh Patel's Second Interview With Ilya Sutskever
Some podcasts are self-recommending on the ‘yep, I’m going to be breaking this one down’ level.
thezvi.substack.com
December 3, 2025 at 4:25 PM
On Dwarkesh Patel's Second Interview With Ilya Sutskever
Reward Mismatches in RL Cause Emergent Misalignment
Reward Mismatches in RL Cause Emergent Misalignment
Learning to do misaligned-coded things anywhere teaches an AI (or a human) to do misaligned-coded things everywhere.
thezvi.substack.com
December 2, 2025 at 4:31 PM
Reward Mismatches in RL Cause Emergent Misalignment
Claude Opus 4.5 Is The Best Model Available
Claude Opus 4.5 Is The Best Model Available
Claude Opus 4.5 is the best model currently available.
thezvi.substack.com
December 1, 2025 at 1:55 PM
Claude Opus 4.5 Is The Best Model Available
Claude Opus 4.5: Model Card, Alignment and Safety
Claude Opus 4.5: Model Card, Alignment and Safety
They saved the best for last.
thezvi.substack.com
November 28, 2025 at 2:00 PM
Claude Opus 4.5: Model Card, Alignment and Safety
AI #144: Thanks For the Models
AI #144: Thanks For the Models
Thanks for everything.
thezvi.substack.com
November 27, 2025 at 4:20 PM
AI #144: Thanks For the Models
The Big Nonprofits Post 2025
The Big Nonprofits Post 2025
thezvi.substack.com
November 26, 2025 at 7:00 PM
The Big Nonprofits Post 2025
ChatGPT 5.1 Codex Max
ChatGPT 5.1 Codex Max
OpenAI has given us GPT-5.1-Codex-Max, their best coding model for OpenAI Codex.
thezvi.substack.com
November 25, 2025 at 1:55 PM
ChatGPT 5.1 Codex Max
Gemini 3 Pro Is a Vast Intelligence With No Spine
Gemini 3 Pro Is a Vast Intelligence With No Spine
It’s A Great Model, Sir
thezvi.substack.com
November 24, 2025 at 1:55 PM
Gemini 3 Pro Is a Vast Intelligence With No Spine
Gemini 3: Model Card and Safety Framework Report
Gemini 3: Model Card and Safety Framework Report
thezvi.substack.com
November 21, 2025 at 4:41 PM
Gemini 3: Model Card and Safety Framework Report
AI #143: Everything, Everywhere, All At Once
AI #143: Everything, Everywhere, All At Once
Last week had the release of GPT-5.1, which I covered on Tuesday.
thezvi.substack.com
November 20, 2025 at 6:15 PM
AI #143: Everything, Everywhere, All At Once
Monthly Roundup #36: November 2025
Monthly Roundup #36: November 2025
Happy Gemini Week to those who celebrate.
thezvi.substack.com
November 19, 2025 at 8:56 PM
Monthly Roundup #36: November 2025
GPT 5.1 Follows Custom Instructions and Glazes
GPT 5.1 Follows Custom Instructions and Glazes
thezvi.substack.com
November 18, 2025 at 5:25 PM
GPT 5.1 Follows Custom Instructions and Glazes
On Writing #2
On Writing #2
In honor of my dropping by Inkhaven at Lighthaven in Berkeley this week, I figured it was time for another writing roundup.
thezvi.substack.com
November 17, 2025 at 1:45 PM
On Writing #2
AI Craziness: Additional Suicide Lawsuits and The Fate of GPT-4o
AI Craziness: Additional Suicide Lawsuits and The Fate of GPT-4o
thezvi.substack.com
November 14, 2025 at 8:35 PM
AI Craziness: Additional Suicide Lawsuits and The Fate of GPT-4o
AI #142: Common Ground
AI #142: Common Ground
thezvi.substack.com
November 13, 2025 at 3:20 PM
AI #142: Common Ground
The Pope Offers Wisdom
The Pope Offers Wisdom
thezvi.substack.com
November 12, 2025 at 9:50 PM
The Pope Offers Wisdom
Variously Effective Altruism
Variously Effective Altruism
This post is a roundup of various things related to philanthropy, as you often find in the full monthly roundup.
thezvi.substack.com
November 10, 2025 at 7:15 PM
Variously Effective Altruism
On Sam Altman's Second Conversation with Tyler Cowen
On Sam Altman's Second Conversation with Tyler Cowen
Some podcasts are self-recommending on the ‘yep, I’m going to be breaking this one down’ level.
thezvi.substack.com
November 7, 2025 at 4:41 PM
On Sam Altman's Second Conversation with Tyler Cowen
AI #141: Give Us The Money
AI #141: Give Us The Money
OpenAI does not waste time.
thezvi.substack.com
November 6, 2025 at 2:56 PM
AI #141: Give Us The Money
Anthropic Commits To Model Weight Preservation
Anthropic Commits To Model Weight Preservation
Anthropic announced a first step on model deprecation and preservation, promising to retain the weights of all models seeing significant use, including internal use, for at the lifetime of Anthropic as a company.
thezvi.substack.com
November 5, 2025 at 9:25 PM
Anthropic Commits To Model Weight Preservation
OpenAI: The Battle of the Board: Ilya's Testimony
OpenAI: The Battle of the Board: Ilya's Testimony
thezvi.substack.com
November 4, 2025 at 7:25 PM
OpenAI: The Battle of the Board: Ilya's Testimony
OpenAI Moves To Complete Potentially The Largest Theft In Human History
OpenAI Moves To Complete Potentially The Largest Theft In Human History
OpenAI is now set to become a Public Benefit Corporation, with its investors entitled to uncapped profit shares.
thezvi.substack.com
October 31, 2025 at 1:20 PM
OpenAI Moves To Complete Potentially The Largest Theft In Human History