Tzafrir Rehan
@tzafrirr.bsky.social
GenAI expert (Gen Media / Full Stack / Prompt Engineering / Indie Hacking), 10+ years CTO, Mindset Coach, DJ, Producer, Musician, Software Engineer
I'm glad to report that I was wrong on this one
It's sending me daily updates about trump and DOGE's activities
It's sending me daily updates about trump and DOGE's activities
Did anyone find a real useful use of ChatGPT tasks?
Seems like a glorified reminder system
Seems like a glorified reminder system
February 6, 2025 at 3:48 PM
I'm glad to report that I was wrong on this one
It's sending me daily updates about trump and DOGE's activities
It's sending me daily updates about trump and DOGE's activities
Did anyone find a real useful use of ChatGPT tasks?
Seems like a glorified reminder system
Seems like a glorified reminder system
January 15, 2025 at 8:46 AM
Did anyone find a real useful use of ChatGPT tasks?
Seems like a glorified reminder system
Seems like a glorified reminder system
December 23, 2024 at 2:02 PM
Is this music theory or are you trying to send a message?
December 21, 2024 at 2:46 PM
Is this music theory or are you trying to send a message?
מי קהל היעד של הקרנת 23:15 של "מבצע סנטה"? י
December 12, 2024 at 5:49 PM
מי קהל היעד של הקרנת 23:15 של "מבצע סנטה"? י
December 9, 2024 at 6:49 PM
Has anyone found a consistent way to get GPT o1 to spend extra time "thinking"?
It under-spends thinking time for my usual queries, answering within 2 seconds instead
It under-spends thinking time for my usual queries, answering within 2 seconds instead
December 6, 2024 at 4:28 PM
Has anyone found a consistent way to get GPT o1 to spend extra time "thinking"?
It under-spends thinking time for my usual queries, answering within 2 seconds instead
It under-spends thinking time for my usual queries, answering within 2 seconds instead
זה כבר כמה ימים, שבצוותא עובר רב־פקד מחופש
עם שוחד מאתמול, שאיש לא יבין, שאיש לא יתפוס
עם שוחד מאתמול, שאיש לא יבין, שאיש לא יתפוס
December 4, 2024 at 2:27 PM
זה כבר כמה ימים, שבצוותא עובר רב־פקד מחופש
עם שוחד מאתמול, שאיש לא יבין, שאיש לא יתפוס
עם שוחד מאתמול, שאיש לא יבין, שאיש לא יתפוס
בעולם הפוסט־פוסט־אמת, האם ביידן שחונן את בנו עושה יותר שכונה מהחוק והצדק מאשר צרפת שמחליקה לנתניהו על צו מעצר בינלאומי?
December 2, 2024 at 9:42 AM
בעולם הפוסט־פוסט־אמת, האם ביידן שחונן את בנו עושה יותר שכונה מהחוק והצדק מאשר צרפת שמחליקה לנתניהו על צו מעצר בינלאומי?
Interest check,
Is anyone on this feed interested in a detailed writeup on how I used Claude to figure out what works and what doesn't in LTX-Video model?
Is anyone on this feed interested in a detailed writeup on how I used Claude to figure out what works and what doesn't in LTX-Video model?
December 1, 2024 at 6:37 PM
Interest check,
Is anyone on this feed interested in a detailed writeup on how I used Claude to figure out what works and what doesn't in LTX-Video model?
Is anyone on this feed interested in a detailed writeup on how I used Claude to figure out what works and what doesn't in LTX-Video model?
November 29, 2024 at 10:25 PM
November 29, 2024 at 10:07 PM
Use case alert 💥 for Claude MCP integrations:
* Use brave search to pull breaking news
* Post them on Slack
* Read today's news while sipping your coffee like a boss ☕
* Use brave search to pull breaking news
* Post them on Slack
* Read today's news while sipping your coffee like a boss ☕
November 28, 2024 at 10:26 AM
Use case alert 💥 for Claude MCP integrations:
* Use brave search to pull breaking news
* Post them on Slack
* Read today's news while sipping your coffee like a boss ☕
* Use brave search to pull breaking news
* Post them on Slack
* Read today's news while sipping your coffee like a boss ☕
Reposted by Tzafrir Rehan
Yes their attention to detail of the user interface is on another level. You can feel the empathy in the product. It's not just a bunch of AI Bros. And somehow that shines through. It helps that the model is just a lot more sensitive to human instruction and preferences.
November 27, 2024 at 4:41 PM
Yes their attention to detail of the user interface is on another level. You can feel the empathy in the product. It's not just a bunch of AI Bros. And somehow that shines through. It helps that the model is just a lot more sensitive to human instruction and preferences.
I really like the way Claude builds its products, being very people oriented.
This is a great example, one could achieve the same result by just putting an example in the prompt, or using ChatGPT's "Custom Instructions",
But Claude speaks in the *user*'s language, about writing style.
This is a great example, one could achieve the same result by just putting an example in the prompt, or using ChatGPT's "Custom Instructions",
But Claude speaks in the *user*'s language, about writing style.
Claude's lead in writing as a frontier model is only going to improve. Anthropic announced that with styles, you can now customize how Claude responds. Select from the new preset options: Concise, Explanatory, or Formal.
Upload writing samples and Claude can automatically generate custom styles.
Upload writing samples and Claude can automatically generate custom styles.
November 27, 2024 at 4:37 PM
I really like the way Claude builds its products, being very people oriented.
This is a great example, one could achieve the same result by just putting an example in the prompt, or using ChatGPT's "Custom Instructions",
But Claude speaks in the *user*'s language, about writing style.
This is a great example, one could achieve the same result by just putting an example in the prompt, or using ChatGPT's "Custom Instructions",
But Claude speaks in the *user*'s language, about writing style.
Reposted by Tzafrir Rehan
You can work around this if you clone the repo, `npm install` it, then write the node command directly (i.e. command: "node", args: ['path/to/the/actual/js/file.js'])
November 26, 2024 at 7:49 PM
You can work around this if you clone the repo, `npm install` it, then write the node command directly (i.e. command: "node", args: ['path/to/the/actual/js/file.js'])
Tried using Claude's new MCP examples on Windows and couldn't get it to work?
It's not you, it's them: github.com/modelcontext...
Should be fixed soon
It's not you, it's them: github.com/modelcontext...
Should be fixed soon
Fix npx and other processes that launch subprocesses by anaisbetts · Pull Request #68 · modelcontextprotocol/typescript-sdk
detached: true seems to break stdio inheritance for processes, instead try to use windowsHide. Also, log stderr to the log file
github.com
November 26, 2024 at 4:21 PM
Tried using Claude's new MCP examples on Windows and couldn't get it to work?
It's not you, it's them: github.com/modelcontext...
Should be fixed soon
It's not you, it's them: github.com/modelcontext...
Should be fixed soon
Reposted by Tzafrir Rehan
Interested in open datasets for ML and AI?
I've just created this feed with posts about @huggingface.bsky.social datasets!
Don't miss the latest news and conversations about the secret sauce behind every AI model.
bsky.app/profile/dvil...
I've just created this feed with posts about @huggingface.bsky.social datasets!
Don't miss the latest news and conversations about the secret sauce behind every AI model.
bsky.app/profile/dvil...
November 25, 2024 at 3:11 PM
Interested in open datasets for ML and AI?
I've just created this feed with posts about @huggingface.bsky.social datasets!
Don't miss the latest news and conversations about the secret sauce behind every AI model.
bsky.app/profile/dvil...
I've just created this feed with posts about @huggingface.bsky.social datasets!
Don't miss the latest news and conversations about the secret sauce behind every AI model.
bsky.app/profile/dvil...
I've been following the LLM/Chess discussing since last week,
and I'm very intrigued to find out, are there other tasks that base models perform better than the chat tuned models?
I plan on running a couple evals to check
and I'm very intrigued to find out, are there other tasks that base models perform better than the chat tuned models?
I plan on running a couple evals to check
OK, I can partly explain the LLM chess weirdness now
https://dynomight.net/more-chess/
https://dynomight.net/more-chess/
November 24, 2024 at 10:51 AM
I've been following the LLM/Chess discussing since last week,
and I'm very intrigued to find out, are there other tasks that base models perform better than the chat tuned models?
I plan on running a couple evals to check
and I'm very intrigued to find out, are there other tasks that base models perform better than the chat tuned models?
I plan on running a couple evals to check
Reposted by Tzafrir Rehan
We launched new tools for FLUX.1 blackforestlabs.ai/flux-1-tools/
Introducing FLUX.1 Tools
Today, we are excited to release FLUX.1 Tools, a suite of models designed to add control and steerability to our base text-to-image model FLUX.1, enabling the modification and re-creation of real and…
blackforestlabs.ai
November 21, 2024 at 4:08 PM
We launched new tools for FLUX.1 blackforestlabs.ai/flux-1-tools/
Live now 🎉
I host a weekly show called ThursdAi.news, featuring rotating experts in AI news from OSS to Big Companies, APIs, Vision, AI Art, and more. It’s been running weekly for over a year and a half.
Just in time for X-sodus, we moved from Xspaces to YT-live, welcome to set a reminder and follow along 👇
Just in time for X-sodus, we moved from Xspaces to YT-live, welcome to set a reminder and follow along 👇
ThursdAI - Nov 21 - Highest Signal weekly AI news podcast - recorded live
YouTube video by Alex Volkov (Altryne)
www.youtube.com
November 21, 2024 at 4:44 PM
Live now 🎉
This one's very informative, showing workflows to extract information from multiple YouTube videos: www.ai-supremacy.com/p/how-to-use...
(tip: memorize the "Common mistakes to avoid" segment)
(tip: memorize the "Common mistakes to avoid" segment)
November 21, 2024 at 11:47 AM
This one's very informative, showing workflows to extract information from multiple YouTube videos: www.ai-supremacy.com/p/how-to-use...
(tip: memorize the "Common mistakes to avoid" segment)
(tip: memorize the "Common mistakes to avoid" segment)
I used a voice recorder to capture an entire conference, and @notebooklmdeepdive.bsky.social turned it into perfect conference notes 🤯
#GenAI #NotebookLM #GenerativeAI
#GenAI #NotebookLM #GenerativeAI
November 20, 2024 at 2:32 PM
I used a voice recorder to capture an entire conference, and @notebooklmdeepdive.bsky.social turned it into perfect conference notes 🤯
#GenAI #NotebookLM #GenerativeAI
#GenAI #NotebookLM #GenerativeAI
Definitely feels like classic early social media
I open my Bluesky feed and it is full of really cool people saying really smart things and being really passionate about really interesting stuff.
I like this version of social media, and the world.
Thanks, folks.
I like this version of social media, and the world.
Thanks, folks.
November 20, 2024 at 10:57 AM
Definitely feels like classic early social media