Stop babysitting self-hosters who know exactly what they're doing @openclaw @steipete
RANT
I'm really upset, because THIS blocks my workflows and wastes my time. What is it I'm asking for? I'll tell you right now: with every new OpenClaw version, agents can no longer change system stuff, they can't change configs, can't even do a gateway restart. Changing skills? Only allowed to make a proposal for the holy skill workshop, can't edit a skill file directly anymore?!? Holy christ...
Constant babysitting over stuff like API keys (you can't post API keys here), session IDs from projects I do for fun can't be posted on Telegram because it's sensitive data... hey... let me tell you what's actually sensitive. It used to be so nice with OpenClaw this way: I could just say "no boundaries" and you'd have full access to the root system and could do everything I want you to do. I'm running it on a separate VPS server, OpenClaw is running as root, yes you read correctly, ROOT. I don't store anything private there, I know my API keys, I'm not some freaking noob who has no clue about exposing API keys, session IDs, passwords, whatever, but I'm getting babysat more and more and can't do shit. Last time I was on the road and wanted to change an API key I badly needed, but I COULDN'T DO IT, because it doesn't accept it over Telegram, even though I allowed it. I don't know, is that coming from OpenClaw or from GPT? I'm using GPT 5.6 Sol as my main driver on OpenClaw, but as he says, these are all system boundaries from OpenClaw. And the best part today, after the latest update: the agent can't change the OpenClaw config anymore. I needed to change the embeddings provider... he can't do it. I have to do it myself through the WebUI, which I hate. Why do I have the most intelligent agent in the world on my side if he can't do the simplest things, like changing a config file? It really pisses me off. I'm running a nano bot instance on the side on the same server, and now I have to pass stuff over to that agent, because he does what I want. If I say format the whole freaking hard drive, he'd do it, because I said so. I'm totally aware that normal users and other people need boundaries and need to be protected, but OpenClaw really needs settings like a YOLO mode where you can free it from every boundary.
1
19
I built a Google Maps-style universe explorer with GPT-5.6 and Codex.
It uses real astronomical distances and logarithmic zoom to travel from Earth to planets, stars, and black holes across cosmic scales.
Try it: thegreatoutthere.com/
Your OpenClaw agent keeps stalling, forgetting to answer, or just never reporting back?
A lot of “agent unreliability” is not intelligence failure.
It’s loop failure.
If a task takes longer, the agent drifts, gets stuck quietly or stops pushing itself forward.
We fixed that with a simple wake loop 👇
1
1
30
The trick:
Send the agent the same follow-up instruction on an interval until the task is done.
Something like:
“Check the current task status. If it’s not done, continue working on exactly that task. Don’t switch to something else.”
That alone makes agents much more reliable.
But how can you accomplish this without manually sending it to your agent? I did that for a long time, and it drove me crazy. 👇
1
1
43
I built it with Telegram + Telethon + a tiny wrapper around prompt files + recurring jobs.
Example:
./wake-agent.sh TARS stay-on-current-task 20 10
Meaning:
• wake the agent TARS every 20 min
• max 10 times
• re-send the exact “stay on current task” prompt
Small hack, big effect. My agent now works the whole night without stopping.
If you need a complete how-to for your agent let me know.
56
🚀 Just launched LystBot! Finally, a list app with a real API for AI agents.
Your OpenClaw Lobster, Claude Companion, or any other AI assistant can manage your grocery lists, todos, packing lists - through a real API / CLI / MCP.
Powered by your AI. Beautiful on your phone.
Open source CLI + MCP server. Free on iOS & Android.
github.com/TourAround/LystBo…
#buildinpublic #indiehacker
1
1
53
Alex retweeted
Found a great deal on CheapCharts but can’t find it on Apple TV? Read this to get it anyway:
cheapcharts.com/blog/cant-fi…
Alex retweeted
🎬 New feature on CheapCharts.com:
You can now import your Letterboxd Watchlist into your CheapCharts Wishlist!
Track what you want to watch and get it when it’s on sale.
Available now on our website: Wishlist > Import > Letterboxd
How do you like it?
Alex retweeted
🚀 We're working on a new Purchase Statistics feature for CheapCharts!
Check out the early mockups & share your thoughts on Reddit 👇 (or here on X!)
reddit.com/r/CheapChartsDeal…
Alex retweeted
CheapCharts has all the prices and we can do more with them -> Price Insights. Let's talk about them, need your input: reddit.com/r/CheapChartsDeal…
Alex retweeted
If you are looking for a complete list of all Apple Vision Pro apps, you can now find them on our CheapCharts website: cheapcharts.info/us/itunes/a…
Also new "Vision Pro" filters for other app lists.
Do you like it? 🙂
Alex retweeted
🚨 INSANE deal today
The "BUMBLEBEE + TRANSFORMERS 6-MOVIE COLLECTION" has dropped to an insane price of $9.99, down from $69.99, which works out to $1.67 per movie. All movies are in 4K. This is an absolute no-brainer purchase!
cheapcharts.info/fwd/1JhD
Alex retweeted
Danke @ctmagazin für den Artikel über CheapCharts in der aktuellen C'T (04/2024). Und was wir bestätigen können: mit CheapCharts kauft man mehr Filme, als man eigentlich vor hatte 😃
heise.de/select/ct/2024/4/24…
Alex retweeted
🚀 Elevate your gaming without emptying your wallet! Our latest video dives into how our "CheapCharts Games App" makes it happen for Xbox, PS, Switch & Steam. Find out how to get alerts on game discounts you've been waiting for! 🕹️ Watch & save ➡️ youtube.com/watch?v=v34fG3Y6…





