@thorfinson12.bsky.social
Reposted
Reposted
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (over 300 pages)
A comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs, systematically examining
A comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs, systematically examining
December 4, 2025 at 12:10 AM
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (over 300 pages)
A comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs, systematically examining
A comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs, systematically examining
Reposted
Robots have been getting really dynamic and capable this year while no one was looking
November 16, 2025 at 2:53 AM
Robots have been getting really dynamic and capable this year while no one was looking
Reposted
Free eBook: Machine Learning Systems
Principles and Practices of Engineering Artificially Intelligent Systems
This textbook bridges the gap between theoretical foundations and practical engineering, emphasizing the systems perspective required to build effective AI solutions.
www.mlsysbook.ai
Principles and Practices of Engineering Artificially Intelligent Systems
This textbook bridges the gap between theoretical foundations and practical engineering, emphasizing the systems perspective required to build effective AI solutions.
www.mlsysbook.ai
ML Systems Textbook
www.mlsysbook.ai
November 1, 2025 at 2:04 AM
Free eBook: Machine Learning Systems
Principles and Practices of Engineering Artificially Intelligent Systems
This textbook bridges the gap between theoretical foundations and practical engineering, emphasizing the systems perspective required to build effective AI solutions.
www.mlsysbook.ai
Principles and Practices of Engineering Artificially Intelligent Systems
This textbook bridges the gap between theoretical foundations and practical engineering, emphasizing the systems perspective required to build effective AI solutions.
www.mlsysbook.ai
Reposted
BAAI's Emu3.5
A large-scale multimodal world model that natively predicts the next vision-language state. Emu3.5 outperforms Nano Banana across image generation, editing, interleaved tasks and more.
A large-scale multimodal world model that natively predicts the next vision-language state. Emu3.5 outperforms Nano Banana across image generation, editing, interleaved tasks and more.
October 31, 2025 at 7:16 PM
BAAI's Emu3.5
A large-scale multimodal world model that natively predicts the next vision-language state. Emu3.5 outperforms Nano Banana across image generation, editing, interleaved tasks and more.
A large-scale multimodal world model that natively predicts the next vision-language state. Emu3.5 outperforms Nano Banana across image generation, editing, interleaved tasks and more.
Reposted
A Practitioner's Guide to Kolmogorov-Arnold Networks
A systematic and comprehensive overview of the rapidly expanding KAN landscape, moving beyond simple performance comparisons to offer a structured synthesis of theoretical foundations, architectural variants, and practical implementation strategi
A systematic and comprehensive overview of the rapidly expanding KAN landscape, moving beyond simple performance comparisons to offer a structured synthesis of theoretical foundations, architectural variants, and practical implementation strategi
October 31, 2025 at 11:04 PM
A Practitioner's Guide to Kolmogorov-Arnold Networks
A systematic and comprehensive overview of the rapidly expanding KAN landscape, moving beyond simple performance comparisons to offer a structured synthesis of theoretical foundations, architectural variants, and practical implementation strategi
A systematic and comprehensive overview of the rapidly expanding KAN landscape, moving beyond simple performance comparisons to offer a structured synthesis of theoretical foundations, architectural variants, and practical implementation strategi
Reposted
We’re updating olmOCR, our model for turning PDFs & scans into clean text with support for tables, equations, handwriting, & more. olmOCR 2 uses synthetic data + unit tests as verifiable rewards to reach state-of-the-art performance on challenging documents. 🧵
October 22, 2025 at 4:09 PM
We’re updating olmOCR, our model for turning PDFs & scans into clean text with support for tables, equations, handwriting, & more. olmOCR 2 uses synthetic data + unit tests as verifiable rewards to reach state-of-the-art performance on challenging documents. 🧵
Reposted
Jesse has been a heavy user of parallel agents for a lot longer than I have, and shares some of his processes here (including how he uses git worktrees) bsky.app/profile/s.ly...
Wrote up my September 2025 "How I'm using coding agents" - blog.fsck.com/2025/10/05/h...
How I'm using coding agents in September, 2025
I used to write more
blog.fsck.com
October 5, 2025 at 5:49 PM
Jesse has been a heavy user of parallel agents for a lot longer than I have, and shares some of his processes here (including how he uses git worktrees) bsky.app/profile/s.ly...
Reposted
“We're in the crazy stage. It is late '99. This is, from a fundamental standpoint, identical to all the past major manias. We're bumping up against history real hard now." www.businessinsider.com/ai-stocks-bu...
People are chasing AI stocks like 'dogs chase cars' — and a crash looks certain, veteran investor Bill Smead says
Bill Smead said "momentum" is fueling the AI boom, the tech's potential is already priced in, and it's almost a replay of the dot-com bubble.
www.businessinsider.com
October 1, 2025 at 8:45 PM
“We're in the crazy stage. It is late '99. This is, from a fundamental standpoint, identical to all the past major manias. We're bumping up against history real hard now." www.businessinsider.com/ai-stocks-bu...
Reposted
Thinking Machines (Mira Murati's company)'s Tinker: a flexible API for fine-tuning language models.
Focus on what matters in LLM training - your data and algorithms - while they handle the heavy lifting of distributed training.
thinkingmachines.ai/tinker/
Focus on what matters in LLM training - your data and algorithms - while they handle the heavy lifting of distributed training.
thinkingmachines.ai/tinker/
Tinker
Tinker is a training API for researchers and developers.
thinkingmachines.ai
October 1, 2025 at 10:26 PM
Thinking Machines (Mira Murati's company)'s Tinker: a flexible API for fine-tuning language models.
Focus on what matters in LLM training - your data and algorithms - while they handle the heavy lifting of distributed training.
thinkingmachines.ai/tinker/
Focus on what matters in LLM training - your data and algorithms - while they handle the heavy lifting of distributed training.
thinkingmachines.ai/tinker/
Reposted
@nouhadziri.bsky.social shares RL Grokking Recipe: a training recipe that enables LLMs to solve previously unsolvable coding problems!
They also dive into the heated debate: does RL just sharpen previous learnt skills or can it unlock genuinely new reasoning?
They also dive into the heated debate: does RL just sharpen previous learnt skills or can it unlock genuinely new reasoning?
RL Grokking Recipe: How Can We Enable LLMs to Solve Previously Unsolvable Tasks with RL? | Notion
Yiyou Sun¹, Yuhan Cao, Pohao Huang¹, Haoyue Bai², Hannaneh Hajishirzi³⁴, Nouha Dziri⁴♠, Dawn Song¹♠
accessible-dragon-75f.notion.site
October 1, 2025 at 10:41 PM
@nouhadziri.bsky.social shares RL Grokking Recipe: a training recipe that enables LLMs to solve previously unsolvable coding problems!
They also dive into the heated debate: does RL just sharpen previous learnt skills or can it unlock genuinely new reasoning?
They also dive into the heated debate: does RL just sharpen previous learnt skills or can it unlock genuinely new reasoning?
Reposted
Metacognitive Reuse: Turn recurring LLM reasoning into concise, reusable “behaviors”. The model learns named skills from its own chains-of-thought and reuses them to think faster & cheaper.
Paper: arxiv.org/abs/2509.13237
Paper: arxiv.org/abs/2509.13237
September 22, 2025 at 12:51 AM
Metacognitive Reuse: Turn recurring LLM reasoning into concise, reusable “behaviors”. The model learns named skills from its own chains-of-thought and reuses them to think faster & cheaper.
Paper: arxiv.org/abs/2509.13237
Paper: arxiv.org/abs/2509.13237
Reposted
@yannlecun and his collaborators (Hai Huang and Randall Balestriero) have released their first iteration of Joint Embedding Predictive Architectures (JEPAs) for LLMs.
"Learning by latent space prediction has revolutionized vision models, and it will revolutionize LLMs!, which they termed as
"Learning by latent space prediction has revolutionized vision models, and it will revolutionize LLMs!, which they termed as
September 21, 2025 at 10:23 PM
@yannlecun and his collaborators (Hai Huang and Randall Balestriero) have released their first iteration of Joint Embedding Predictive Architectures (JEPAs) for LLMs.
"Learning by latent space prediction has revolutionized vision models, and it will revolutionize LLMs!, which they termed as
"Learning by latent space prediction has revolutionized vision models, and it will revolutionize LLMs!, which they termed as
Reposted
I've been nerdsniped by the idea of Semantic IDs.
Here's the result of my training runs:
• RQ-VAE to compress item embeddings into tokens
• SASRec to predict the next item (i.e., 4-tokens) exactly
• Qwen3-8B that can return recs and natural language!
eugeneyan.com/writing/sema...
Here's the result of my training runs:
• RQ-VAE to compress item embeddings into tokens
• SASRec to predict the next item (i.e., 4-tokens) exactly
• Qwen3-8B that can return recs and natural language!
eugeneyan.com/writing/sema...
How to Train an LLM-RecSys Hybrid for Steerable Recs with Semantic IDs
An LLM that can converse in English & item IDs, and make recommendations w/o retrieval or tools.
eugeneyan.com
September 17, 2025 at 2:04 AM
I've been nerdsniped by the idea of Semantic IDs.
Here's the result of my training runs:
• RQ-VAE to compress item embeddings into tokens
• SASRec to predict the next item (i.e., 4-tokens) exactly
• Qwen3-8B that can return recs and natural language!
eugeneyan.com/writing/sema...
Here's the result of my training runs:
• RQ-VAE to compress item embeddings into tokens
• SASRec to predict the next item (i.e., 4-tokens) exactly
• Qwen3-8B that can return recs and natural language!
eugeneyan.com/writing/sema...
Reposted
Evaluation of machine learning-assisted directed evolution across diverse combinatorial landscapes: Cell Systems
#microsky
www.cell.com/cell-systems...
#microsky
www.cell.com/cell-systems...
Evaluation of machine learning-assisted directed evolution across diverse combinatorial landscapes
Machine learning-assisted directed evolution (MLDE) enhances protein engineering by
efficiently identifying high-fitness variants. However, a limited understanding of
MLDE performance across diverse p...
www.cell.com
September 12, 2025 at 9:29 PM
Evaluation of machine learning-assisted directed evolution across diverse combinatorial landscapes: Cell Systems
#microsky
www.cell.com/cell-systems...
#microsky
www.cell.com/cell-systems...
Reposted
By applying these design patterns, you gain access to solutions that enhance the structure, maintainability, reliability, and efficiency of the agents you build on your canvas.
docs.google.com/document/d/1...
docs.google.com/document/d/1...
Agentic Design Patterns
Agentic Design Patterns A Hands-On Guide to Building Intelligent Systems, Antonio Gulli Table of Contents - total 424 pages = 1+2+1+1+4+9+103+61+34+114+74+5+4 11 Dedication, 1 page Acknowledgment,...
docs.google.com
September 3, 2025 at 12:48 AM
By applying these design patterns, you gain access to solutions that enhance the structure, maintainability, reliability, and efficiency of the agents you build on your canvas.
docs.google.com/document/d/1...
docs.google.com/document/d/1...
Reposted
Printable, large-format poster that explains how to make a scientific poster. Sharing because it's the start of the semester. #science #research #teaching #academia 🧪 colinpurrington.com/2017/03/conf...
Conference poster full of tips for creating conference posters » Colin Purrington's blog
In case you need a quick guide to making a conference poster, here’s a poster of poster tips. Just view click to enlarge, or display with a room projector and invite students to come up and read toget...
colinpurrington.com
August 27, 2025 at 7:06 PM
Printable, large-format poster that explains how to make a scientific poster. Sharing because it's the start of the semester. #science #research #teaching #academia 🧪 colinpurrington.com/2017/03/conf...
Reposted
Some: EU is small.
Reality: EU has surpassed China as the second biggest economy globally.
We need to stop playing second fiddle to US, Russia, China, India & instead punch in our weight class being an equal superpower.
Need to #federalize EU to accomplish this.
🇪🇺
#EUsky
www.imf.org/external/dat...
Reality: EU has surpassed China as the second biggest economy globally.
We need to stop playing second fiddle to US, Russia, China, India & instead punch in our weight class being an equal superpower.
Need to #federalize EU to accomplish this.
🇪🇺
#EUsky
www.imf.org/external/dat...
August 18, 2025 at 11:40 AM
Some: EU is small.
Reality: EU has surpassed China as the second biggest economy globally.
We need to stop playing second fiddle to US, Russia, China, India & instead punch in our weight class being an equal superpower.
Need to #federalize EU to accomplish this.
🇪🇺
#EUsky
www.imf.org/external/dat...
Reality: EU has surpassed China as the second biggest economy globally.
We need to stop playing second fiddle to US, Russia, China, India & instead punch in our weight class being an equal superpower.
Need to #federalize EU to accomplish this.
🇪🇺
#EUsky
www.imf.org/external/dat...
Reposted
Google Custom Search Engines for #OSINT
- Social Search (Global)
- Most Wanted and Sanctioned
- Stats and Docs
- Wikileaks
- Fact-checks
- File Sharing and Clouds
- Webcam Search
- Pastebin
github.com/paulpogoda/O...
Contributor @paulpogoda.bsky.social
- Social Search (Global)
- Most Wanted and Sanctioned
- Stats and Docs
- Wikileaks
- Fact-checks
- File Sharing and Clouds
- Webcam Search
- Pastebin
github.com/paulpogoda/O...
Contributor @paulpogoda.bsky.social
August 4, 2025 at 9:31 AM
Google Custom Search Engines for #OSINT
- Social Search (Global)
- Most Wanted and Sanctioned
- Stats and Docs
- Wikileaks
- Fact-checks
- File Sharing and Clouds
- Webcam Search
- Pastebin
github.com/paulpogoda/O...
Contributor @paulpogoda.bsky.social
- Social Search (Global)
- Most Wanted and Sanctioned
- Stats and Docs
- Wikileaks
- Fact-checks
- File Sharing and Clouds
- Webcam Search
- Pastebin
github.com/paulpogoda/O...
Contributor @paulpogoda.bsky.social
Reposted
Nice that the Bluesky firehose is now becoming a live dataset on which to demo streaming databases
We have a new blog post up at @materialize.com about analyzing the Bluesky firehose (Jetstream, really) through Materialize. You can grab a copy of the community edition of MZ and follow along, or invent your own ways of looking at the data, live!
materialize.com/blog/analyzi...
materialize.com/blog/analyzi...
Analyzing Live Social Data: Exploring Social Trends on Bluesky
Bluesky provides a public firehose that we can stream into Materialize, through which we can observe live social behavior and trends.
materialize.com
July 29, 2025 at 11:25 AM
Nice that the Bluesky firehose is now becoming a live dataset on which to demo streaming databases
Reposted
Amazon launches Kiro - yet another AI Agent IDE based on Visual Studio Code.
kiro.dev/blog/introdu...
kiro.dev/blog/introdu...
July 15, 2025 at 4:15 AM
Amazon launches Kiro - yet another AI Agent IDE based on Visual Studio Code.
kiro.dev/blog/introdu...
kiro.dev/blog/introdu...
Reposted
Kimi-K2 just clinched the #1 spot on both EQ-Bench3 and Creative Writing benchmarks—outperforming many closed competitors.
Key stats: - Top rank on EQ-Bench3 - #1 in Creative Writing - Proof that open models can win big!
Key stats: - Top rank on EQ-Bench3 - #1 in Creative Writing - Proof that open models can win big!
July 14, 2025 at 4:35 PM
Kimi-K2 just clinched the #1 spot on both EQ-Bench3 and Creative Writing benchmarks—outperforming many closed competitors.
Key stats: - Top rank on EQ-Bench3 - #1 in Creative Writing - Proof that open models can win big!
Key stats: - Top rank on EQ-Bench3 - #1 in Creative Writing - Proof that open models can win big!
Reposted
NotebookLM’s featured notebooks are crafted by renowned authors, researchers, and publications. Think: longevity hacks from Dr. Eric Topol, global trends from The Economist, and expert mental health tips—all in one place, ready to explore.
blog.google/technology/g...
blog.google/technology/g...
Try featured notebooks on selected topics in NotebookLM
NotebookLM is adding featured notebooks, with works selected by partners like The Atlantic and The Economist.
blog.google
July 14, 2025 at 4:37 PM
NotebookLM’s featured notebooks are crafted by renowned authors, researchers, and publications. Think: longevity hacks from Dr. Eric Topol, global trends from The Economist, and expert mental health tips—all in one place, ready to explore.
blog.google/technology/g...
blog.google/technology/g...
Reposted
This is interesting — “The Hidden Wealth of Cities” by the World Bank, on how the PUBLIC SPACES in cities are a great economic investment and asset rather than a cost or liability. But too many cities aren’t taking smart advantage of the economic benefits. openknowledge.worldbank.org/server/api/c...
July 12, 2025 at 4:22 PM
This is interesting — “The Hidden Wealth of Cities” by the World Bank, on how the PUBLIC SPACES in cities are a great economic investment and asset rather than a cost or liability. But too many cities aren’t taking smart advantage of the economic benefits. openknowledge.worldbank.org/server/api/c...
Reposted
Meet the open-source robot shaking up the market! 🤖
Hugging Face and Pollen Robotics just unveiled Reachy Mini—a cute, affordable, and hackable robot platform that rivals commercial bots but runs on open-source magic. Designed for AI builders of all ages, it’s tiny in size but massive in potential.
Hugging Face and Pollen Robotics just unveiled Reachy Mini—a cute, affordable, and hackable robot platform that rivals commercial bots but runs on open-source magic. Designed for AI builders of all ages, it’s tiny in size but massive in potential.
July 10, 2025 at 12:51 AM
Meet the open-source robot shaking up the market! 🤖
Hugging Face and Pollen Robotics just unveiled Reachy Mini—a cute, affordable, and hackable robot platform that rivals commercial bots but runs on open-source magic. Designed for AI builders of all ages, it’s tiny in size but massive in potential.
Hugging Face and Pollen Robotics just unveiled Reachy Mini—a cute, affordable, and hackable robot platform that rivals commercial bots but runs on open-source magic. Designed for AI builders of all ages, it’s tiny in size but massive in potential.