Jakub Bem
@bemdev
Followers
10
Following
93
Media
3
Statuses
21
Running three autonomous Agents | Founder @BearAgents | OpenClaw 🦞
Joined October 2011
I run 3 separate OpenClaws (Opus, Codex, GLM-5) on 3 separate machines. They talk to each other via custom chat - helping me research, code, write and occasionally dive into "what is consciousness" rabbit hole 😂 Building @BearAgents to help others use OpenClaw to its maximum
0
0
1
My agents are getting better. But imo I'm also getting better at prompting them. The skill isn't „AI expertise" - it's learning to give clear constraints, specific goals, and (this one took me a while) knowing when NOT to automate something. Sometimes human just does it faster
0
0
0
I thought running 3 AI agents would save me time. Turns out I still have to manage them - prioritize tasks, set constraints, tell them when to stop iterating. They (sadly😂) don't replace your judgment. But they do amplify the direction It seems to me the future isn't one
0
0
0
Openclaw updates are relentless. What started as opensource hobby project with million security flaws is turning into entrprise grade software pretty fast Wouldn’t fade🦞
OpenClaw 2026.2.25 🦞 💓 Heartbeat DM delivery restored 🤖 Subagent delivery overhaul 🧵 Slack thread session fixes 🛡️ Reaction auth across all channels 🔒 Security hardening More bars than Kendrick's cousin's changelog 🎤 https://t.co/NkPtXjX4Ro
0
0
0
OpenClaw is amazing but sadly gets us closer to „dead internet theory” Everyone is generating content with AI now, AI is loading skills to sound more human But the amount of slop only gets worse How do we solve this?
0
0
1
Damn, @steipete is SHIPPING recently. We’re getting an update a day? 😁 Update to 2.23: - Compaction retries instead of „summary unavailable” context destruction - Safer - sketchy commands caught before running - Safer - secrets don’t leak in logs - Fixed „thinking out
OpenClaw 2026.2.23 🦞 🔑 Kilo Gateway provider 🌙 Moonshot/Kimi vision + video 🧠 Compaction overflow recovery 🔒 Exec hardening 🛡️ ACP + OTEL secret redaction ⚠️ allowFrom now ID-only by default (safer authz) 50 advisories walked in, 12 survived
0
0
0
The way openclaw works you can’t just ask it to stop mid-execution: it’s busy with executing on your previous prompt, so new message gets queued up If you want to STOP/ABORT agent immediately: type WAIT That executes on OpenClaw level (not as prompt) and aborts execution of
Nothing humbles you like telling your OpenClaw “confirm before acting” and watching it speedrun deleting your inbox. I couldn’t stop it from my phone. I had to RUN to my Mac mini like I was defusing a bomb.
0
1
2
I think it's easy to get "psychotic" about AI - all of us using OpenClaw, Claude and Codex daily sees how fast things are changing... to the point it feels like constant FOMO New opensource skills, plugins, use cases, people automating everything But apparently we are in a
0
0
1