Ryan Wigley

automating the computer has made the computer radically more fun and its even harder to go outside now

@petergyang been too busy to try; report back!

I caved and downloaded Hermes to try. For those of you who have tried both Hermes and OpenClaw what difference do you notice? No shilling please, just want some honest opinions

@signulll the modern version of “have your people call my people”

coordinating meet ups with ppl is still so so hard. :( which ai agent does this well? like have your agent talk to my agent type stuff.

@Dimillian @steipete Well said, been thinking the same after working it with my projects and system this weekend

I've used Codex /goal all weekend on side projects/games and I can definitely feel that this is something that will be part of our toolbelt from now on and will change how we work with agent quite a bit 🧵

@thsottiaux Just let us use / for skills & features Former WoW players will love you

What are we obviously not getting right with Codex?

starting to use this as a way for agents to present plans back to me so I know we’re aligned works great for codex goals and Claude loops github.com/hurleywgly/ink

@thsottiaux could be! the bounded code execution is a clear step-change in my mind liking it a ton so far

/goal might be the most consequential thing we have shipped in codex The value of good instructions has never been higher.

@AlexFinn looks great, how was the gameplay?

Pretty incredible You have to try the new '/goal' feature in Codex It worked for over an hour and built me an entire complex extraction shooter video game You give it a goal, then it works endlessly until the goal is complete. It's like a Ralph loop. Can run for days If you enable the image gen skill before you run the goal, it will even generate ALL the assets for your game autonomously. I didn't manually create ANY of the assets you see in the video Recommendations: enable the image gen skill, put on skip all permissions, and give the prompt as much detail as you can. It will accomplish ALL of it This has to be the sickest way to build games/ long running app tasks ever

Codex’s /goal is a great model to tie together a bunch of Claude /loop skills (like Karpathy’s autoresearch) this synergy is going to make for many more effective working loops now the hard part will be what to focus on and what follow up decision to make

@petergyang great skill idea. Having it send Marie kondo quotes after clean up would be 👌👌

It's very satisfying to get Codex or Claude Code to "marie kondo" your local files and Google Drive. I give these apps full access to my computer and gws (google workspace cli), then prompt things like: "Tell me what apps load on computer bootup. Give me a plan to clean this up." "Look at my downloads folder. Give me a plan to clean up and organize it." "Help me organize my Google Drive. Let's review your plan first before doing anything." Note that I always ask it for a plan first. These are semi-dangerous operations so try them at your own risk. Anyway, my files and Drive now spark joy 🤣

It’s 8AM and codex & Claude have teamed up to plan a multi hour rework of my looping PM-family-builder operating system what a world

@danshipper exciting times no doubt

@rywigs it's definitely not good at this yet! whole new paradigm

clear that this is how we'll be doing most of our work for the next 10 years: agent running continuously on the left, application that you + the agent use on the right

@nicbstme Going to try out the snapshot technique here, like your turn based approach! I’ve found it helpful to have a few mid session checkpoint skills (I have too many multi day sessions) that make changes to varying levels of fidelity. This will mix in nicely!

@sama working pretty well on strategy debating! definitely noticing a step up in the inference & nuance.

big upgrade for codex today! try it for non-coding computer work.

really well-written piece from @clairevo No amount of lazy AI content or instant gratification short cuts the path to credibility. you gotta earn it and I’m convinced if you take too many short cuts now, you’ll be inseparable from the bots

@corbin_braun the switching costs are wildly low, wow

what are people using to build now?

@danshipper Dope. Thinking you must have a testing suite for a bunch of these role-based bench marks? I hope there’s a pelican riding a bicycle somewhere in there

i just kicked off my Senior Engineer bench on Codex's /goal feature. we'll see how well it compares to a senior engineer rewriting a slop codebase. current high score on this benchmark is 66/100 achieved by GPT-5.5 with an Opus 4.6 plan—but with an agent baby sitter to make sure it completed. let's see how far it gets!

i'm hopeful for a future where people who want to work really hard have incredibly fulfilling things to do, and people who don't want to work hard don't have to and can still have an amazing life of prosperity.

context is the leverage you have over an LLM