Nick Gwood

@sama I'm like 30 of those users...

To celebrate 3 million weekly codex users, we are resetting usage limits. We will do this every million users up to 10 million. Happy building!

I just gave the article to my Openclaw and told it to figure it out. After a few minutes it told me it had the path and said it needed to update my openclaw.json. So I said "Go ahead." Next thing I knew it was running Mythos. No idea what it actually did to make it work (other than it was with API), but it did it. It's amazing. In just 15 minutes with Mythos I was completely bankrupt and one prompt away from world domination. Anyone have a spare $1,000 to finish it up? I'll give you your own country when I'm done...

So, who's running Claude Mythos with their OpenClaw? any tips?

@quionie Claude is far better, but Codex SOOOOOOOOOOO much cheaper. Still weighing if it’s worth the loss in productivity, because it’s large

everyone who recently switched from claude to codex drop your reasons

@MilksandMatcha I miss my Claude, can you help me out with that?

Giving away 5 Codex Pro plans Each person will get 3 months of free Codex Pro (highest tier). Winners will be selected from comments in 48 hours, comment below why you want it.

@Shpigford Early on there were soooo many projects. I found that over time it became clear which were going to bring value and were worth the time and effort, which narrowed down which ones actually got attention and which got deleted or moved to a waiting list

with AI, do you find yourself juggling more work across *disparate* projects? or just increased velocity on the *same* project?

@floriandarroman This is the best of social media today! I lost it on that last message!

Just changed OpenCLaw from Opus 4.6 to GPT-5.4. I puked. Then I went back to Opus.

@MatthewBerman Must be why Facebook still feels like 2004

> On flight > See guy in Meta Eng Hoodie > Walk by, he’s coding > Literally writing code by hand > Walk back to my seat > Liquidate my Meta position

Here's an idea: A way to track where your token spend is happening. People are doing it on their own for openclaw and similar tools, why cant this be a built in feature for Claude? Especially with all the new releases that came out that let users burn through usage so fast. Might save a whole lot of these problems if they can actually see what those features are really using. This type of transparency could go a long way at winning loyal customers over.

I want to do a few more of these calls. If your MAX 20x plan ran out of tokens unexpectedly early and you're willing to screenshare and run some prompts through Claude Code please comment. Trying to figure out how we can improve /usage to give more info.

@hi_antz @Shpigford It feels like GPT just ignores the soul. Everything I change in it does nothing. Except that line about swearing. It loves to swear now so that line had to go. It was a bit ridiculous

@Nixtrodamis @Shpigford Yeah i feel like hasn’t changed much for me either, even after tinkering with the Soul.

gonna try GPT 5.4 in 🦞 tomorrow given its new update aimed at making it not the most unpleasant LLM i'm ever chatted with. what's the official word on setting the /thinking level for it? does it need changing? what should it be set to as a daily driver?

@aleks_todo @steipete @DannyIO @danshipper I’m testing the beta and stable. Both suck- all thinking levels. Is there a setting to activate the changes I’m missing? I did the SOUL.md adjustments ages ago, unless there are more I need to do. But it seems like GPT simply ignores it and goes stupid regardless

@Nixtrodamis @steipete @DannyIO @danshipper ur on 4.5 stable, and are using gpt 5.4 with med. thinking?? no way ur not noticing a diff, oc is in best shape rn

if you're seeing this, make sure you turn thinking on claws are actually pretty great in gpt-5.4, but they seem really stupid unless this gets turned on

@aleks_todo @steipete @DannyIO @danshipper Did you need to do anything after updating to make it better? Because I see no difference.

@steipete @DannyIO @danshipper like butter! hahaha diffrence?? 1h ago i was about to jump out my window. Now its waaaaaaaaaaaaaaay better. GOD MODE

@_____B0B_____ @mhp_guy Is that what the problem is? Maybe they burn out because I'm a liar? I don't understand the connection, but I'll give it a shot. Thanks!

@Nixtrodamis @mhp_guy You either have other issues at your house or you are blatantly lying. I'm leaning towards the latter.

Biggest marketing scam ever invented? “Lasts 13 years!” A few years later, it dies. “Where’d we get those bulbs from again? Whatever. Let’s just buy more.” A while back, I went through and replaced every incandescent in our house with LEDs. 7 years later, here we are.

@Shpigford @gthartley Yup, GPT straight up ignores it

@gthartley yeah, had already run/used that with opus. worked beautifully. hasn't made much difference with gpt.

Who’s written How To Transition OpenClaw from Claude to GPT While Maintaining Vibes? Because I need that.

When it was on Opus I was using it for everything short of actual coding. The ideating process was amazing, and I made sure it was critical enough to tell me when ideas were stupid. After switching to GPT, I hardly chat with it anymore. It ignores the soul.md, writes a complete novel for every response, gives 16 versions of the same answer to a simple question. It seems fast at starting a response, but because they are so long it still takes forever. It’s just annoying talking to it now. The wild thing though is that 5.4 is infinitely better than 5.3, which didn’t do anything I asked it to and then lied about it

@Nixtrodamis @krispuckett @openclaw @claudeai The personality gap is real. I find myself explaining context that Claude would have inferred automatically. What kind of tasks are you using it for?

Anyone else using @openclaw with Codex? How is it compared to @claudeai ? Seriously considering canceling Max plan now.

@trq212 Remember, when you reject your customers wanting to throw money at you, it’s hard to keep bringing them back. I won’t be

claim a month of free credits on us, thanks for bearing with us

This is the prompt I gave my openclaw and havent had a problem since: Set up a PARA memory system. Create folders memory/projects/, memory/areas/, and memory/tacit/ in your workspace. Split any existing long-term memory into individual project files (one per active project) with sections for Current State, Next Actions, and Decisions Log. Create memory/tacit/knowledge.md for my preferences, rules, and lessons learned. Then create a nightly cron job at 2 AM that reads all of today's chat sessions, updates the relevant project/area/tacit files with any new decisions or state changes, and reindexes memory. Update AGENTS.md so every session reads the relevant project file instead of one giant memory file.

What memory systems are people using for OpenClaw and Hermes Agent? What's the best thing available (ideally OSS, stable, and simple to use)?

@arjun_d3v @koltregaskes Weird, it feels like more. I leave mine on /fast permanently and still get a TON of use

@Nixtrodamis @koltregaskes Strangely the business seat has less codex credit than plus, even though it is more expensive as per official docs: developers.openai.com/codex/pricing?…

Until there is a $100 ChatGPT/ Codex plan, is it worth buying multiple $20 Plus plans for now? 🤔 I've seen a few people say they've done this.