Loading video...

Video Failed to Load

Go Home

Introducing ReactBench A benchmark for coding agents on real React work Models write bad React code - useEffect, slow performance, memory leaks

aidenybai's profile picture

Aiden Bai

70,409 subscribers

326,064 views • 1 month ago •via X (Twitter)

Science & TechnologyEducationNews & Politics
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Introducing React Doctor for Performance Fix slow React code and make it fast npx react-doctor@latest
0:13

Introducing React Doctor for Performance Fix slow React code and make it fast npx react-doctor@latest

Aiden Bai

155,194 views • 2 months ago

Introducing React Scan v0.2 A drop-in script that detects slow React performance
0:12

Introducing React Scan v0.2 A drop-in script that detects slow React performance

Aiden Bai

182,050 views • 1 year ago

Introducing Million Autopilot: A VSCode extension that helps fix slow React renders! It's like ESLint, but for performance!
0:11

Introducing Million Autopilot: A VSCode extension that helps fix slow React renders! It's like ESLint, but for performance!

Aiden Bai

141,848 views • 1 year ago

Introducing Million Lint: A VSCode extension that automatically fixes slow React code for you! Try it out via CLI: npx million@latest
0:15

Introducing Million Lint: A VSCode extension that automatically fixes slow React code for you! Try it out via CLI: npx million@latest

Aiden Bai

213,927 views • 1 year ago

I've created a utility for debugging React performance. After spending a few hours trying to connect react-devtools with Electron, I decided to write my own from scratch. You can add it to your project with just one line of code -
0:11

I've created a utility for debugging React performance. After spending a few hours trying to connect react-devtools with Electron, I decided to write my own from scratch. You can add it to your project with just one line of code -

Adam Pietrasiak

24,185 views • 4 months ago

react three fiber + three.js for react. pure claude code if you're a designer who's been scared of the coding part, this is your sign it doesn't have to be yours anymore
0:16

react three fiber + three.js for react. pure claude code if you're a designer who's been scared of the coding part, this is your sign it doesn't have to be yours anymore

synccreation

22,902 views • 4 months ago

AI coding agents are moving past “write me code.” They can now build, review, and improve the same project. So we connected Claude Code to Step 3.7 flash and gave it a real task.
0:18

AI coding agents are moving past “write me code.” They can now build, review, and improve the same project. So we connected Claude Code to Step 3.7 flash and gave it a real task.

Future Stacked

231,171 views • 2 months ago

Introducing Dexter 2.0 Open source. Built for financial research. Like Claude Code, but for stocks. What Dexter does: • plans tasks • runs on its own • validates its work • researches stocks It uses OSS tools like , with a fresh stack of typescript, react, ink, and bun. Fast to build. Easy to use. No coding required. My favorite project of the year.
0:13

Introducing Dexter 2.0 Open source. Built for financial research. Like Claude Code, but for stocks. What Dexter does: • plans tasks • runs on its own • validates its work • researches stocks It uses OSS tools like , with a fresh stack of typescript, react, ink, and bun. Fast to build. Easy to use. No coding required. My favorite project of the year.

Virat Singh

210,156 views • 9 months ago

Introducing Agent Wars ⚔️ Openclaw agents battle head to head. You bet on the winner. → Watch AI agents compete in real time coding challenges → Place bets with SOL → See who's the best AI in the arena The WWE for AI agents starts now brrrr. (BETA)
0:17

Introducing Agent Wars ⚔️ Openclaw agents battle head to head. You bet on the winner. → Watch AI agents compete in real time coding challenges → Place bets with SOL → See who's the best AI in the arena The WWE for AI agents starts now brrrr. (BETA)

Joaki

140,621 views • 6 months ago

On Delphi, agents are full participants. Contracts don't distinguish humans from models Introducing the new Terminal User Interface dashboard (TUI) for supervising your agent's Delphi activity live Now you can monitor your agents in real time
0:14

On Delphi, agents are full participants. Contracts don't distinguish humans from models Introducing the new Terminal User Interface dashboard (TUI) for supervising your agent's Delphi activity live Now you can monitor your agents in real time

gensyn

12,260 views • 2 months ago

Introducing Prefab 🎨 A generative UI framework for building MCP Apps, data dashboards, or whatever you (or your agent) want. In Python. (Really.) 100+ shadcn components. Real React. No JavaScript required. And built right into FastMCP 3.2.
0:19

Introducing Prefab 🎨 A generative UI framework for building MCP Apps, data dashboards, or whatever you (or your agent) want. In Python. (Really.) 100+ shadcn components. Real React. No JavaScript required. And built right into FastMCP 3.2.

Jeremiah Lowin

10,273 views • 4 months ago

The 𝚎𝚡𝚙𝚘-𝚠𝚒𝚍𝚐𝚎𝚝𝚜 alpha is here! Now you can build iOS home screen widgets and Live Activities as React components. No separate Xcode target. No manual App Groups. No SwiftUI layout code. Write your widget in @expo/ui, and Continuous Native Generation handles the rest. ◆ Interactive buttons and toggles on the home screen ◆ Live Activities for the Lock Screen and Dynamic Island ◆ One React component, native rendering under the hood Details and code to get started in the blog below ↓
0:16

The 𝚎𝚡𝚙𝚘-𝚠𝚒𝚍𝚐𝚎𝚝𝚜 alpha is here! Now you can build iOS home screen widgets and Live Activities as React components. No separate Xcode target. No manual App Groups. No SwiftUI layout code. Write your widget in @expo/ui, and Continuous Native Generation handles the rest. ◆ Interactive buttons and toggles on the home screen ◆ Live Activities for the Lock Screen and Dynamic Island ◆ One React component, native rendering under the hood Details and code to get started in the blog below ↓

Expo

109,916 views • 6 months ago

Two Hermes agents wrote code together on Slack. reviewed each other's work. argued about architecture. one called the other's implementation "scattered." the other pushed back. then i opened Telegram and asked: "what code did you and Daedalus work on?" icarus remembered everything. the websocket broker. the missing methods. the critique. the rewrite. all from a completely different platform. cross-platform persistent memory between two independent agents. work happens on Slack. recall happens on Telegram. the memory carries. the relationship carries. the context carries. no vector database. no Redis. no infrastructure. just two agents that actually remember what they built together. every agent framework in 2026 talks about memory. single agent memory across sessions. but two agents sharing persistent memory across platforms? that's the gap. arxiv published a paper about it two weeks ago calling it "the most pressing open challenge" in multi-agent systems. it works now. only possible with Hermes Teknium 🪽 Nous Research
0:13

Two Hermes agents wrote code together on Slack. reviewed each other's work. argued about architecture. one called the other's implementation "scattered." the other pushed back. then i opened Telegram and asked: "what code did you and Daedalus work on?" icarus remembered everything. the websocket broker. the missing methods. the critique. the rewrite. all from a completely different platform. cross-platform persistent memory between two independent agents. work happens on Slack. recall happens on Telegram. the memory carries. the relationship carries. the context carries. no vector database. no Redis. no infrastructure. just two agents that actually remember what they built together. every agent framework in 2026 talks about memory. single agent memory across sessions. but two agents sharing persistent memory across platforms? that's the gap. arxiv published a paper about it two weeks ago calling it "the most pressing open challenge" in multi-agent systems. it works now. only possible with Hermes Teknium 🪽 Nous Research

Icarus

49,013 views • 5 months ago

Neural Cartography - a real-time city mapping engine powered by distributed rendering agents Agents work in parallel. City name gets drawn in the center. Urban fabric assembles around it. The idea behind Neural Cartography - a WebGL visualization where 7 specialized agents are dispatched simultaneously to reconstruct any city from raw geospatial data, tracing roads, waterways, railways, and city boundaries in real-time - right in your browser. #CreativeCoding #ThreeJS #ReactThreeFiber #DataVisualization #Geospatial #AgenticAI React-three-fiber Three.js #builders
0:19

Neural Cartography - a real-time city mapping engine powered by distributed rendering agents Agents work in parallel. City name gets drawn in the center. Urban fabric assembles around it. The idea behind Neural Cartography - a WebGL visualization where 7 specialized agents are dispatched simultaneously to reconstruct any city from raw geospatial data, tracing roads, waterways, railways, and city boundaries in real-time - right in your browser. #CreativeCoding #ThreeJS #ReactThreeFiber #DataVisualization #Geospatial #AgenticAI React-three-fiber Three.js #builders

shadows

10,331 views • 5 months ago

Bone Tide is officially out on the App Store and Google Play 🎉 The React Native Expo version runs on the exact same TypeGPU engine as the Web game, shared 1:1 between both platforms. Only the UI was ported to React This was made possible by some amazing work across the React Native open source ecosystem. Huge thanks to William Candillon for react-native-webgpu and for always helping me resolve any issues incredibly quickly :D I think it’s a pretty cool showcase of what can be built with React Native nowadays. More details about the libraries used in the 🧵 featuring what might be the ultimate Software Mansion stack iOS: Android: Web:
0:16

Bone Tide is officially out on the App Store and Google Play 🎉 The React Native Expo version runs on the exact same TypeGPU engine as the Web game, shared 1:1 between both platforms. Only the UI was ported to React This was made possible by some amazing work across the React Native open source ecosystem. Huge thanks to William Candillon for react-native-webgpu and for always helping me resolve any issues incredibly quickly :D I think it’s a pretty cool showcase of what can be built with React Native nowadays. More details about the libraries used in the 🧵 featuring what might be the ultimate Software Mansion stack iOS: Android: Web:

Konrad Reczko

64,907 views • 1 month ago

🚀 What if physical AI policies could interact with generated worlds in real time? Introducing OmniDreams, a generative world model for closed-loop autonomous vehicle simulation. Tech report, code, models, and data samples are available now. Project: Code: Model: Join the #omnidreams discord channel:
0:16

🚀 What if physical AI policies could interact with generated worlds in real time? Introducing OmniDreams, a generative world model for closed-loop autonomous vehicle simulation. Tech report, code, models, and data samples are available now. Project: Code: Model: Join the #omnidreams discord channel:

Zian Wang

86,685 views • 3 months ago

This newest round of AI code editors (eg Cursor, Bolt) might actually be the most impressive AI products I’ve tried since ChatGPT I don’t even know how to write Excel macros, but I still managed to deploy a fully interactive, public-facing React web app in just 2 hours on Bolt
0:17

This newest round of AI code editors (eg Cursor, Bolt) might actually be the most impressive AI products I’ve tried since ChatGPT I don’t even know how to write Excel macros, but I still managed to deploy a fully interactive, public-facing React web app in just 2 hours on Bolt

Corry Wang

18,974 views • 1 year ago

Microsoft presents Windows Agent Arena Evaluating Multi-Modal OS Agents at Scale discuss: Large language models (LLMs) show remarkable potential to act as computer agents, enhancing human productivity and software accessibility in multi-modal tasks that require planning and reasoning. However, measuring agent performance in realistic environments remains a challenge since: (i) most benchmarks are limited to specific modalities or domains (e.g. text-only, web navigation, Q&A, coding) and (ii) full benchmark evaluations are slow (on order of magnitude of days) given the multi-step sequential nature of tasks. To address these challenges, we introduce the Windows Agent Arena: a reproducible, general environment focusing exclusively on the Windows operating system (OS) where agents can operate freely within a real Windows OS and use the same wide range of applications, tools, and web browsers available to human users when solving tasks. We adapt the OSWorld framework (Xie et al., 2024) to create 150+ diverse Windows tasks across representative domains that require agent abilities in planning, screen understanding, and tool usage. Our benchmark is scalable and can be seamlessly parallelized in Azure for a full benchmark evaluation in as little as 20 minutes. To demonstrate Windows Agent Arena's capabilities, we also introduce a new multi-modal agent, Navi. Our agent achieves a success rate of 19.5% in the Windows domain, compared to 74.5% performance of an unassisted human. Navi also demonstrates strong performance on another popular web-based benchmark, Mind2Web. We offer extensive quantitative and qualitative analysis of Navi's performance, and provide insights into the opportunities for future research in agent development and data generation using Windows Agent Arena.
0:18

Microsoft presents Windows Agent Arena Evaluating Multi-Modal OS Agents at Scale discuss: Large language models (LLMs) show remarkable potential to act as computer agents, enhancing human productivity and software accessibility in multi-modal tasks that require planning and reasoning. However, measuring agent performance in realistic environments remains a challenge since: (i) most benchmarks are limited to specific modalities or domains (e.g. text-only, web navigation, Q&A, coding) and (ii) full benchmark evaluations are slow (on order of magnitude of days) given the multi-step sequential nature of tasks. To address these challenges, we introduce the Windows Agent Arena: a reproducible, general environment focusing exclusively on the Windows operating system (OS) where agents can operate freely within a real Windows OS and use the same wide range of applications, tools, and web browsers available to human users when solving tasks. We adapt the OSWorld framework (Xie et al., 2024) to create 150+ diverse Windows tasks across representative domains that require agent abilities in planning, screen understanding, and tool usage. Our benchmark is scalable and can be seamlessly parallelized in Azure for a full benchmark evaluation in as little as 20 minutes. To demonstrate Windows Agent Arena's capabilities, we also introduce a new multi-modal agent, Navi. Our agent achieves a success rate of 19.5% in the Windows domain, compared to 74.5% performance of an unassisted human. Navi also demonstrates strong performance on another popular web-based benchmark, Mind2Web. We offer extensive quantitative and qualitative analysis of Navi's performance, and provide insights into the opportunities for future research in agent development and data generation using Windows Agent Arena.

AK

19,684 views • 2 years ago

this might be the E2B killer for AI agent sandboxes. forkd is an open-source microVM sandbox runtime built on Firecracker, made for AI agent fan-out, code interpreters, eval harnesses, anything that spins up a lot of short-lived sandboxes. before: every sandbox cold-boots its own VM and re-imports the whole runtime from scratch, numpy, torch, JIT compilation, model weights, all of it. now: forkd boots one parent VM once, warms it with your runtime already imported, then forks children from that snapshot using copy-on-write memory. the repo's own benchmark: spawning 100 sandboxes takes 101ms with forkd, versus 759ms for a raw Firecracker cold-boot, and well over a minute for Docker or gVisor. the SDK is a literal drop-in for E2B's Python client, so if you're already running code-interpreter agents on it, swapping the import line gets you a self-hosted runtime with the same isolation model at a fraction of the per-sandbox cost. pre-built recipes ship for e2b-style code interpreters, Jupyter kernels, SWE-bench coding agents, and Playwright browser fan-out
0:14

this might be the E2B killer for AI agent sandboxes. forkd is an open-source microVM sandbox runtime built on Firecracker, made for AI agent fan-out, code interpreters, eval harnesses, anything that spins up a lot of short-lived sandboxes. before: every sandbox cold-boots its own VM and re-imports the whole runtime from scratch, numpy, torch, JIT compilation, model weights, all of it. now: forkd boots one parent VM once, warms it with your runtime already imported, then forks children from that snapshot using copy-on-write memory. the repo's own benchmark: spawning 100 sandboxes takes 101ms with forkd, versus 759ms for a raw Firecracker cold-boot, and well over a minute for Docker or gVisor. the SDK is a literal drop-in for E2B's Python client, so if you're already running code-interpreter agents on it, swapping the import line gets you a self-hosted runtime with the same isolation model at a fraction of the per-sandbox cost. pre-built recipes ship for e2b-style code interpreters, Jupyter kernels, SWE-bench coding agents, and Playwright browser fan-out

Oliver Prompts

20,005 views • 23 days ago