Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

trying a TDD + git workflow npm i task-master-ai → test-driven development → github branch for each task → commit + test each subtask once done → push PR when task is feature complete → full code coverage → nearly autopilot will polish and ship as rules first then... show more

Eyal Toledano

17,911 subscribers

14,586 views • 1 year ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

11 Comments

Dennison1 year ago

@TaskmasterAI It's good, I got it working, only I think we need to solve: "Project Drift"- when the project needs to be implemented in some way not explicitly described the tasks, the tasks suddenly "Drift". - There are ton of tools, an Cursor is complaining there are too many

Eyal Toledano1 year ago

@TaskmasterAI task drift is fixable using the `update` command if things are changing as of task 6, run `tm update from=6 prompt=<describe the change>` it will update task 6 with that change and every/any task/subtask thereafter that might depend or mention task 6 solves the drift easily

Manticore Search2 years ago

🚀 Star Manticore Search on #GitHub today and join our quest for the ultimate search solution!

Fabio Vedovelli1 year ago

@TaskmasterAI Let me tell you this: yesterday was the very first time I used TM for real and I'm beyond impressed. Thanks for this tool!

Eyal Toledano1 year ago

@TaskmasterAI glad to hear it! share your experience and feedback, i wanna know more

Dennison1 year ago

@TaskmasterAI I think you can combine tools as the LLM will figure out how to use a combined tool (I think)

Eyal Toledano1 year ago

@TaskmasterAI that's why i'm focused on primitives first then you can link things up in detailed workflows

Ismaël1 year ago

@TaskmasterAI using this pattern + background agents to achieve tasks that can be achieved concurrently is the next meta to ship at blazing levels speed imo

Fackler1 year ago

@TaskmasterAI fantastic! looking forward to trying this out. 🙏

⚖️Justice⚖️1 year ago

@TaskmasterAI Wen update ?? 👊🏼

Eyal Toledano1 year ago

@TaskmasterAI they not comin out fast enough for ya? 😂

Related Videos

Ex-Google engineer built a graph-based workflow that made Claude Code do nearly 10 days of work in about an hour. Here's the exact setup he uses. step 1. create a separate Git worktree for every task. run: git worktree add ../task-1 -b task-1 open a new terminal, cd into it, and launch Claude Code. repeat this for every agent. each one gets its own branch, so nothing conflicts. step 2. give every Claude agent a different task. same repository. different branches. they all work in parallel. step 3. once they're done, review the pull requests. keep the good ones. merge what works. step 4. add a review agent. paste each PR into Claude and ask: "does this implementation match the original specification" every agent now has another agent reviewing its work before it gets merged. step 5. scale it up. 48GB of RAM can comfortably run around 20 Claude Code agents. 128GB can handle roughly 50. most engineers run Claude Code once and wait. he runs dozens of Claude Code agents in parallel, then merges the best results. this 60-minute lecture is one of the best free resources on graph engineering and multi-agent workflows. bookmark it and watch it below.

Ex-Google engineer built a graph-based workflow that made Claude Code do nearly 10 days of work in about an hour. Here's the exact setup he uses. step 1. create a separate Git worktree for every task. run: git worktree add ../task-1 -b task-1 open a new terminal, cd into it, and launch Claude Code. repeat this for every agent. each one gets its own branch, so nothing conflicts. step 2. give every Claude agent a different task. same repository. different branches. they all work in parallel. step 3. once they're done, review the pull requests. keep the good ones. merge what works. step 4. add a review agent. paste each PR into Claude and ask: "does this implementation match the original specification" every agent now has another agent reviewing its work before it gets merged. step 5. scale it up. 48GB of RAM can comfortably run around 20 Claude Code agents. 128GB can handle roughly 50. most engineers run Claude Code once and wait. he runs dozens of Claude Code agents in parallel, then merges the best results. this 60-minute lecture is one of the best free resources on graph engineering and multi-agent workflows. bookmark it and watch it below.

MIKE

32,313 views • 2 months ago

Claude Code v2.1.16 adds new task management system with full dependency tracking I demoed this through CC Mirror a few weeks ago and it was extremely powerful Watch the full video - this is going to change the way you work forever Each task can be assigned to sub agents in parallel and they can easily pick up tasks as they become unblocked

Claude Code v2.1.16 adds new task management system with full dependency tracking I demoed this through CC Mirror a few weeks ago and it was extremely powerful Watch the full video - this is going to change the way you work forever Each task can be assigned to sub agents in parallel and they can easily pick up tasks as they become unblocked

Numman Ali

51,261 views • 9 months ago

You lose customers every time an agent like Codex or Claude Code uses your product and fails to complete a task. At Scope (YC P26) we built a feature to find those errors for you and fix them. Let’s use Vercel as an example to show you how it works! Want us to run a workflow for you? Drop it in the comments below :)

You lose customers every time an agent like Codex or Claude Code uses your product and fails to complete a task. At Scope (YC P26) we built a feature to find those errors for you and fix them. Let’s use Vercel as an example to show you how it works! Want us to run a workflow for you? Drop it in the comments below :)

Lin

17,190 views • 1 month ago

Git Merge vs. Rebase vs. Squash Commit What are the differences and when should you use each? As I explain in more detail in the video, there are some key distinctions. Git Merge This creates a new commit in the target branch. The new commit ties the histories of both main and feature branches together. Git merge is non-destructive - it introduces a new commit without altering existing ones. Git Rebase Rebase transplants commits to the tip of another branch. It creates new commits for each one moved over. The benefit is linear history. But be cautious with shared branches to avoid confusing collaborators. Squash Commit Squashing condenses multiple commits into one, streamlining the commit history. Watch the video to see when to use merge vs rebase vs squash when incorporating changes between main and feature branches. – Subscribe to our weekly newsletter to get a Free System Design PDF (158 pages):

Git Merge vs. Rebase vs. Squash Commit What are the differences and when should you use each? As I explain in more detail in the video, there are some key distinctions. Git Merge This creates a new commit in the target branch. The new commit ties the histories of both main and feature branches together. Git merge is non-destructive - it introduces a new commit without altering existing ones. Git Rebase Rebase transplants commits to the tip of another branch. It creates new commits for each one moved over. The benefit is linear history. But be cautious with shared branches to avoid confusing collaborators. Squash Commit Squashing condenses multiple commits into one, streamlining the commit history. Watch the video to see when to use merge vs rebase vs squash when incorporating changes between main and feature branches. – Subscribe to our weekly newsletter to get a Free System Design PDF (158 pages):

Sahn Lam

65,234 views • 2 years ago

I managed to unlock a crazy new hidden feature in Claude Code called Swarms. You're not talking to an AI coder anymore. You're talking to a team lead. The lead doesn't write code - it plans, delegates, and synthesizes. When you approve a plan, it enters a new "delegation mode" and spawns a team of specialists who: - Share a task board with dependencies - Work in parallel as teammates - Message each other to coordinate work Workers do the heavy lifting, coordinate amongst themselves, then report back.

I managed to unlock a crazy new hidden feature in Claude Code called Swarms. You're not talking to an AI coder anymore. You're talking to a team lead. The lead doesn't write code - it plans, delegates, and synthesizes. When you approve a plan, it enters a new "delegation mode" and spawns a team of specialists who: - Share a task board with dependencies - Work in parallel as teammates - Message each other to coordinate work Workers do the heavy lifting, coordinate amongst themselves, then report back.

Mike Kelly

566,747 views • 8 months ago

how to use claude code mods like a top 1% user, step by step: give this to your agent before everyone catches on👇 1. set up Jev connect Jev to the model registry you want to use. check that the connection works and the listed models are available. 2. build your claude code mod open claude code 2.1.287 or later and paste this prompt: “build a mod called run-ledger. load plugin-authoring and use the API types for my installed version. create a dashboard that shows: - estimated cost per run, model, and source plugin where known - which model handles each task - each subagent’s status, latest action, and elapsed time include token counts, cache usage, and reported retries. count each request once. keep background tasks linked to their original run. use dated prices. label costs as API estimates, not subscription charges. show unknown when data is missing. connect the mod to my existing Jev setup. give Jev the task, available models, prices, budget, and relevant past results. ask it to recommend a model and explain why. start with recommendations. make automatic routing optional for eligible subagents. show the recommended model, actual model, result, and cost. include Jev’s own cost. keep state across hot reloads. add details and export. the dashboard itself must make no model calls. validate the plugin. test rendering, costs, attribution, and duplicate counting. give me steps for a live test.” 3. test the setup allow hot reload when prompted. run a simple task, a subagent task, and a mod-triggered model call. check that the dashboard updates, each request counts once, and Jev’s recommended model can actually run the task. 4. install the working mod ask claude to copy it out of the temporary folder and install it as a persistent plugin. see the cost. choose the model. check the result.

how to use claude code mods like a top 1% user, step by step: give this to your agent before everyone catches on👇 1. set up Jev connect Jev to the model registry you want to use. check that the connection works and the listed models are available. 2. build your claude code mod open claude code 2.1.287 or later and paste this prompt: “build a mod called run-ledger. load plugin-authoring and use the API types for my installed version. create a dashboard that shows: - estimated cost per run, model, and source plugin where known - which model handles each task - each subagent’s status, latest action, and elapsed time include token counts, cache usage, and reported retries. count each request once. keep background tasks linked to their original run. use dated prices. label costs as API estimates, not subscription charges. show unknown when data is missing. connect the mod to my existing Jev setup. give Jev the task, available models, prices, budget, and relevant past results. ask it to recommend a model and explain why. start with recommendations. make automatic routing optional for eligible subagents. show the recommended model, actual model, result, and cost. include Jev’s own cost. keep state across hot reloads. add details and export. the dashboard itself must make no model calls. validate the plugin. test rendering, costs, attribution, and duplicate counting. give me steps for a live test.” 3. test the setup allow hot reload when prompted. run a simple task, a subagent task, and a mod-triggered model call. check that the dashboard updates, each request counts once, and Jev’s recommended model can actually run the task. 4. install the working mod ask claude to copy it out of the temporary folder and install it as a persistent plugin. see the cost. choose the model. check the result.

Avid

47,276 views • 17 days ago

100,000 GMT and much more to win with #STEPNEMBER3 🎂 3 years in web3 feel like 10, but we’re here and stronger than ever 💪 For #STEPN's third anniversary, we will host a 2 weeks-long in-app event, celebrating the journey we’ve traveled together and remembering those good old memories! During this event, there will be 4 different tasks, representing our accomplishments as a thriving community: 💎 Task 1: Level up your g̶a̶m̶e̶ Gems 👟 Task 2: Grow your c̶o̶m̶m̶u̶n̶i̶t̶y̶ inventory ⚡️ Task 3: Go farther, go longer 🪄 Task 4: Enhance your experience Each task will come with its own set of prizes and by completing all of the 4 tasks above, you will enter a raffle to win a portion of the prize pool of 100,000 GMT as follows: 🥇 5 winners will get 5,000 GMT each 🥈 20 winners will get 2,000 GMT each 🥉 25 winners will get 1,000 GMT each 🍫 50 winners will get 200 GMT each 📆 The first task will start on September 16, 8 AM UTC! 🎁 We will also organise many more challenges across our social media with additional rewards, make sure to activate the notifications and join our Discord servers to participate (link in BIO) Notes: 🔹 You have to complete the 4 tasks to enter the FINAL raffle 🔹 The winners will be picked randomly among the eligible participants 🔹 You can participate in any Realms Happy #STEPNEMBER3 all 💚

100,000 GMT and much more to win with #STEPNEMBER3 🎂 3 years in web3 feel like 10, but we’re here and stronger than ever 💪 For #STEPN's third anniversary, we will host a 2 weeks-long in-app event, celebrating the journey we’ve traveled together and remembering those good old memories! During this event, there will be 4 different tasks, representing our accomplishments as a thriving community: 💎 Task 1: Level up your g̶a̶m̶e̶ Gems 👟 Task 2: Grow your c̶o̶m̶m̶u̶n̶i̶t̶y̶ inventory ⚡️ Task 3: Go farther, go longer 🪄 Task 4: Enhance your experience Each task will come with its own set of prizes and by completing all of the 4 tasks above, you will enter a raffle to win a portion of the prize pool of 100,000 GMT as follows: 🥇 5 winners will get 5,000 GMT each 🥈 20 winners will get 2,000 GMT each 🥉 25 winners will get 1,000 GMT each 🍫 50 winners will get 200 GMT each 📆 The first task will start on September 16, 8 AM UTC! 🎁 We will also organise many more challenges across our social media with additional rewards, make sure to activate the notifications and join our Discord servers to participate (link in BIO) Notes: 🔹 You have to complete the 4 tasks to enter the FINAL raffle 🔹 The winners will be picked randomly among the eligible participants 🔹 You can participate in any Realms Happy #STEPNEMBER3 all 💚

STEPN

159,265 views • 2 years ago

Unpopular opinion: VS Code is holding back AI coding. I’ve been spinning up multiple Cline instances across tmux panes and terminal tabs. Each agent has its own isolated state and can run a different task, branch, or idea in parallel. They just keep going while I focus on something else, no context collisions, no babysitting a single chat thread. You can pipe input, chain output, or run them headless in CI/CD. It fits into the workflow I already use in the terminal. Getting started literally took me 30 seconds: npm install -g cline AI is finally just another process in my workflow, instead of a panel or plugin.

Unpopular opinion: VS Code is holding back AI coding. I’ve been spinning up multiple Cline instances across tmux panes and terminal tabs. Each agent has its own isolated state and can run a different task, branch, or idea in parallel. They just keep going while I focus on something else, no context collisions, no babysitting a single chat thread. You can pipe input, chain output, or run them headless in CI/CD. It fits into the workflow I already use in the terminal. Getting started literally took me 30 seconds: npm install -g cline AI is finally just another process in my workflow, instead of a panel or plugin.

Tech with Mak

43,106 views • 8 months ago

Anthropic just dropped 1-hour workshop on loop engineering: 1 spec → the loop → 200 features from 0% to 100%: • 10% → 5:31 - Building memory for agentic loops • 30% → 8:22 - Four files for each Loop: features.json, standards.md, progress.md, • 55% → 11:02 - The Loop: check environment → run tests → pick feature → build → test → commit • 80% → 13:06 - one feature per loop. try everything at once and context dies mid-build • 100% → 33:03 - MVP in 1 hour, every next loop under 30 minutes most people build an agent that finishes a task -Anthropic ships agentic loops that can run for days without you. watch it today - then read the full loop engineering playbook below ↓

Anthropic just dropped 1-hour workshop on loop engineering: 1 spec → the loop → 200 features from 0% to 100%: • 10% → 5:31 - Building memory for agentic loops • 30% → 8:22 - Four files for each Loop: features.json, standards.md, progress.md, • 55% → 11:02 - The Loop: check environment → run tests → pick feature → build → test → commit • 80% → 13:06 - one feature per loop. try everything at once and context dies mid-build • 100% → 33:03 - MVP in 1 hour, every next loop under 30 minutes most people build an agent that finishes a task -Anthropic ships agentic loops that can run for days without you. watch it today - then read the full loop engineering playbook below ↓

Codez

18,398 views • 2 months ago

Maximizing your token throughput is one of the biggest productivity unlocks in Claude Code. So here's my guide to use multiple Claude Codes at the same time, effectively: - Run multiple Claude Code sessions at the same time, each on a different feature (Ghostty is great for this) - Use Claude Code's built in git worktrees to keep each instance isolated on its own branch (-w or --worktree flag) - Let Claude handle creating PRs or merging the worktree branches when you're done - Use the Claude Code Desktop app as your multi-agent control center Full guide below:

Maximizing your token throughput is one of the biggest productivity unlocks in Claude Code. So here's my guide to use multiple Claude Codes at the same time, effectively: - Run multiple Claude Code sessions at the same time, each on a different feature (Ghostty is great for this) - Use Claude Code's built in git worktrees to keep each instance isolated on its own branch (-w or --worktree flag) - Let Claude handle creating PRs or merging the worktree branches when you're done - Use the Claude Code Desktop app as your multi-agent control center Full guide below:

Avthar

10,754 views • 6 months ago

Instant $20 Bnb Reward distribution update bPay The $20 BNB offer is still live, but the rules just changed, Follow me now about more update Eligible user Reward distribution live tomorrow (09/27) — 8:00 PM UTC 👀 New requirement to qualify for $20 : 1. Complete both Binance Square tasks first 2. Then complete at least 2 more from TikTok / Douyin / X / Instagram task ( check full video ) Many people can’t find the new task submission on the airdrop page right now. Follow me — I’ll drop the full update as soon as it’s live. Honestly I want Everyone eligible here also get paid , iam trying full night, iam not seeing new task submitte in the page you can check yours if shown then complete it 📝Don’t wait. Old applications may not count.reapply under the new rules Sources : bPay #bPay #BNB #BinanceSquare #Airdrop #BSC Dyor.

Instant $20 Bnb Reward distribution update bPay The $20 BNB offer is still live, but the rules just changed, Follow me now about more update Eligible user Reward distribution live tomorrow (09/27) — 8:00 PM UTC 👀 New requirement to qualify for $20 : 1. Complete both Binance Square tasks first 2. Then complete at least 2 more from TikTok / Douyin / X / Instagram task ( check full video ) Many people can’t find the new task submission on the airdrop page right now. Follow me — I’ll drop the full update as soon as it’s live. Honestly I want Everyone eligible here also get paid , iam trying full night, iam not seeing new task submitte in the page you can check yours if shown then complete it 📝Don’t wait. Old applications may not count.reapply under the new rules Sources : bPay #bPay #BNB #BinanceSquare #Airdrop #BSC Dyor.

Crypto Kopakupi (Team)

16,601 views • 23 days ago

I stopped doing my weekly competitor research manually. I gave Qoder one recurring task: → Track Cursor, Windsurf & GitHub Copilot → Check official updates, X, Reddit & Product Hunt → Find the week’s important releases, news & partnerships → Turn it all into a structured report and save it automatically The best part? I set the workflow once. Qoder handles the research → analysis → report. That’s the kind of AI workflow I actually want to automate. Qwen 3.8 Flash is currently at 0.0x Credits through September 30. Enter my code 9797YB at first, then download Qoder and sign in. You will get 600 Credits: 300 from the trial and 300 from my code.

I stopped doing my weekly competitor research manually. I gave Qoder one recurring task: → Track Cursor, Windsurf & GitHub Copilot → Check official updates, X, Reddit & Product Hunt → Find the week’s important releases, news & partnerships → Turn it all into a structured report and save it automatically The best part? I set the workflow once. Qoder handles the research → analysis → report. That’s the kind of AI workflow I actually want to automate. Qwen 3.8 Flash is currently at 0.0x Credits through September 30. Enter my code 9797YB at first, then download Qoder and sign in. You will get 600 Credits: 300 from the trial and 300 from my code.

Kalsoom (ghotai )

118,841 views • 24 days ago

Full-circle Test-driven Firmware Development with OpenClaw Ladyada: "I've only had OpenClaw installed on this Raspberry Pi 5 for a couple of days, but boy, have we burned through a lot of tokens and learned a lot. Including what I think is a really fun improvement in my development process: “Agentic test-driven firmware development.” I've used LLMs for writing code as a sort of pair-programming setup, where I dictate exactly what I want done. But this is the first time that I'm giving full access to the hardware to the LLMs and letting Claude Opus 4.5 as a manager to control Codex subagents. Not only does it parse the datasheet for the register map and functionality, Claude also comes up with a full development and test plan, writes the library, tests it on existing hardware, and then also works up a test suite that covers all of the hardware registers to make sure that the library is exercising the entire chip capability. For example, here I give it an APDS-9999 color sensor and a Neopixel ring and tell it, “hey use the Neopixel ring to verify that we're really reading red, green, and blue data properly from the sensor,” and it will do the whole thing completely autonomously… no humans involved! I still review the final code and ensure the tests genuinely validate the functionality, not just take shortcuts. There is a phenomenon known as "reward hacking" (also called "specification gaming"). The model may optimize for passing tests as a metric, rather than ensuring the code truly works as intended. So far, the results have been excellent... no surprise, since these LLMs are trained on Adafruit open-source GitHub repositories!"

Full-circle Test-driven Firmware Development with OpenClaw Ladyada: "I've only had OpenClaw installed on this Raspberry Pi 5 for a couple of days, but boy, have we burned through a lot of tokens and learned a lot. Including what I think is a really fun improvement in my development process: “Agentic test-driven firmware development.” I've used LLMs for writing code as a sort of pair-programming setup, where I dictate exactly what I want done. But this is the first time that I'm giving full access to the hardware to the LLMs and letting Claude Opus 4.5 as a manager to control Codex subagents. Not only does it parse the datasheet for the register map and functionality, Claude also comes up with a full development and test plan, writes the library, tests it on existing hardware, and then also works up a test suite that covers all of the hardware registers to make sure that the library is exercising the entire chip capability. For example, here I give it an APDS-9999 color sensor and a Neopixel ring and tell it, “hey use the Neopixel ring to verify that we're really reading red, green, and blue data properly from the sensor,” and it will do the whole thing completely autonomously… no humans involved! I still review the final code and ensure the tests genuinely validate the functionality, not just take shortcuts. There is a phenomenon known as "reward hacking" (also called "specification gaming"). The model may optimize for passing tests as a metric, rather than ensuring the code truly works as intended. So far, the results have been excellent... no surprise, since these LLMs are trained on Adafruit open-source GitHub repositories!"

adafruit industries

25,532 views • 8 months ago

Anthropic dropped a free 59-minute Claude Code course From autocomplete to real agents: 00:00 - From autocomplete to agents 04:50 - How the agentic loop works 14:07 - CLAUDE.md as project memory 26:53 - Why Plan Mode comes first 33:31 - Live task from brief to commit 54:46 - Skills vs CLAUDE.md Most people still use Claude Code like a smarter autocomplete This shows how Anthropic uses it to plan, call tools, test, and ship real work Worth more than 99% of paid Claude Code tutorials Bookmark and watch it later

Anthropic dropped a free 59-minute Claude Code course From autocomplete to real agents: 00:00 - From autocomplete to agents 04:50 - How the agentic loop works 14:07 - CLAUDE.md as project memory 26:53 - Why Plan Mode comes first 33:31 - Live task from brief to commit 54:46 - Skills vs CLAUDE.md Most people still use Claude Code like a smarter autocomplete This shows how Anthropic uses it to plan, call tools, test, and ship real work Worth more than 99% of paid Claude Code tutorials Bookmark and watch it later

rari

120,767 views • 3 months ago

I gave Qoder a task instead of a prompt. And it actually did the work. It read the context, made a plan, used the tools it needed, executed the task, and verified the result before finishing. That’s the big difference with Qoder. It feels less like chatting with AI and more like handing a task to an AI agent on your computer. You describe the outcome you want. Qoder can then: • Read your workspace and context • Plan how to complete the task • Execute commands and make file changes • Use the browser or desktop when needed • Let you pause, adjust, or take over • Verify the result when it’s done And right now, Qwen3.8-Flash runs at 0.0x Credits through September 30. New users can also get 600 Credits in total. Try Qoder here: If you already use AI for coding or everyday computer work, this is worth experimenting with.

I gave Qoder a task instead of a prompt. And it actually did the work. It read the context, made a plan, used the tools it needed, executed the task, and verified the result before finishing. That’s the big difference with Qoder. It feels less like chatting with AI and more like handing a task to an AI agent on your computer. You describe the outcome you want. Qoder can then: • Read your workspace and context • Plan how to complete the task • Execute commands and make file changes • Use the browser or desktop when needed • Let you pause, adjust, or take over • Verify the result when it’s done And right now, Qwen3.8-Flash runs at 0.0x Credits through September 30. New users can also get 600 Credits in total. Try Qoder here: If you already use AI for coding or everyday computer work, this is worth experimenting with.

Shushant Lakhyani

143,473 views • 22 days ago

This is my assignment for a SF startup as a backend and infra engineer the main goal was to build a workflow engine that executes directed acyclic graphs of tasks for this assessment the task would arbitrary HTTP endpoints. But I replaced it with my own usage of news aggregator from hackernews, github (top repos for VAE, and libs related to Pytorch) and threads on reddit related to topic ML This I can use in many paid gigs too as a CD deployements like test > build > deploy staging > tests > deploy prod > notify but for now it handles simple tasks and its built on go, redis, grpc with htmx and tailwindcss

This is my assignment for a SF startup as a backend and infra engineer the main goal was to build a workflow engine that executes directed acyclic graphs of tasks for this assessment the task would arbitrary HTTP endpoints. But I replaced it with my own usage of news aggregator from hackernews, github (top repos for VAE, and libs related to Pytorch) and threads on reddit related to topic ML This I can use in many paid gigs too as a CD deployements like test > build > deploy staging > tests > deploy prod > notify but for now it handles simple tasks and its built on go, redis, grpc with htmx and tailwindcss

Mrinal

145,017 views • 8 months ago

$ANTHROPIC JUST QUIETLY SHIPPED A FEATURE THAT LETS CLAUDE SPAWN A WHOLE TEAM OF AGENTS THAT MESSAGE EACH OTHER AND REVIEW EACH OTHER'S WORK. It's a Claude Code feature called agent teams. The team lead spawns multiple agents that share a task list and message each other directly, not subagents reporting back, actual peers. In the demo a QA agent caught three bugs, sent the work back to the front-end and back-end devs, they fixed it, app shipped in one pass. How to run it: 1. Enable it. Needs Claude Code v2.1.32+. Add to settings.json: "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }. Or paste that to Claude and say "add this to my settings." Restart. 2. Prompt in plain English. Start with a goal (agents wake with zero context), then "create a team of 3 using Sonnet," describe each role, its deliverable, and who it messages when done. 3. The rules: each agent owns its own files, define exact outputs, name who talks to who, keep it to 3-5 agents. Use it for complex work with separate parts running in parallel. Skip it for simple or sequential tasks, teams cost 3-4x the tokens. Bookmark this.$

ANTHROPIC JUST QUIETLY SHIPPED A FEATURE THAT LETS CLAUDE SPAWN A WHOLE TEAM OF AGENTS THAT MESSAGE EACH OTHER AND REVIEW EACH OTHER'S WORK. It's a Claude Code feature called agent teams. The team lead spawns multiple agents that share a task list and message each other directly, not subagents reporting back, actual peers. In the demo a QA agent caught three bugs, sent the work back to the front-end and back-end devs, they fixed it, app shipped in one pass. How to run it: 1. Enable it. Needs Claude Code v2.1.32+. Add to settings.json: "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }. Or paste that to Claude and say "add this to my settings." Restart. 2. Prompt in plain English. Start with a goal (agents wake with zero context), then "create a team of 3 using Sonnet," describe each role, its deliverable, and who it messages when done. 3. The rules: each agent owns its own files, define exact outputs, name who talks to who, keep it to 3-5 agents. Use it for complex work with separate parts running in parallel. Skip it for simple or sequential tasks, teams cost 3-4x the tokens. Bookmark this.

Yarchi

463,029 views • 4 months ago