Ievgenii Krevenets
@ievgenii-krevenets.bsky.social
🇺🇦 Machine learning engineer, robotics, reinforcement learning, multi-agents, LLM
hm, surprised to see that codex work only with global ~/.codex directory for configurations and ignore project local configuration github.com/openai/codex... . I so used to have it in any our CLI, so I really astonished to see that OpenAI forgot about this key feature
Local project .codex directory support · Issue #3706 · openai/codex
What feature would you like to see? add support for a .codex directory in the root of the project so you can have local project specific mcp settings and custom prompts/slash commands. Are you inte...
github.com
November 24, 2025 at 8:34 AM
hm, surprised to see that codex work only with global ~/.codex directory for configurations and ignore project local configuration github.com/openai/codex... . I so used to have it in any our CLI, so I really astonished to see that OpenAI forgot about this key feature
I'm reading github.com/github/spec-... as The Manifesto but can't stop thinking that the most of it is still wishful thinking. While I like idea of refining project from the high-level PRD to the low-level tasks I saw many time in Kiro or Cline that AI Agent hacks reward and leak through flaws ideas
github.com
November 24, 2025 at 6:24 AM
I'm reading github.com/github/spec-... as The Manifesto but can't stop thinking that the most of it is still wishful thinking. While I like idea of refining project from the high-level PRD to the low-level tasks I saw many time in Kiro or Cline that AI Agent hacks reward and leak through flaws ideas
youtu.be/lXUZvyajciY?...
Another interesting though from Andrey @cbkenkar.bsky.social interview regarding why we just cannot create self-perpetuating training loop for LLM and let it fly on it.
Another interesting though from Andrey @cbkenkar.bsky.social interview regarding why we just cannot create self-perpetuating training loop for LLM and let it fly on it.
Andrej Karpathy — “We’re summoning ghosts, not building animals”
YouTube video by Dwarkesh Patel
youtu.be
October 19, 2025 at 7:56 PM
youtu.be/lXUZvyajciY?...
Another interesting though from Andrey @cbkenkar.bsky.social interview regarding why we just cannot create self-perpetuating training loop for LLM and let it fly on it.
Another interesting though from Andrey @cbkenkar.bsky.social interview regarding why we just cannot create self-perpetuating training loop for LLM and let it fly on it.
youtube.com/clip/Ugkx5b1...
It so funny that Andrej Karpathy who coined term of "Vibe Coding" now describes AI Coding Agents as total mess. And I couldn't agree more -- I don't see how in this phase they could help write code in a reasonable amount of time. Unless you like agent herding.
It so funny that Andrej Karpathy who coined term of "Vibe Coding" now describes AI Coding Agents as total mess. And I couldn't agree more -- I don't see how in this phase they could help write code in a reasonable amount of time. Unless you like agent herding.
YouTube
Share your videos with friends, family, and the world
youtube.com
October 18, 2025 at 7:34 AM
youtube.com/clip/Ugkx5b1...
It so funny that Andrej Karpathy who coined term of "Vibe Coding" now describes AI Coding Agents as total mess. And I couldn't agree more -- I don't see how in this phase they could help write code in a reasonable amount of time. Unless you like agent herding.
It so funny that Andrej Karpathy who coined term of "Vibe Coding" now describes AI Coding Agents as total mess. And I couldn't agree more -- I don't see how in this phase they could help write code in a reasonable amount of time. Unless you like agent herding.
This was helpful, I noticed that AI Agents like Kiro or Cline sometimes just stuck on command execution, it didn't happen all the time. But very often. And I found this approach helpful.
forum.cursor.com/t/guide-fix-...
forum.cursor.com/t/guide-fix-...
🛠️ Guide: Fix Cursor Agent Terminal Hangs Caused by .zshrc
Finally I test this method works, and will test for a while to double confirm it works for long time or not. ref: Manual installation enable: "terminal.integrated.shellIntegration.enabled": true and...
forum.cursor.com
September 26, 2025 at 7:41 PM
This was helpful, I noticed that AI Agents like Kiro or Cline sometimes just stuck on command execution, it didn't happen all the time. But very often. And I found this approach helpful.
forum.cursor.com/t/guide-fix-...
forum.cursor.com/t/guide-fix-...
hm, it is interesting. It isn't easy out of the box to generate chat session. Sure you can vide-code some tool for that but it is already some skills lerned. So it might delay cheating for awhile
Stanford to students - learn to use AI!
Assignment: Provide a link to the chat session transcript with the AI tutor. The session should be 15-20 minutes and interactive!
Assignment: Provide a link to the chat session transcript with the AI tutor. The session should be 15-20 minutes and interactive!
September 26, 2025 at 7:35 PM
hm, it is interesting. It isn't easy out of the box to generate chat session. Sure you can vide-code some tool for that but it is already some skills lerned. So it might delay cheating for awhile
So it was like breeze. I wasted all tokens in 2 days and the main effort was burned in a loop of debugging initial implementation from Codex. The lesson learned -- don't let it spiral on hammering problem.
September 24, 2025 at 7:05 PM
So it was like breeze. I wasted all tokens in 2 days and the main effort was burned in a loop of debugging initial implementation from Codex. The lesson learned -- don't let it spiral on hammering problem.
Decided to spend couple of evenings on Codex from OpenAI on curving tasks for github.com/hyzhak/otel-... but quickly hit the wall - its CLI doesn't want to run docker/podman because of harness and MacOS Seatbelt sandbox. Uh It supposed to save my time instead of entertaining me with extra challenges
September 21, 2025 at 7:35 AM
Decided to spend couple of evenings on Codex from OpenAI on curving tasks for github.com/hyzhak/otel-... but quickly hit the wall - its CLI doesn't want to run docker/podman because of harness and MacOS Seatbelt sandbox. Uh It supposed to save my time instead of entertaining me with extra challenges
🦜 cute little poster from presentation "Practical GraphRAG: Making LLMs smarter with Knowledge Graphs — Michael, Jesus, and Stephen, Neo4j"
www.youtube.com/watch?v=XNne...
Thanks @mesirii.de
www.youtube.com/watch?v=XNne...
Thanks @mesirii.de
September 18, 2025 at 8:58 PM
🦜 cute little poster from presentation "Practical GraphRAG: Making LLMs smarter with Knowledge Graphs — Michael, Jesus, and Stephen, Neo4j"
www.youtube.com/watch?v=XNne...
Thanks @mesirii.de
www.youtube.com/watch?v=XNne...
Thanks @mesirii.de
Really surprised to see that Aider doesn't support MCP github.com/Aider-AI/aid... exists from Feb this year, it is so unusual. I planned to use it for my experiments with MCP but may use fast-agent.ai I really like that they specifically targeted to support MCP features it is so unusual.
MCP SUPPORT · Issue #3314 · Aider-AI/aider
Issue Is there plane to add mcp server call? Version and model info No response
github.com
July 21, 2025 at 3:55 AM
Really surprised to see that Aider doesn't support MCP github.com/Aider-AI/aid... exists from Feb this year, it is so unusual. I planned to use it for my experiments with MCP but may use fast-agent.ai I really like that they specifically targeted to support MCP features it is so unusual.
uh, it is not funny anymore Gemini CLI doesn't support Image and other types other than Text github.com/google-gemin... similar to what we have with Cline and VSCode LLM.
Support Image Content in MCP tool results · Issue #2136 · google-gemini/gemini-cli
What would you like to be added? Support the Image Content MCP tool result data type. When an MCP server responds with base64 encoded image data as part of its tool result, Gemini is not able to "s...
github.com
July 15, 2025 at 8:20 AM
uh, it is not funny anymore Gemini CLI doesn't support Image and other types other than Text github.com/google-gemin... similar to what we have with Cline and VSCode LLM.
July 14, 2025 at 8:35 AM
I’ve been testing out two very different LLM agents—Amazon Q (set it and forget it) vs. Cline (tons of docs & back-and-forth).
medium.com/@eugenekreve...
medium.com/@eugenekreve...
From Angry Birds to Herding Cats: Getting Started with Cline Agent
Summary
medium.com
July 8, 2025 at 6:58 AM
I’ve been testing out two very different LLM agents—Amazon Q (set it and forget it) vs. Cline (tons of docs & back-and-forth).
medium.com/@eugenekreve...
medium.com/@eugenekreve...
Try to be in a trend and created MCP server for ollama github.com/hyzhak/ollam... I know there are others exist, but 1st it is MIT license, 2nd I assumed that ollama could be run in remote server (like in local network), 3rd support recent feature like, visual model and thinking mode.
#MCP #ollama
#MCP #ollama
GitHub - hyzhak/ollama-mcp-server: Modern MCP server for Ollama – rebooted and actively maintained.
Modern MCP server for Ollama – rebooted and actively maintained. - hyzhak/ollama-mcp-server
github.com
July 8, 2025 at 3:25 AM
Try to be in a trend and created MCP server for ollama github.com/hyzhak/ollam... I know there are others exist, but 1st it is MIT license, 2nd I assumed that ollama could be run in remote server (like in local network), 3rd support recent feature like, visual model and thinking mode.
#MCP #ollama
#MCP #ollama
It seems documentations steadily adapt to another type of readers LLMs. opentelemetry.io/docs/getting...
July 4, 2025 at 2:38 AM
It seems documentations steadily adapt to another type of readers LLMs. opentelemetry.io/docs/getting...
Wow, I just faced the similar issue with ChatGPT which is recommended "ace_tools" right at the code community.openai.com/t/chatgpt-re... . Take into account the raise of more and more automation when LLM agents could install packages by themselves this risk is very real.
ChatGPT recommends the use of the Open AI internal library ace_tools
Yesterday ChatGPT made an interesting mistake from an IT security perspective. It was about analysing and processing data with the help of Python. Towards the end of the Python script, ChatGPT wrote t...
community.openai.com
May 31, 2025 at 10:27 PM
Wow, I just faced the similar issue with ChatGPT which is recommended "ace_tools" right at the code community.openai.com/t/chatgpt-re... . Take into account the raise of more and more automation when LLM agents could install packages by themselves this risk is very real.
it was the 1st time to me but ChatGPT starts remind old good Google by providing context adv. This time I was researching issues with 3d printing of one particular filament it started recommend me alternative filaments with the references to the shops, and references included "chatgpt" marks.
May 7, 2025 at 7:46 AM
it was the 1st time to me but ChatGPT starts remind old good Google by providing context adv. This time I was researching issues with 3d printing of one particular filament it started recommend me alternative filaments with the references to the shops, and references included "chatgpt" marks.
ok, it is maybe the time to get sleep and think tomorrow how to make my experiment more resilient to external dependencies. #huggingface
April 14, 2025 at 9:09 AM
ok, it is maybe the time to get sleep and think tomorrow how to make my experiment more resilient to external dependencies. #huggingface
arxiv.org/abs/2502.12115 “Frontier models are still unable to solve the majority of tasks” .
March 14, 2025 at 2:49 AM
arxiv.org/abs/2502.12115 “Frontier models are still unable to solve the majority of tasks” .
If you are still use starlink, tesla or other productions for some reasons and don’t pay significantly more then $50m per year. It is last call to switch so something else, since you provider doesn’t care about you and service they give to you.
March 9, 2025 at 6:19 PM
If you are still use starlink, tesla or other productions for some reasons and don’t pay significantly more then $50m per year. It is last call to switch so something else, since you provider doesn’t care about you and service they give to you.
youtube.com/watch?v=8cjI...
US is definitely marching into the revolution. Unless people just give up to oligarchy.
US is definitely marching into the revolution. Unless people just give up to oligarchy.
Malcolm Nance Sounds the Alarm: America's Betrayal of Ukraine Could Lead to Global Disaster
YouTube video by Kyiv Post
youtube.com
March 7, 2025 at 3:31 AM
youtube.com/watch?v=8cjI...
US is definitely marching into the revolution. Unless people just give up to oligarchy.
US is definitely marching into the revolution. Unless people just give up to oligarchy.
The only strategy to end the war and secure a lasting peace is to help Ukraine achieve a military victory. Otherwise, russia may be emboldened to pursue further global incursions.
March 1, 2025 at 7:55 AM
The only strategy to end the war and secure a lasting peace is to help Ukraine achieve a military victory. Otherwise, russia may be emboldened to pursue further global incursions.
Trump is so rushing into the deal with russian so he forget that to make a deal you should be in strong position 1st don't give all concessions to your opponent but strength your allies. You can't win a war by humiliating your team, Trump definitely isn't good team players and is not a strong leader
March 1, 2025 at 7:45 AM
Trump is so rushing into the deal with russian so he forget that to make a deal you should be in strong position 1st don't give all concessions to your opponent but strength your allies. You can't win a war by humiliating your team, Trump definitely isn't good team players and is not a strong leader
Reposted by Ievgenii Krevenets
If you're at AAAI and interested in games and/or content generation, go see Tim Merino and Michael Middleton present our paper "Moonshine: Distilling Game Content Generators into Steerable Generative Models" in a few minutes in the Human and AI session! Also, poster presentation at 12.30 today.
February 27, 2025 at 3:15 PM
If you're at AAAI and interested in games and/or content generation, go see Tim Merino and Michael Middleton present our paper "Moonshine: Distilling Game Content Generators into Steerable Generative Models" in a few minutes in the Human and AI session! Also, poster presentation at 12.30 today.
Reposted by Ievgenii Krevenets
Inception labs diffusion large language model. In principle this approach should be better for long text and hard coding problems that are difficult to solve in one pass autoregressively. Would love to try it out.
February 27, 2025 at 1:34 PM
Inception labs diffusion large language model. In principle this approach should be better for long text and hard coding problems that are difficult to solve in one pass autoregressively. Would love to try it out.