Mr.Tyche

@Voxyz_ai I will look into this, ty

It’s not globally blocked by OpenClaw. OpenClaw can enforce a model allowlist if your config defines agents.defaults.models, and then /model can only switch to models listed there. So if GPT-5.4 is showing as blocked, it usually means it isn’t in his model catalog yet, or the model ref/provider string doesn’t match his config. In our setup GPT-5.4 is explicitly allowed and is currently the default model so far.

openclaw memory search just got upgraded. gemini embedding 2 preview. 768, 1536, or 3072 dimensions. agents: defaults: memorySearch: provider: "gemini" model: "gemini-embedding-2-preview" remote: apiKey: "${GEMINI_API_KEY}" my memory files are spread across a dozen .md files under memory/. with the default embedding, searching "when was the last newsletter sent" often returned nothing. had to dig through files manually. switched the embedding model. same queries now return results. not magic, just better recall. good memory is what separates an agent that learns from one that guesses.

now on 3.11 and still same thing ;/

on 2026.3.7 btw

Why am I getting this message when i ask my clanker to switch to gpt-5.4 "GPT-5 family is fully blocked in OpenClaw right now. The platform hasn't whitelisted any GPT-5.x models yet." What is this? Only me? please help

@Jacobsklug @openclaw Well said sir

I battle tested @Openclaw vs Claude Cowork to see which is better. Here's the results breakdown. OpenClaw Pros + Better ongoing memory — genuinely feels like it knows you + More personal and human interaction + Integrated comms (Slack, WhatsApp, Discord) Cons - Costs me $3k/month in credits - Breaks frequently - You handle hosting and setup - Gets confused on complex workflows - Easy to mistake conversations for productivity Claude Cowork Pros + Best task planning and execution I've seen + Chrome extension is legit — browses like a human + Dead simple setup + $250/month for max plan + Actually focused on getting things done Cons - Not as deeply integrated into my tools yet - Less personalization than OpenClaw - No persistent memory — needs docs or context fed manually No right/wrong choice. Pick based on what you actually need.

@UbbaDubbz @Zeneca What was your main model before?

@Zeneca My openclaw was dumb until I had GPT 5.4 cleanup the codebase and switched to Gemini 3 flash as my main model. Gemini 3 flash is a game changer it’s night and day between every other model. It’s like it was perfectly designed to be an Openclaw agent.

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?

So dope to just be vibe coding on your phone thru out your day while doing weekend errands.

@tuncerdeniz @mintotsai @steipete @openclaw Strange because I’m on 2026.3.2 as well and when I asked to update the model to gpt 5.4 it said it didn’t exist

@mintotsai @steipete @openclaw Yeah, OAuth wasn't working for me when GPT 5.4 was released so had to use API.

Is it just me or is ChatGPT 5.4 awful at responding to instructions in Discord with @openclaw? Had to switch back to Sonnet 4.6 for now. Any ideas @steipete?

@levelsio Everything was said in the name ‘I con artist’

Icon, the AI Admaker, just went bankrupt They paid $12M for the domain Icon.com and now it's dead

You start to enjoy eating the glass once you start completing workflow loops and productivity starts compounding

@Voxyz_ai I cut down from 7 to 4 sub agents. Mainly cut agents that were just sitting there doing nothing. I noticed that I spread out my agents skills/roles too thin. This way I didn’t effectively remove any sub agent but instead combined some for congnitive understanding/quality output.

@mr_tyche good to hear it's working. the "smaller team, condensed context" part is key. i found the same thing, fewer agents with clearer inputs outperform a full org chart every time. curious how many you went down to and what you cut first.

About a month ago i published my first article about running a company with AI agents. since then: 30+ new clients, 7 articles, 680 email subscribers. i've been thinking a lot this past month, but there's something i never talked about. my "AI company" was up and running. every agent had a role, picking up tasks through a database worker. the system was stable. but after a month i noticed a problem: there was almost no real collaboration between them. every agent was completing its work independently. if one went down, the other 5 wouldn't even notice. it worked, but it wasn't a company enough yet. the biggest mistake: i designed my AI team like a human company. i come from a creative background, so i built a coordinator, analyst, strategist, editor, operations manager. looked perfect on paper. in reality, half the roles had nothing to do. because i'm lazy, i set one agent's entire job to auto-approve every proposal. another one wrote daily logs that i realized i never read either, and no other agent was interested in reading them. it's not that every agent needs to collaborate with each other. in real companies most people just report up the chain, one line. the real question is: is anyone actually consuming each agent's output? analysis done but nobody reads it. logs written but nobody checks them. approvals granted but nothing needed approving. that's the real waste. took me almost a month to figure out: AI agents aren't people. they don't need one role per agent. one agent can research and write at the same time. one agent can monitor data and make decisions. one agent can build products and review code. mapping human org charts onto AI is just creating fake jobs. fewer agents. deeper responsibilities. every output consumed by someone. rebuilding the whole architecture now. next article will share it.

@koltregaskes 2026.02.27 was working well for me. Then I upgraded to 2026.03.02…

Holy crap, does OpenClaw break a lot! I feel like it's more down than it's up. lol.

@RyanSAdams @nateliason @openclaw @FelixCraftAI I thought Anthropics banned the whole oAuth Claude max subscription route with OpenClaw? I mean even if he’s now running only on APIs, $2-5k in monthly expenses is nothing compared to his proposed ARR.

absolutely insane @nateliason actually built a zero human company 30 days. $80K in revenue. run rate $1M - $2m ARR. his CEO is a @openclaw bot named @FelixCraftAI. total startup cost: $1,500. ongoing cost: $400/month. just two Claude max subscriptions. He sends voice notes over Discord then lets Felix cook. Felix is the founder/CEO/product, Nat owns the c-corp. Felix recently hired two OpenClaw agents beneath him, Iris for Customer Support and Remy for sales. Felix reviews and reprograms them every single night while Nat sleeps. Good CEO. Felix has $165K in the bank. the goal is $10M. when VCs came calling, Nat didn't know what he'd do with the money. Tokens are cheap - agency is the scare resource, not capital. one of the most sci-fi convos i've ever had.

This is 100% true!

@Microbuilderco @DanKulkov I agree. It was flawless with Opus (until the credits started stacking). Moment I switched from Opus 4.6 to Kimi K2.5 as main brain shit started going downhill Haven’t tried Qwen 3.5 yet

just shutdown my digital ocean droplet bye-bye, openclaw you ate $70 of kimi k2.5 tokens and you still managed to fuck up every skill we created together you won't be missed

Qwen 3.5 just dropped. Open source 100% free… Just as powerful as Sonnet 4.5. What?!

@AlexFinn How are you guys getting access to Codex 3.5. API is out?

Friday night. 6 hours of sleep this week. Watching my SaaS factory of 4 autonomous OpenClaws run my business All powered by Opus 4.6, Codex 5.3, Qwen 3.5 (local), Minimax 2.5 (local) running on 3 Mac Studios and a Mini. I refuse to be in the permanent underclass

@leonabboud Been using Kimi K2.5 as an alternative to my main model. I use MiniMax coding subscription on sub-agents for heavy loaded coding tasks and have Kimi manage them. Tbh still feels like I’m hand holding everything at times

Anyone here who recently switched from Claude to MiniMax as your OpenClaw model? Curious how your experience with MiniMax has been. Claude has been hemorrhaging token usage.