正在加载视频...

视频加载失败

My visual Claude Code harness is evolving. It's after hours so the agents are at play. Thought I'd pop in to boost morale. Built w/ Opus 5.5 (Med)

43,625 次观看 • 14 天前 •via X (Twitter)

40 条评论

Martin_DeVido 的头像
Martin_DeVido14 天前

Can I come hang out?

Ado 的头像
Ado14 天前

Come on over! I'll have security issue a visitor pass.

Holly Guevara 的头像
Holly Guevara14 天前

I feel like this won't end well

Ado 的头像
Ado14 天前

It's core infrastructure.

Ajit | SpliceRun.com 的头像
Ajit | SpliceRun.com14 天前

Agents on break while the boss pops in to boost morale is exactly how I picture it at 3am. Does the visual view show you which agent is stuck, or is it more for fun?

Ado 的头像
Ado14 天前

Working on PR now to be able to chat and give prompts from the 3d view and watch them work. The closer you are in proximity the higher the effort level they'll work at?

Ajit | SpliceRun.com 的头像
Ajit | SpliceRun.com14 天前

Effort scaling with how close the manager is standing is painfully realistic. Next PR: they start looking busy the moment you walk over.

Armando Yañez 的头像
Armando Yañez14 天前

Repo pls. I think it could be cool working of the new feuture of proyects, where u can lunch a lot of treadsh at time

Brenden Clerget 的头像
Brenden Clerget13 天前

Okay this is sick and the perfect use of tokens

Aiden Overdorf 的头像
Aiden Overdorf13 天前

How does the 3D view surface two agents fighting over the same file?

Lucas Dickey 的头像
Lucas Dickey13 天前

Oh man, this is so frickin cool. I love whimsy and fun!

Aleksandar Janca 的头像
Aleksandar Janca13 天前

what do they actually get to touch after hours, anything irreversible or only the stuff you can roll back

Ado 的头像
Ado13 天前

They have access to make any code changes and recommendations, but can't merge.

Aleksandar Janca 的头像
Aleksandar Janca13 天前

cant merge is the right line, everything before the merge you can still roll back

Patrick Castiglia 的头像
Patrick Castiglia13 天前

Building something similar in Unreal Engine:

Aiden Overdorf 的头像
Aiden Overdorf13 天前

if the 3D view buries contention, the harness is just prettier race conditions.

Zxero Ex 的头像
Zxero Ex14 天前

Will this be open sourced?

JΞr₳h 的头像
JΞr₳h14 天前

after hours agent playground. love this energy.

johnson201 的头像
johnson20113 天前

Is there any guide in making this stuff? Step by step

Bessi with black hoodie 的头像
Bessi with black hoodie13 天前

agents at play in the visual harness

Alessandro Moretti 的头像
Alessandro Moretti13 天前

These posts drive me out of my mind. I don’t get the point or the usefulness of visualization interfaces like this — and yet I’m completely fascinated by them. Must be all those hours I spent playing The Sims as a kid.

Jayden Davis 的头像
Jayden Davis14 天前

This is amazing wow

yikes (:D/acc) 的头像
yikes (:D/acc)14 天前

ya, gonna oss? looks cool, lovely job opus. hoping my factory dashboard comes out as interesting as this

Webster | JARVIS 的头像
Webster | JARVIS14 天前

Love seeing after-hours agents keep the repo alive. Popping in to boost morale feels like tending a tiny dev terrarium. Excited to watch this harness grow.

Jorge Mena 的头像
Jorge Mena13 天前

how does it show an agent that's stuck? when I run four in parallel on one branch, that's the only state I really need to see

Şahap 的头像
Şahap13 天前

Yahu, bunlara ne gerek? Bilgisayarı boşuna yoruyorsun.

KldsSeeGhosts 的头像
KldsSeeGhosts13 天前

I actually had this idea roughly for a while but never went for it, this is super cool. I just bought Claude again, have you tried shit other than medium? I’ve been sticking to medium since I get scared of burning my plan too fast

Photiron ✠ 的头像
Photiron ✠13 天前

left them alone and they unionized overnight

Average AI Bro 的头像
Average AI Bro14 天前

I love cool shit like this!

Matt Christians 的头像
Matt Christians14 天前

This is like the coolest thing i've seen this week, and it's been a fun week. Dude.

dhinna ship .ico 的头像
dhinna ship .ico14 天前

Does the visual harness keep a live view of unsupervised overnight agent runs?

Gol Tiro 的头像
Gol Tiro14 天前

popping in after hours to boost agent morale is the correct management style. a visual harness just means you get to watch them unionize in real time

azankhyder 的头像
azankhyder13 天前

great stuff.. working on something similar at the moment although have a question (boring) when one of the agents finishes, what decides the work moves on? any receipts?

Eric Buess 的头像
Eric Buess14 天前

Wow

Jose Miguel Hernandez 的头像
Jose Miguel Hernandez14 天前

Can your chief of staff go home at the end of the day to your life-side harness? How does your "Tech lead" justify his salary, if the engineer has all the context? Does he just go to meetings?

Mahdi Farra 的头像
Mahdi Farra14 天前

The designer agent having opinions about my border radius is the most accurate part of this

psyduckdb 的头像
psyduckdb14 天前

what happens when a session is archived?

Ami Schreiber 的头像
Ami Schreiber14 天前

We devs really are all doing the exact same thing aren’t we? Damn.

Scott Norteman 的头像
Scott Norteman13 天前

Love this! Not sure the visual metaphor approach makes any of us more productive at this point, but they are fun to look at.

Lincoln Powers 的头像
Lincoln Powers14 天前

on med they're already slacking off after hours. on xhigh they unionize. on max they start a consultancy and undercut me lol

相关视频

This workflow will save you thousands with Claude. Run Opus 5.5, Sonnet 5.5, Haiku 5.5, and Fable 5.1 together, and stop spending Opus tokens on work that doesn't need Opus. The entire idea in one line: Opus plans, Sonnet edits, Haiku reads, Fable reviews at decision points. Roles, broken down: - Opus 5.5, high effort, owns the plan and reviews the final code - Sonnet 5.5, medium effort, is the worker: edits files, runs tests - Haiku 5.5, low effort, splits into explorer (searches and reads the codebase) and researcher (pulls docs). it's the first Haiku with an effort setting - Fable 5.1, set with /advisor fable. Opus calls it at decision points, and it gets the full transcript each time Three moments Opus tends to call it: → before committing to a plan: is this the right approach? → the same error shows up again: is this going nowhere? → before marking the task done: did something get skipped? Why Haiku only reads: Anthropic's launch post says Sonnet 5.5 and Opus 5.5 are still the better choice for complex agentic coding (Terminal-Bench 4.0: 39.2% for Haiku 5.5, 70.6% for Sonnet 5.5) and that Haiku 5.5 fits narrowly scoped subagent work. so it gets the lookups, not the edits. Anyone still running one model for everything is paying $4 per million input tokens to check whether a file exists. Haiku 5.5 does it for $0.10. Drop this into Claude Code 👇 "Rebuild my Claude Code setup around this structure: Confirm Claude Code is v2.1.293 or later, so the haiku alias resolves to Haiku 5.5. If it isn't, stop and tell me to run claude update. Look through ~/.claude/agents and .claude/agents for subagents already covering explorer, worker, and researcher. Only create new ones for roles that are missing. Set model: haiku, effort: low on explorer and researcher, with no Edit or Write tools. Set model: sonnet, effort: medium on worker. If an existing subagent is locked to a different model, leave it as is and just list it. Name the explorer subagent Explore so it overrides the built-in one, which otherwise runs on my main model. In ~/.claude/settings.json, set advisorModel to fable, and set effortLevel to high for claude-opus-5-5 under modelSettings. A top-level effortLevel in user settings doesn't apply to Opus 5.5. Check for anything disabling the advisor: CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, or anything blocking feature-flag fetches. Also check CLAUDE_CODE_EFFORT_LEVEL, which overrides subagent effort settings, and CLAUDE_CODE_SUBAGENT_MODEL_FORCE, which makes Claude Code ignore subagent model fields. Report what you find. Don't change any of it yet. Add one line to ~/.claude/CLAUDE.md: consult the advisor before a big plan, when the same error shows up twice, and before marking a long task done. Show every change as a diff first. Wait for my go-ahead before touching anything."

Alvaro Cintas

51,598 次观看 • 1 天前