Zach Landes
@zachlandes.bsky.social
Partner at http://2ml.ai, an AI and data engineering consulting firm. Especially interested in LLM-as-judge, human-simulating AI, and our AI-powered future. Recovering economist, active rock climber and street photographer. San Francisco.
Knowing a lot is always going to be useful. But the marginal value of knowledge is going to fall and the value of getting AI to produce the best output will be more important. Outsourcing our brains to accomplish more work and work more broadly
October 16, 2025 at 8:03 AM
Knowing a lot is always going to be useful. But the marginal value of knowledge is going to fall and the value of getting AI to produce the best output will be more important. Outsourcing our brains to accomplish more work and work more broadly
Once workers start orchestrating AI to be able to manage more responsibilities and areas they are not expert in, those who do so most effectively become the most productive. It won’t be how much you know, it will be how well you can elicit expertise from AI to get the best output and outcomes.
October 16, 2025 at 7:59 AM
Once workers start orchestrating AI to be able to manage more responsibilities and areas they are not expert in, those who do so most effectively become the most productive. It won’t be how much you know, it will be how well you can elicit expertise from AI to get the best output and outcomes.
DolphinGemma! Google deepmind used 40 years of dolphin research data to collaborate on an LLM that predicts the next…whistle? Click? Of dolphin communication. We don’t yet know if it’s language. blog.google/technology/a...
DolphinGemma: How Google AI is helping decode dolphin communication
Dolphin researchers are using Gemma and Google Pixel phones to try to decipher how dolphins talk to one another.
blog.google
April 14, 2025 at 7:18 PM
DolphinGemma! Google deepmind used 40 years of dolphin research data to collaborate on an LLM that predicts the next…whistle? Click? Of dolphin communication. We don’t yet know if it’s language. blog.google/technology/a...
Uncertainty about tariffs is not as bad as tariffs—but it’s still very bad: why would a company invest in U.S. domestic production if the tariffs are off and on (and mostly off?). The behavior will just promote conservative business moves as a hedge, and reduce growth. Consumers stop spending, too
April 9, 2025 at 7:19 PM
Uncertainty about tariffs is not as bad as tariffs—but it’s still very bad: why would a company invest in U.S. domestic production if the tariffs are off and on (and mostly off?). The behavior will just promote conservative business moves as a hedge, and reduce growth. Consumers stop spending, too
Reposted by Zach Landes
Government abductions are unconstitutional.
Denying anyone the right to due process in America is unconstitutional.
If you’ve rationalized your silence or support for any of it because it’s happening to those you disagree with, give it a minute…
They’ll get to you soon enough.
Denying anyone the right to due process in America is unconstitutional.
If you’ve rationalized your silence or support for any of it because it’s happening to those you disagree with, give it a minute…
They’ll get to you soon enough.
March 27, 2025 at 10:09 PM
Government abductions are unconstitutional.
Denying anyone the right to due process in America is unconstitutional.
If you’ve rationalized your silence or support for any of it because it’s happening to those you disagree with, give it a minute…
They’ll get to you soon enough.
Denying anyone the right to due process in America is unconstitutional.
If you’ve rationalized your silence or support for any of it because it’s happening to those you disagree with, give it a minute…
They’ll get to you soon enough.
Sesame just open sourced the base model for their crazy impressive voice ai demo. You’ll have to finetune this and figure out the text generation component to reproduce Maya from the demo.
github.com/SesameAILabs...
github.com/SesameAILabs...
GitHub - SesameAILabs/csm: A Conversational Speech Generation Model
A Conversational Speech Generation Model. Contribute to SesameAILabs/csm development by creating an account on GitHub.
github.com
March 14, 2025 at 12:42 AM
Sesame just open sourced the base model for their crazy impressive voice ai demo. You’ll have to finetune this and figure out the text generation component to reproduce Maya from the demo.
github.com/SesameAILabs...
github.com/SesameAILabs...
When operator agents are asked to complete tasks on the web, they can either search or use their knowledge to select websites to visit. If search, then as before search rankings matter. But if knowledge, then training matters. Will AI labs (or inference providers) take $ to make LLMs prefer a brand?
March 5, 2025 at 6:54 AM
When operator agents are asked to complete tasks on the web, they can either search or use their knowledge to select websites to visit. If search, then as before search rankings matter. But if knowledge, then training matters. Will AI labs (or inference providers) take $ to make LLMs prefer a brand?
Nitpick all you want, but this is the most lifelike voice AI I’ve ever spoken to. Hoping they keep their promise of open source “soon”: www.sesame.com/research/cro...
Crossing the uncanny valley of conversational voice
At Sesame, our goal is to achieve “voice presence”—the magical quality that makes spoken interactions feel real, understood, and valued.
www.sesame.com
March 3, 2025 at 8:15 AM
Nitpick all you want, but this is the most lifelike voice AI I’ve ever spoken to. Hoping they keep their promise of open source “soon”: www.sesame.com/research/cro...
@404media.co should cover this: xAI has added censorship and bias to its system prompt for grok3 models, instructing them not to say Musk or Trump spread misinformation and to consider Musk, along with Aristotle, as paragons of rational thinking methods. www.reddit.com/r/OpenAI/s/D...
From the OpenAI community on Reddit: elon musk is trying to censor Grok 3. which the thoughts feature conveniently manages to entirely bypass.
Explore this post and more from the OpenAI community
www.reddit.com
February 23, 2025 at 5:31 PM
@404media.co should cover this: xAI has added censorship and bias to its system prompt for grok3 models, instructing them not to say Musk or Trump spread misinformation and to consider Musk, along with Aristotle, as paragons of rational thinking methods. www.reddit.com/r/OpenAI/s/D...
There is a theory that almost all decisions people make are not rationally-derived, but emotionally-derived: Rationality developed as a demonstrative tool for consensus building. If this is true, will reasoning AI models ever align with human decision making?
February 18, 2025 at 6:29 PM
There is a theory that almost all decisions people make are not rationally-derived, but emotionally-derived: Rationality developed as a demonstrative tool for consensus building. If this is true, will reasoning AI models ever align with human decision making?
An LLM that thinks in recurrent blocks, I.e. it doesn’t output it’s thoughts as tokens before the answer: arxiv.org/abs/2502.05171
May outperform Chain of thought because some ideas aren’t well represented with tokens. But alignment is out the window: the AI will kill us while promising roses
May outperform Chain of thought because some ideas aren’t well represented with tokens. But alignment is out the window: the AI will kill us while promising roses
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling...
arxiv.org
February 12, 2025 at 8:12 AM
An LLM that thinks in recurrent blocks, I.e. it doesn’t output it’s thoughts as tokens before the answer: arxiv.org/abs/2502.05171
May outperform Chain of thought because some ideas aren’t well represented with tokens. But alignment is out the window: the AI will kill us while promising roses
May outperform Chain of thought because some ideas aren’t well represented with tokens. But alignment is out the window: the AI will kill us while promising roses
Amazing and terrifying: clone a voice with 10-30 seconds of audio, run at 2x realtime on 4090. don’t answer calls from unknown numbers now or your parents will be robbed by phone www.linkedin.com/posts/vaibha...
Vaibhav Srivastav on LinkedIn: HOLY SHITT! @ZyphraAI just dropped Zonos - Apache 2.0 licensed… | 34 comments
HOLY SHITT! @ZyphraAI just dropped Zonos - Apache 2.0 licensed, Multilingual, Text to Speech model with INSTANT voice cloning! 🔥
> Zero-shot TTS with Voice… | 34 comments on LinkedIn
www.linkedin.com
February 10, 2025 at 8:38 PM
Amazing and terrifying: clone a voice with 10-30 seconds of audio, run at 2x realtime on 4090. don’t answer calls from unknown numbers now or your parents will be robbed by phone www.linkedin.com/posts/vaibha...
Reposted by Zach Landes
"It’s disgusting and, frankly, counterproductive to point out that I voted for this."
I wrote this bit that is entirely unrelated to current things
Voting for the Mayor Who Promised to Blow Up the City Doesn’t Mean I Approve of the Mayor Blowing Up the City
It’s so easy to label people these days. From the way folks have been talking, you’d think everyone falls into two buckets: those who voted against...
www.mcsweeneys.net
February 10, 2025 at 6:50 PM
"It’s disgusting and, frankly, counterproductive to point out that I voted for this."
Short and sweet letter from two Yale psychologists about how online outrage could be detrimental to effective collective action: par.nsf.gov/servlets/pur...
Signal:noise falls, collective action is misdirected to sources of ire rather than effective targets, and more. Recommended 2 page read.
Signal:noise falls, collective action is misdirected to sources of ire rather than effective targets, and more. Recommended 2 page read.
par.nsf.gov
February 6, 2025 at 7:12 AM
Short and sweet letter from two Yale psychologists about how online outrage could be detrimental to effective collective action: par.nsf.gov/servlets/pur...
Signal:noise falls, collective action is misdirected to sources of ire rather than effective targets, and more. Recommended 2 page read.
Signal:noise falls, collective action is misdirected to sources of ire rather than effective targets, and more. Recommended 2 page read.
A U.S. sovereign wealth fund despite our annual deficit? Okay, so the money comes from tariffs? Your plan is to unilaterally raise prices during an affordability crisis in America and then spend the tariff proceeds not on said affordability crisis but on…buying TikTok. It boggles the mind.
February 3, 2025 at 8:03 PM
A U.S. sovereign wealth fund despite our annual deficit? Okay, so the money comes from tariffs? Your plan is to unilaterally raise prices during an affordability crisis in America and then spend the tariff proceeds not on said affordability crisis but on…buying TikTok. It boggles the mind.
We’ve got Bluesky as a promising alternative to X with a sizable and growing user base. I think this is because a) people are truly disgusted with Elon and the X product has gotten worse. And b) Bluesky offers a comparable experience and now has enough users to keep growing on its own merits.
February 1, 2025 at 12:22 AM
We’ve got Bluesky as a promising alternative to X with a sizable and growing user base. I think this is because a) people are truly disgusted with Elon and the X product has gotten worse. And b) Bluesky offers a comparable experience and now has enough users to keep growing on its own merits.
It’s still stunning to me how much higher my productivity is with AI-assisted coding. This week I added authentication with magic links, a Postgres database, tests, and much more to a client’s app. While this is only a pilot app, the fact that I was able to do this in about 20 hours is…a miracle.
January 24, 2025 at 11:45 PM
It’s still stunning to me how much higher my productivity is with AI-assisted coding. This week I added authentication with magic links, a Postgres database, tests, and much more to a client’s app. While this is only a pilot app, the fact that I was able to do this in about 20 hours is…a miracle.
I think Gen AI is the thing that will finally make AR go mainstream. Just need someone to come up with a battery breakthrough or we’ll be wearing batteries like the Walkman era.
January 17, 2025 at 6:32 AM
I think Gen AI is the thing that will finally make AR go mainstream. Just need someone to come up with a battery breakthrough or we’ll be wearing batteries like the Walkman era.
AI generated images of people are now practically indistinguishable from real. Will this shine a light on the unreality of influencers, reducing their sway on our culture? www.reddit.com/r/ChatGPT/s/...
From the ChatGPT community on Reddit: We are doomed
Explore this post and more from the ChatGPT community
www.reddit.com
January 6, 2025 at 7:05 PM
AI generated images of people are now practically indistinguishable from real. Will this shine a light on the unreality of influencers, reducing their sway on our culture? www.reddit.com/r/ChatGPT/s/...
Wrote a little intro to speculative decoding and why you probably want it for its 10-300% speed up of LLM inference: www.linkedin.com/posts/zachla...
Zach Landes on LinkedIn: AI practitioners and leaders, do you know about 𝘀𝗽𝗲𝗰𝘂𝗹𝗮𝘁𝗶𝘃𝗲…
AI practitioners and leaders, do you know about 𝘀𝗽𝗲𝗰𝘂𝗹𝗮𝘁𝗶𝘃𝗲 𝗱𝗲𝗰𝗼𝗱𝗶𝗻𝗴?
As a fan of running open-source AI, I wanted to try speculative…
www.linkedin.com
December 19, 2024 at 10:05 PM
Wrote a little intro to speculative decoding and why you probably want it for its 10-300% speed up of LLM inference: www.linkedin.com/posts/zachla...
Reposted by Zach Landes
I recommend no fewer than 4 copies of any beloved book. A paperback for traveling and lending to friends, an eBook for reading with greasy snack fingers, an audio book so you know how the characters' names are actually pronounced, and a pristine hardcover to be buried with you like a pharaoh
December 19, 2024 at 6:51 PM
I recommend no fewer than 4 copies of any beloved book. A paperback for traveling and lending to friends, an eBook for reading with greasy snack fingers, an audio book so you know how the characters' names are actually pronounced, and a pristine hardcover to be buried with you like a pharaoh