Ryan Wigley

@morganlinton @jxnlco keeping it clean-ish

@rywigs @jxnlco Thanks for sharing and yes, would love to see a screenshot too!

For people running 20+ agents with Codex, I'm curious how you're doing it in an organized(ish) way. I'm starting to find myself playing around with spinning up a lot more agents, but am having a difficult time keeping things organized. When I have 3-5 agents I can just spawn them in their own window, but I feel like I'm missing some optimization when I want to really spin up a lot of agents. What's the best way to do this in Codex? Thanks in advance for anyone who can share their workflow for this.

Design language is context engineering for taste. My site kept looking like whichever agent built it. So I codified my visual language in Figma, DESIGN.md, and an HTML guide. Then ChatGPT Voice turned a rough 5-minute talk into a sharp deck in 45 seconds. It looked like mine.

the wife and I share a ton of @ChatGPT threads back and forth about childcare, housework, random thoughts would be mighty awesome if we could do the same thing with Codex threads, like having the ability to @ tag her and it appears in her app view only permissions would be a nice starting point imo

@mil000 imagine the shredded cheese

Holy shit it’s so pretty, (256gb ram btw)

@petergyang I wonder how much damage the years of Apple trust pop ups have set this back. It’s not surprising people feel burned out and confused by permissions asked every step of the way Going to be a long hard reset of people’s boundaries online

This is the number one question I get asked by my non-AI pilled friends. Most people haven't unlocked all the amazing things that agents can do because they simply don't understand or trust how AI will use their personal info. There should be more in-product education and re-assurances to overcome this hurdle.

@thsottiaux help me insta-kill when audio is playing in random browser tabs Codex opened for some reason this is happening a lot more lately

What should we improve on Codex to improve the everyday experience? Nothing too small

@rileybrown the goods! Sick man, thanks for making this video. Will be stashing for the weekend

@Dimillian 8-12. Yes. Data pulse checks and standing long-running repeating topics (a new one is a discount finder that absolutely rocks)

How many pinned tasks do you have? Do you even use pinned tasks? What do you use them for?

@james406 such a great point and couldn't agree more with your last sentence. I am laser focused on intent data these days

people forget how early we still are. most people have chatgpt. that's it. they're not using agents. even inside engineering, where AI adoption is highest, the most sophisticated users achieve more through automation than 95% of engineers. banging out prompt->code->prompt->code isn't the future. it's systems of agents working for you, toward goals you decide. describe what you want product-wise, the system will take you there. engineers can build and maintain some of this themselves, but it's hard. we're building this system for you so you don't have to. there are lots of hard problems to solve here (e.g. mergeability, costs, pricing, etc.), but they're not impossible. (sidenote: this is why i think intent data (what a user wants) and context (product goals etc.) are going to be ultra important in the next year or so. you need both to make your product self driving)

@Dimillian I love this approach. Had agents doing something similar: told them make a tester-report website with recorded videos and annotations, all done in html at a link I can access anywhere Fun times

How good it is. Building on the go and asking Codex to record a small gameplay video so I can see the progress as the game gets built. Codex Remote.

@nurijanian There is something about the fact it’s less daunting to work with as a result Fable 5 got bad boy vibes lol

i got the ick from Fable now, due to all the scaremongering of "it's gonna cost ya!", "we're gonna take it away soon! anytime now!" despite the claims that it's "SO GOOOOOD" Anthropic definitely damaged perception with their messaging here 5.6 Sol doesn't have that vibe, and the performance differences seem negligible, so I'm using it way more often

@yoheinakajima 🎯

if you are trying to solve long-running agent problems, you will eventually start using immutable event logs

@zarazhangrui Keeping it all in your thoughts and the cloud would literally be the lightest way we could work. I would also bet there will be people working this way for years before the majority catches on

My bet on how we'll work: you don't type, you don't write documents, you talk and the work gets done in real time, from a beach if you want. The latency between talking and doing should be zero.

@adocomplete ok, great to know

@rywigs Honestly I would have either opus or fable plan and sonnet implement. I don’t think the difference between opus and fable is wide enough to designate one as planner and one as executor.

After spending ~24 hours working with Fable 5 and Opus 5 on a coding project these are my initial thoughts: - Code quality for both is excellent - Fable 5 does significantly better with open-ended prompts and requests and nails it the first time - Opus 5 can do just as well, just needs a bit more direction - Fable 5 spends more time understanding the code base and being thoughtful with implementation and pushing back when a better option exists - Opus 5 seems to prefer to do what it's told even if a better route exists - They're both really really good Claude's

Efficiency gains are everywhere. 10x this, 100x that. Focus on solving the problems you have and obsessing over your customers. Do not break your brain trying to do more for efficiency’s sake, “just because”

I disagree here. There are plenty of things we want agents to do that require using and access information you would never want hosted in the cloud, managed by a third party. I know the efficiency gains make sense on paper, but the cloud forward strategy is not a one-size-fits-all. It is frankly, not what most customers will need or want with AI. Sensitive data, pii, personal taste, daily habits / trends…. Maybe that’s something you, me, and Ryan are comfortable with in some cases but definitely not the majority of adopters going forward. m2c

If your whole workflow is one Claude Code instance on your MacBook, you're leaving 10x on the table. Here's why. On your laptop, running 2 agents at once means two copies of your code, and you're constantly fighting over which worktree you're on and whether everything's synced. It caps out fast. Cloud agents in VMs let you run as many as you want without worrying about any of that. Each one gets its own machine, so the work never collides. Spin up 10, let them all run, and nothing steps on anything else. Watch full ep below on @startupideaspod. Thanks to @ryancarson for sharing the sauce. youtube.com/watch?v=vJEy3n… Cloud agents are probably the future. So is doing half your work from your phone, which sounds insane until you try it once. Then you don't go back.

@signulll would love seeing some examples, I'm at my limits! how does it do with an established style guide?

opus 5 is absolutely insane, holy shit. visual construction + design is noticeably much much better. paired with claude design, it’s an absolute beauty. wow.

@gregisenberg Think I’m in the pending person camp

Are you an Opus 5 person or a Fable 5 person?