@zephyrarc0x Makes troubleshooting feel a lot less like guesswork.
English
Unknown™ | clipur.com
4.7K posts
Unknown™ | clipur.com
@UnknowN_Ops5
clipper | discord randon_207, RXLHZXZU
Katılım Haziran 2022
463 Takip Edilen1.3K Takipçiler
seeing why an agent picked a tool is almost as useful as the tool call itself, debugging gets a lot easier when you can inspect the selection instead of guessing, going to dig through the repo.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@idolo_Adio Lower costs without changing models is exactly the kind of optimization teams want.
English
cutting input tokens 70-85% at realistic tool catalog sizes without changing models is huge. public benchmark and code on github: ratel-ai/ratel
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@kingofzynx Better accuracy with larger toolsets is where the real leverage is.
English
The local model result is what caught my attention. going from 8% to 77% tool accuracy with 100 tools is enough to make me want to read through the benchmark and see how they measured it.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@matyCly Cutting token usage without touching the model is a solid optimization.
English
my agent once tried solving a simple task with half the tool catalog sitting in context. turns out more context isn’t always better. 70–85% fewer input tokens without changing models is worth digging into. repo’s linked in the post.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@derxvyral Public benchmarks help. I’d still like to test it on real workloads.
English
Cutting tokens 70-85% without swapping models still feels off. Their benchmarks are public though. Fits under your current stack.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
Seeing why an agent picked a specific tool is something I’ve wanted for a while. Makes debugging a lot less like guessing. Looking through the repo today: github.com/ratel-ai/ratel
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
A year ago this probably would’ve launched as a paid service with a waitlist. Instead it’s just sitting on GitHub. Going to spend some time reading through it. github.com/ratel-ai/ratel
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
The part that caught my eye wasn’t the token savings. It was local models going from 8% to 77% tool accuracy with 100 tools in their benchmark. That’s worth reading. github.com/ratel-ai/ratel
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@ratelai Been looking for something that reduces context instead of adding more. Going to read through the repo.
English
if you’re keeping an eye on $HANDL, here’s the listing. you can also claim the $10 signup bonus: mexc.com/register?utm_s…
English
if you’re keeping an eye on $HANDL, here’s the listing. you can also claim the $10 signup bonus: mexc.com/register?utm_s…
English
if you’re keeping an eye on $HANDL, here’s the listing. you can also claim the $10 signup bonus: mexc.com/register?utm_s…
English
@elonmusk One Haley Stevens clip. Thousands of reactions. The comments might be even more entertaining than the video.
x.com/clipurculture/…
CLIPUR CULTURE@ClipurCulture
Our debt and deficit problem has a name: a billionaire class that isn't paying its fair share. Haley Stevens has a plan to fix it.
English
Grok Build improves almost every day
Tech Dev Notes@techdevnotes
Grok Build now has /timeline command to Toggle the timeline sidebar
English
@techdevnotes One Haley Stevens clip. Thousands of reactions. The comments might be even more entertaining than the video.
x.com/clipurculture/…
CLIPUR CULTURE@ClipurCulture
Our debt and deficit problem has a name: a billionaire class that isn't paying its fair share. Haley Stevens has a plan to fix it.
English
Grok Build now has /timeline command to Toggle the timeline sidebar
English
@elonmusk Is Haley Stevens being judged for her message, or is everyone focused on the delivery instead?
x.com/clipurculture/…
CLIPUR CULTURE@ClipurCulture
Our debt and deficit problem has a name: a billionaire class that isn't paying its fair share. Haley Stevens has a plan to fix it.
English
@elonmusk Haley Stevens just became the latest politician to go viral for how she said it, not just what she said.
x.com/clipurculture/…
CLIPUR CULTURE@ClipurCulture
Our debt and deficit problem has a name: a billionaire class that isn't paying its fair share. Haley Stevens has a plan to fix it.
English
@elonmusk You can agree or disagree with Haley Stevens, but this clip is guaranteed to get people arguing.
x.com/clipurculture/…
CLIPUR CULTURE@ClipurCulture
Our debt and deficit problem has a name: a billionaire class that isn't paying its fair share. Haley Stevens has a plan to fix it.
English