Aleksei Grebenkin
@koshadze.bsky.social
Software engineer. Full stack, iOS developer. IE. AI and ML enthusiast.
Reposted by Aleksei Grebenkin
Every time you declare open-source AI dead another team rise up
like a Whac-A-Mole.
Intelligence will be commoditized!
The remaining moat will be the quality of the product and how it integrate with co-factors. That’s were billions dollars AI companies will be built.
like a Whac-A-Mole.
Intelligence will be commoditized!
The remaining moat will be the quality of the product and how it integrate with co-factors. That’s were billions dollars AI companies will be built.
January 27, 2025 at 8:24 AM
Every time you declare open-source AI dead another team rise up
like a Whac-A-Mole.
Intelligence will be commoditized!
The remaining moat will be the quality of the product and how it integrate with co-factors. That’s were billions dollars AI companies will be built.
like a Whac-A-Mole.
Intelligence will be commoditized!
The remaining moat will be the quality of the product and how it integrate with co-factors. That’s were billions dollars AI companies will be built.
Reposted by Aleksei Grebenkin
BREAKING 🚨: Perplexity released US-hosted DeepSeek R1 reasoning option 👀
January 27, 2025 at 2:50 PM
BREAKING 🚨: Perplexity released US-hosted DeepSeek R1 reasoning option 👀
Reposted by Aleksei Grebenkin
January 27, 2025 at 5:21 PM
Today I learned that the US imposed restrictions on AI chip distribution )) This happened not too long ago, I just found out about it from this article.
blog.heim.xyz/deepseek-wha...
blog.heim.xyz/deepseek-wha...
The Rise of DeepSeek: What the Headlines Miss
TLDR: timing is political but tech is real, compute constraints bite differently than you think, and the story is more complex than "export controls failed."
blog.heim.xyz
January 27, 2025 at 1:47 PM
Today I learned that the US imposed restrictions on AI chip distribution )) This happened not too long ago, I just found out about it from this article.
blog.heim.xyz/deepseek-wha...
blog.heim.xyz/deepseek-wha...
🚀 OpenAI launches Operator - an AI agent that can browse the web & make real-world bookings. Built on new GPT-4o with CUA tech, outperforming competitors. Pro users get early access.
operator.chatgpt.com
#AI #OpenAI #FutureOfTech #AIAgent
operator.chatgpt.com
#AI #OpenAI #FutureOfTech #AIAgent
Operator
An agent that can use its own browser to perform tasks for you.
operator.chatgpt.com
January 25, 2025 at 6:32 AM
🚀 OpenAI launches Operator - an AI agent that can browse the web & make real-world bookings. Built on new GPT-4o with CUA tech, outperforming competitors. Pro users get early access.
operator.chatgpt.com
#AI #OpenAI #FutureOfTech #AIAgent
operator.chatgpt.com
#AI #OpenAI #FutureOfTech #AIAgent
Reposted by Aleksei Grebenkin
“Anthropic CEO Dario Amodei said Thursday (Jan. 23) that accelerated advances in artificial intelligence (AI), particularly in biology, can lead to a doubling of human lifespans in as little as five to 10 years ‘if we really get this AI stuff right.’” 💀
Anthropic CEO Sees AI-Powered Advances Doubling Human Lifespans | PYMNTS.com
Anthropic CEO Dario Amodei said Thursday (Jan. 23) that accelerated advances in artificial intelligence (AI), particularly in biology, can lead to a
www.pymnts.com
January 24, 2025 at 1:30 PM
“Anthropic CEO Dario Amodei said Thursday (Jan. 23) that accelerated advances in artificial intelligence (AI), particularly in biology, can lead to a doubling of human lifespans in as little as five to 10 years ‘if we really get this AI stuff right.’” 💀
It's absolutely insane - the DeepSeek R1 model was released 72 hours ago but has already made huge waves.
huggingface.co/deepseek-ai/...
It's been used on mobile devices and private servers, even simple PCs, and its quality is comparable to that of OpenAI's solutions. ->
huggingface.co/deepseek-ai/...
It's been used on mobile devices and private servers, even simple PCs, and its quality is comparable to that of OpenAI's solutions. ->
deepseek-ai/DeepSeek-R1 · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
January 24, 2025 at 7:27 AM
It's absolutely insane - the DeepSeek R1 model was released 72 hours ago but has already made huge waves.
huggingface.co/deepseek-ai/...
It's been used on mobile devices and private servers, even simple PCs, and its quality is comparable to that of OpenAI's solutions. ->
huggingface.co/deepseek-ai/...
It's been used on mobile devices and private servers, even simple PCs, and its quality is comparable to that of OpenAI's solutions. ->
Impressive
DeepSeek released a whole family of inference-scaling / "reasoning" models today, including distilled variants based on Llama and Qwen
Here are my notes on the new models, plus how I ran DeepSeek-R1-Distill-Llama-8B on my Mac using Ollama and LLM
simonwillison.net/2025/Jan/20/...
Here are my notes on the new models, plus how I ran DeepSeek-R1-Distill-Llama-8B on my Mac using Ollama and LLM
simonwillison.net/2025/Jan/20/...
DeepSeek-R1 and exploring DeepSeek-R1-Distill-Llama-8B
DeepSeek are the Chinese AI lab who dropped the best currently available open weights LLM on Christmas day, DeepSeek v3. That model was trained in part using their unreleased R1 …
simonwillison.net
January 21, 2025 at 7:14 AM
Impressive
Reposted by Aleksei Grebenkin
The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange.
January 20, 2025 at 7:26 PM
The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange.
🤖 AI coding boost but won't replace devs.
Hit 70% fast, last 30% needs human expertise 🧠
newsletter.pragmaticengineer.com/p/how-ai-wil...
#AI #Dev #Tech #Engineering
Hit 70% fast, last 30% needs human expertise 🧠
newsletter.pragmaticengineer.com/p/how-ai-wil...
#AI #Dev #Tech #Engineering
How AI-assisted coding will change software engineering: hard truths
A field guide that also covers why we need to rethink our expectations, and what software engineering really is. A guest post by software engineer and engineering leader Addy Osmani
newsletter.pragmaticengineer.com
January 21, 2025 at 7:09 AM
🤖 AI coding boost but won't replace devs.
Hit 70% fast, last 30% needs human expertise 🧠
newsletter.pragmaticengineer.com/p/how-ai-wil...
#AI #Dev #Tech #Engineering
Hit 70% fast, last 30% needs human expertise 🧠
newsletter.pragmaticengineer.com/p/how-ai-wil...
#AI #Dev #Tech #Engineering
Reposted by Aleksei Grebenkin
Wait! Over 35 *thousands* people already registered for the AI Agent course
in a few days...
time to find the biggest conference center in the world and gather everyone for a giga-conference on agents?
you can still join the online course here bit.ly/hf-learn-age...
in a few days...
time to find the biggest conference center in the world and gather everyone for a giga-conference on agents?
you can still join the online course here bit.ly/hf-learn-age...
Hugging Face
Hugging Face Email Forms
bit.ly
January 20, 2025 at 9:15 AM
Wait! Over 35 *thousands* people already registered for the AI Agent course
in a few days...
time to find the biggest conference center in the world and gather everyone for a giga-conference on agents?
you can still join the online course here bit.ly/hf-learn-age...
in a few days...
time to find the biggest conference center in the world and gather everyone for a giga-conference on agents?
you can still join the online course here bit.ly/hf-learn-age...
👹 Meet Nepenthes: The AI Crawler Trap
This mad lad created a tool that generates endless pages of gibberish to trap AI web crawlers in an infinite maze. Goal? Mess up future LLM training datasets lol
🤔
zadzmo.org/code/nepenth...
#AI #TechMemes #ChaosMode
This mad lad created a tool that generates endless pages of gibberish to trap AI web crawlers in an infinite maze. Goal? Mess up future LLM training datasets lol
🤔
zadzmo.org/code/nepenth...
#AI #TechMemes #ChaosMode
ZADZMO code
zadzmo.org
January 20, 2025 at 7:45 AM
👹 Meet Nepenthes: The AI Crawler Trap
This mad lad created a tool that generates endless pages of gibberish to trap AI web crawlers in an infinite maze. Goal? Mess up future LLM training datasets lol
🤔
zadzmo.org/code/nepenth...
#AI #TechMemes #ChaosMode
This mad lad created a tool that generates endless pages of gibberish to trap AI web crawlers in an infinite maze. Goal? Mess up future LLM training datasets lol
🤔
zadzmo.org/code/nepenth...
#AI #TechMemes #ChaosMode
Reposted by Aleksei Grebenkin
Evidence of AI being good for education remains really weak - but it's still not going to stop Microsoft absolutely cramming "copilots" and "agents" into schools, as I'm unhappy to learn they're due announce in the UK this week www.microsoft.com/en-us/educat...
“The best-performing LLM was GPT-4 Turbo, but it only achieved about 46% accuracy — not much higher than random guessing.”
AI isn’t very good at history, new paper finds | TechCrunch
Top LLMs performed poorly on a high-level history test, a new paper has found.
techcrunch.com
January 19, 2025 at 4:49 PM
Evidence of AI being good for education remains really weak - but it's still not going to stop Microsoft absolutely cramming "copilots" and "agents" into schools, as I'm unhappy to learn they're due announce in the UK this week www.microsoft.com/en-us/educat...
Reposted by Aleksei Grebenkin
"Treatment Effect Estimation with Observational Network Data using Machine Learning"
Arxiv: arxiv.org/abs/2206.14591
#rstats code: github.com/corinne-rahe...
#stats
Arxiv: arxiv.org/abs/2206.14591
#rstats code: github.com/corinne-rahe...
#stats
January 20, 2025 at 3:02 AM
"Treatment Effect Estimation with Observational Network Data using Machine Learning"
Arxiv: arxiv.org/abs/2206.14591
#rstats code: github.com/corinne-rahe...
#stats
Arxiv: arxiv.org/abs/2206.14591
#rstats code: github.com/corinne-rahe...
#stats
Reposted by Aleksei Grebenkin
We are hiring! We currently have several great job openings at the Interaction Design Lab at the Potsdam University of Applied Sciences: idl.fh-potsdam.de/de/jobs/ Apply Now! #DesignResearch #HumanComputerInteraction #AI #Teaching #DigitalLiteracy #Design
January 20, 2025 at 5:34 AM
We are hiring! We currently have several great job openings at the Interaction Design Lab at the Potsdam University of Applied Sciences: idl.fh-potsdam.de/de/jobs/ Apply Now! #DesignResearch #HumanComputerInteraction #AI #Teaching #DigitalLiteracy #Design
Google presents an architecture named Titan - a neural long-term memory that, as a meta in-context learner, learns to memorize at test time.
Here is a good explanation- www.youtube.com/watch?v=6iEg...
Source: arxiv.org/pdf/2501.00663
Here is a good explanation- www.youtube.com/watch?v=6iEg...
Source: arxiv.org/pdf/2501.00663
Google Research Paper: Titans Architecture Solves AI Memory?
YouTube video by Nate B Jones
www.youtube.com
January 16, 2025 at 8:14 AM
Google presents an architecture named Titan - a neural long-term memory that, as a meta in-context learner, learns to memorize at test time.
Here is a good explanation- www.youtube.com/watch?v=6iEg...
Source: arxiv.org/pdf/2501.00663
Here is a good explanation- www.youtube.com/watch?v=6iEg...
Source: arxiv.org/pdf/2501.00663
Reposted by Aleksei Grebenkin
fantechno.com/nvidia-inves...
Nvidia Invests in MetAI: A Breakthrough in Digital Twin Creation
#AI #ArtificialIntelligence #Business #MetAI #Partnership #Taiwan
Nvidia Invests in MetAI: A Breakthrough in Digital Twin Creation
#AI #ArtificialIntelligence #Business #MetAI #Partnership #Taiwan
Nvidia Invests in MetAI: A Breakthrough in Digital Twin Creation | FanTechno.com
The American tech giant Nvidia has announced a strategic investment in Taiwanese startup MetAI, which specializes in developing innovative solutions for creating digital twins using artificial intelli...
fantechno.com
January 15, 2025 at 7:45 AM
fantechno.com/nvidia-inves...
Nvidia Invests in MetAI: A Breakthrough in Digital Twin Creation
#AI #ArtificialIntelligence #Business #MetAI #Partnership #Taiwan
Nvidia Invests in MetAI: A Breakthrough in Digital Twin Creation
#AI #ArtificialIntelligence #Business #MetAI #Partnership #Taiwan
Reposted by Aleksei Grebenkin
US President Joe Biden signed an executive order to provide federal support to address massive energy needs for fast-growing advanced artificial intelligence data centers, the White House said reut.rs/42eyQMt
Biden signs executive order to ensure power for AI data centers
President Joe Biden signed an executive order on Tuesday to provide federal support to address massive energy needs for fast-growing advanced artificial intelligence data centers, the White House said.
reut.rs
January 14, 2025 at 3:17 PM
US President Joe Biden signed an executive order to provide federal support to address massive energy needs for fast-growing advanced artificial intelligence data centers, the White House said reut.rs/42eyQMt
Reposted by Aleksei Grebenkin
Paper: ai.nejm.org/doi/full/10....
Using Large Language Models to Promote Health Equity
While the discussion about the effects of large language models (LLMs) on health equity
has been largely cautionary, LLMs also present significant opportunities for improving
health equity. We high...
ai.nejm.org
January 14, 2025 at 5:48 PM
Paper: ai.nejm.org/doi/full/10....
Seems cool 😎
It also comes with cheap API prices and good evals 🤖
January 15, 2025 at 6:05 AM
Seems cool 😎
It is interesting how it will use a battery.
January 15, 2025 at 5:43 AM
It is interesting how it will use a battery.
Reposted by Aleksei Grebenkin
I am seeing a lot of bad takes on passkeys by technologists and even security engineers that don’t work on end user authentication.
I’m not an expert either, but if you trust my opinion on security, STFU and listen to people actually dealing with lockouts and ATOs on a daily basis.
I’m not an expert either, but if you trust my opinion on security, STFU and listen to people actually dealing with lockouts and ATOs on a daily basis.
January 13, 2025 at 8:21 PM
I am seeing a lot of bad takes on passkeys by technologists and even security engineers that don’t work on end user authentication.
I’m not an expert either, but if you trust my opinion on security, STFU and listen to people actually dealing with lockouts and ATOs on a daily basis.
I’m not an expert either, but if you trust my opinion on security, STFU and listen to people actually dealing with lockouts and ATOs on a daily basis.
Reposted by Aleksei Grebenkin
📦 automatisch / automatisch
⭐ 6,873 (+21)
🗒 JavaScript
The open source Zapier alternative. Build workflow automation without spending time and money.
⭐ 6,873 (+21)
🗒 JavaScript
The open source Zapier alternative. Build workflow automation without spending time and money.
GitHub - automatisch/automatisch: The open source Zapier alternative. Build workflow automation without spending time and money.
The open source Zapier alternative. Build workflow automation without spending time and money. - automatisch/automatisch
github.com
January 14, 2025 at 12:01 AM
📦 automatisch / automatisch
⭐ 6,873 (+21)
🗒 JavaScript
The open source Zapier alternative. Build workflow automation without spending time and money.
⭐ 6,873 (+21)
🗒 JavaScript
The open source Zapier alternative. Build workflow automation without spending time and money.
Reposted by Aleksei Grebenkin
How to deploy (the easy way) to Kubernetes
www.realworldml.net/blog/deploy-...
www.realworldml.net/blog/deploy-...
Let's deploy our REST API to Kubernetes 🚀
www.realworldml.net
January 14, 2025 at 7:07 AM
How to deploy (the easy way) to Kubernetes
www.realworldml.net/blog/deploy-...
www.realworldml.net/blog/deploy-...
The most relevant and serious ML and AI courses are deeplearning.ai and www.nvidia.com/en-us/traini.... Ain't it?
January 14, 2025 at 5:54 AM
The most relevant and serious ML and AI courses are deeplearning.ai and www.nvidia.com/en-us/traini.... Ain't it?