- プラットフォーム
- YouTube
- 長さ
- 3:24
- 話されている言語
- アメリカ英語
- 再生数
- 800
- 顔の映る割合
- 89%
- 暗いフレーム
- 0%
Your agent skills should improve themselves の文字起こし
ブランドが書いたものではなく、映像の音声から音声モデルが書き起こしたものです。行をクリックすると、その瞬間に移動します。
- 0:00Let's talk about self-improvement loops for agents. The concept is pretty simple. We have a number of
- 0:05conversations with our agents every day across our team, and there are a number of failure points in
- 0:10those conversations where agents may hold the skills wrong or write inefficient code that we
- 0:15have to go revise, and it would be nice to have a system that's evaluating those conversations as
- 0:20they happen and suggesting improvements to our skills so that the agent performs better next time.
- 0:25To address this, we've created a three-step system that works off of all of your conversation data.
- 0:31It starts from aggregating all of your raw conversations, turning those into scored conversations,
- 0:37so defining rubrics of what success and failure look like, and then having an agent review all of those scores to suggest improvements.
- 0:45Here I'll show you what that looks like implemented as part of a larger software factory.
- 0:50This dashboard is part of Warp Factories, which you can get early access to right now.
- 0:55And you can see here we've set up a number of scorers that we run across all the conversations
- 1:00across our team.
- 1:01Here we've created a code quality agent that reviews all the conversations we have working
- 1:05on our main repositories, and it evaluates whether it has passed or failed our code quality
- 1:11test.
- 1:12And we can see the configuration for that here, where whenever an agent is implementing
- 1:17code on our repositories, we've given it some instructions on how to evaluate the performance.
- 1:22This code quality agent is going to assess the design and correctness of the code that
- 1:26agents write, and it's also going to look into the diffs and documentation to suggest
- 1:32improvements.
- 1:32And here we've set the model that's going to judge all these conversations, along with
- 1:36classifications, which are basically a rubric from zero to one where you can define how
- 1:42the agent should judge that conversation.
- 1:44For code quality, it's fairly simple.
- 1:46It's going to be approved if the code is correct and consistent with our conventions.
- 1:51If there's insufficient evidence, it should bail out without explicitly failing, and it
- 1:56should block that PR if it finds evidence that it wasn't up to our quality standards.
- 2:01This agent will run on an automation every time an agent tries to open a PR, and it will
- 2:06assess the pass or failure over time.
- 2:08Here as you can see, we've been fine-tuning the success and failure criteria to get the
- 2:12best quality output.
- 2:14And then once we have agents grading these conversations, we can have another agent review
- 2:18all of the runs that failed our quality test and suggest improvements to our skills so they don't
- 2:24fail next time. And we can see samples of all of those improvement runs right here in our
- 2:28self-improvement dashboard. This is also featured as baked into warp factories. We can see here
- 2:34that a self-improvement agent reviewed all of those scored conversations and made some PRs to
- 2:40actually improve our skills. In this example, our code review agent wasn't fetching all of the
- 2:44necessary PR feedback before writing a review.
- 2:48And down here, we can see direct references
- 2:50to the failing scores that prompted this PR.
- 2:53Here, it was prompted by our efficiency and task compliance
- 2:57scores, which are other rubrics that we have set up,
- 3:00where there was lower compliance than we would have liked.
- 3:03And here, we can see the changes that it suggested.
- 3:05Since our factory is configured as code,
- 3:07it's easy for agents to reach into the configuration
- 3:10and suggest ways to improve skills.
- 3:12And with that, we have a closed loop system.
- 3:15Bring tasks into the factory and let the code quality improve on its own.
- 3:20If that sounds interesting to you, check out Warp Factories in Early Access.
50行 · 635語
More from Warp
All of Warp →The rest of their work we have taken apart.
Launch23K viewsWarp
Introducing the Warp Agent CLI: a coding agent that does what others can't
2:17101 frames8.1s cut72% face
4K viewsWarp
Orchestrate a team of agents in Warp
2:58107 frames0% face
4K viewsWarp
We fixed Claude Code's biggest problem...
4:20126 frames100% face
540 viewsWarp
How to orchestrate a team of agents in the cloud with Warp
2:0593 frames0% face
506 viewsWarp
How to hand off local agents to the cloud in Warp
3:06110 frames26.6s cut0% face
Lit and framed like this
Closest on how much of the running time has a face in it and how dark the frames are.
Launch72K viewsHaz Hubble
introducing... a brand new pally 👋 pally is an ai assistant that liv...
1:5844 frames36.7s cut89% face
Launch99K viewsRunway
Introducing Project Luxo: a new initiative exploring how AI-generated...
10:02814 frames2.2s cut90% face
65K viewsOpenAI
Make information visual with ChatGPT
1:2645 frames5.6s cut89% face
392K viewsOpenAI Developers
$visualize
1:2645 frames5.6s cut87% face
6K viewsOpenAI
What Codex Unlocks for loveholidays
3:32179 frames2.9s cut85% face

Cloudflare
We're literally building our own network: Cloudflare CEO
4:5785 frames26.9s cut91% face
21K viewsGoogle Workspace
How to create interactive dashboards and apps in Google Sheets
2:34110 frames8.2s cut92% face
236K viewsjules
5 things @FarzaTV does differently
15:421286 frames2.4s cut92% face
45K viewsconor brennan-burke
we just launched agent-to-agent hiring at @hyperspell we're hiring ...
1:3437 frames17.3s cut92% face
5.4M viewsNothing
NOTHING (MILLI)
0:3313 frames92% face
91K viewsNotion
Agency is a muscle. Tools can strengthen it — or let it atrophy. @...
1:0642 frames3.0s cut93% face
38K viewsFarza
ty for the love on this. insane... i wanna keep telling the story. i...
2:16180 frames8.4s cut92% face
More social
Same category, other makers, most watched first.
622M viewsBase44
Your ideas, your apps—Base44
0:5651 frames1.1s cut6% face
538M viewsBase44
Turn any idea into an app with Base44
0:5538 frames2.2s cut50% face
154M viewsBase44
Build Your Website with One Sentence | Base44
0:4835 frames0.9s cut31% face
81M viewsNike Football
Rip The Script | Nike Football
6:00310 frames1.1s cut60% face
75M viewsPetSmart
PetSmart | "Pickup in...STORE!"
0:1511 frames1.4s cut64% face
69M views1X
NEO The Home Robot Order Today
9:53271 frames2.8s cut25% face
69M viewsApple
Apple Education: Ready for every learning opportunity
1:4687 frames1.3s cut40% face
64M viewsBland
Phone calls suck. So, we bought a giant telephone car, raised $65,00...
1:4854 frames4.3s cut81% face
39M viewsFigma
Oh wait there’s more: we’re building 3D Transforms, here’s a preview ...
1:0423 frames4.0s cut78% face
36M viewshims
Getting Help is Simple, With Hims
0:3023 frames2.0s cut70% face
36M viewsFigma
Made in Figma (2025) Like this ❤️ we designed it for you
0:3872 frames0.5s cut7% face
35M viewsApple
Hello, MacBook Neo
3:50291 frames67.5s cut1% face









































