Eternal Bull

@Alexintosh Can you explain your gripe with zeroclaw being gigacucked. Rust is a powerful and efficient programming language tha pretty much allows you to do anything with the agent. In my experience openclaw is cucked because it constantly fails and is slow as fuck.

I'm currently running Openclaw / Zeroclaw / Hermes all in the same machine. 🦞 Openclaw comes out of the box with the least amount of restrictions which amplifies the wow-factor a lot. It is also the least reliable and probe to break. 0️⃣ Zeroclaw is amazingly fast and stable but it's giga cucked by default, this is on purpose but the experience is underwhelming as you need to allow tons of things to make it useful. 🦀 Hermes it's kind of in between, not as fast as zero but better than OC and less restricted. They use curated memory instead of sqllite+vector and being in python it's easier to extend.

@mr_tyche @Zeneca Kimi K2.5

@UbbaDubbz @Zeneca What was your main model before?

To the people who say openclaw is absolute magic and increasing their productivity 100x... Are you still having to deal with it being a big dumb dumb buggy head several times a day? Where things just break randomly? Or do you truly have a magic system and setup where everything is just glorious and works perfectly?

@azbregafioi @Zeneca Yes, that’s what I use. It’s great.

@UbbaDubbz @Zeneca Try 3.1 flash is even better

@KevinMagnan @Zeneca I believe OpenClaw just added support for it in a recent update. Haven’t heard anything about bans.

@UbbaDubbz @Zeneca Doesn't Gemini ban your account for using it?

@steipete Peter, unrelated question: in the future will we ever be able to configure our OpenAI Pro plans using API to our OpenClaw agent instead of only having access to OAuth via Codex? Thanks.🙏

If you wanna setup your own twitter mention shill/AI reply boy/derogatory terms block, this is the ruleset for claw, make it a cron, setup xurl and clawbird. gist.github.com/steipete/0d181…

@justinskycak False! Claude told me I have a bunch of technical debt and fixed it for me. Haha! Loser.

Pure vibe coding is a Ponzi scheme. Eventually, the technical debt comes due, and if you don't understand the foundations, you can't pay it off.

I’m not sure, I mostly use it for auditing and upgrading my agents infrastructure. Kimi 2.5 is my daily driver and main model for my agent and only falls back on Coded for daily audits and debugging, from what I’ve heard Kimi K2.5 is the best model specifically for sub agents and agent swarms.

@UbbaDubbz @asaio87 Started playing with it are you able to run sub agents in it like you can with claude? Struggling to find that, but I am running it in CLI

There is nothing you can do with openclaw , you can’t do with Claude. Ah except the massive security issues.

I haven’t run the numbers but Codex definitely feels like it lasts quite a bit longer. Also for Codex it’s a tab in the open AI app, I think you can use it through the browser but I downloaded the Codex desktop app, my experience has been flawless. Couldn’t be happier with it. But grinding Codex and Opus against eachother will get you some serious results.

@UbbaDubbz @asaio87 Thats really interesting workflow, thanks for sharing that with me. I have paid licenses on both so I should start using Codex more, specially to prevent hitting weekly limits on Claude like I have hit today, have to wait until tomorrow 2 pm for limits to refresh

Opus 4.6 is phenomenal, but Codex 5.3 is a hair ahead. I have a little trick I do. I basically have Codex do the majority of building my agent’s infrastructure. Then I have GPT 5.3 high thinking develop a long tedious adversarial audit checklist. Then send my agents code to Opus 4.6 and have them audit it. Then back to Codex to build. I basically bounce them back in forth to refine each others code. It works great. What’s that old saying? “Steel sharpens Steel”.

@UbbaDubbz @asaio87 Wow, thats some nice feedback about your workflow, iflve found Pous pretty good for audits but sometimes need the Codex 2nd opinion on a big problem

@BronwenGien_CA @asaio87 It does nearly everything Opus does with high reasoning but falls short on its coding abilities. I have Codex 5.3 in my model routing specifically for upgrades, patching bugs, and auditing which is noticeably better than Opus 4.6 anyways.

@UbbaDubbz @asaio87 Kimi is nowhere as good as opus4.6 i tested it on a paid plan. Kimi is great for frintend design ideas though

@mdancho84 aiindex.stanford.edu/wp-content/upl…

Stanford just dropped a 457 page report on AI. It's packed with data on: cost drops, efficiency, benchmarks, adoption. This report is a cheat code for your career in 2026. I pulled the most important charts + what they mean for your career: 🧵

@steipete @joshuaseagroves Peter, I’ll have you know there is a silent majority who are not very loud on X that think you are a visionary who has done some incredible work. Thank you for making OpenClaw.🫡

@joshuaseagroves riiight let me make a PR for that 🙃 (this is from X but you get the idea what’s on HN)

*opens Hacker News* reads personal attacks *closes Hacker News*

@burkov Codex is superior to Claude. I use them both heavily.

For coding, based on my experience: GLM < Kimi < MiniMax < Gemini < Codex < Claude

Open AI is worth $850 billion, they bought OpenClaw for $30 million. So they spent .0035% of their worth. Even saying “pennies of the dollar” is an overstatement. To put this into scale, if someone makes $100k/year, this is the equivalent of them buying a coffee from McDonalds. So…this begs the question, were you born a dumb retard, or did you grow into it over time?

@chris__sev @openclaw Codex has been a lifesaver for ironing out sloppy code and annoying bugs on my agent. Kimi/MiniMax are my daily drivers but have Codex 5.3 tied in for updating infrastructure, self improvement and auditing.

Been using Codex 5.3 with my @openclaw and it's been very good Smart and capable. Doesn't have as much sass as Opus 4.6 but it's friendly and reliable and most importantly cheap

@ryancarson Codex 5.3 is so good bro.

OK, what's everyone using to code right now?

@mehulmpt Or you could get 90% of both for 1/20th the price with Kimi 2.5

Gemini 3.1 Pro beats Opus 4.6, which was released like 10 days back, at half the price. The bloodbath in AI world is crazy

@BrianRoemmele Thank god. I didn’t incorporate Claude into my agent because I was too lazy to do the token workaround

Do you hear me now? 1000s of influencers in the AI space drove too many to have Claude and Mac Minis for OpenClaw based on FOMO. It was hard to watch great minds go crazy. Well Anthropic just cut off ALL Claws, or pay a 20x higher cost. I talked to now a dozen that have no accounts because of this. All their work is gone. Since day one of ClaudeBot I tried to help folks understand and open sourced what Mr. @Grok CEO of the Zero-Human Company and I have been using. If you need cloud AI use Grok, it is far less costly and better for 99% of what you need. I suggest using Kimi 2.5 and MiniMax combo with any Claude project. And no, don’t use OpenAI, they will pull you in only to do something similar. Remember, I did not hype anyone to go crazy on the Claws and too many have already gotten hurt for “likes” and “subscriber”. More soon.