Anton Tesluk (@anton_tesluk) header
Anton Tesluk (@anton_tesluk) profile picture
UAE / USA
Joined May 2022
Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)Photo by Anton Tesluk (@anton_tesluk)
Can't stress it enough
152
With Anthropic limitations on OpenClaw, proper automatic testing becomes more important than ever. And that’s a good thing - you are not supposed to use the strongest models for everything. You should select the cheapest/smallest that reaches the acceptable performance threshold
115
Exploring options on how to trace and test my OpenClaw. That's the biggest bottleneck right now, since every time I want to add another capability, I am spending at least 5 back and forth interactions with it to make sure it works. Testing and evals should automate it
78
Internet slowly becomes unusable anymore. Most of the posts are baits, optimized for clicks, reaction and engagement. Most of the time rage-bait. Looks like @nikitabier is right, we don't have much time left before traditional information sources become obsolete
83
Thinking "am I going to regret it when I am 80?" is a useful way for decision making and self reflection
1
1
86
A problem I see on X (and any other social network) - not many reliable sources. Most of the accounts are overdramatizing and act ling as clickbait machines. Would be great to build a tool that would go through an account’s posts and score their accuracy and give reliability score. So you can filter out quality sources
1
1
120
Spent a day researching how different plastic components affect our body. It is way worse than I thought, it's actually really messed up, especially the hormone mimicry. It just ruthlessly shuffles your carefully balanced internal system.
1
1
87
I have "sleepless notes" where I write random thoughts when I cannot fall asleep. Some of the thoughts from there: - AI simulations will be one of the most important technologies - Everything is about trial and error, and simulation is a cheap way of doing it
86
Seems like AI agent security will be a big thing, so much to tighten up and align
56
Setting OpenClaw as a node on a personal Mac is not easy, need to really carefully approach what it can access and what it cannot. Also to give certain files and directory permissions to tighten up the security, as well as creating intermediary scripts
1
66
Tighting up security for OpenClaw, switching it from root to openclaw user is a pain in the ass, had to carefully copy and replace paths in different files
1
1
87
Last 7 days: Dallas -> Dubai -> Abu Dhabi -> Budapest -> Tenerife -> Krakow -> Riga Busy week
1
68
There's a lot of phishing in ClawHub comments lately - base64-encoded payloads disguised as software updates that download and execute remote scripts. Hope ClawHub steps up monitoring for this. If you see a command with base64 | bash in an issue thread, don't run it.
1
54
How I was using Spotify before: search for specific songs, add them to the queue one by one. How I am using it now: texting my OpenClaw in Telegram "Add 10 Kanye or J. Cole songs next to the queue". It's that simple.
102
Clawhub is really helpful for quickly installing skills. For more customization towards my needs I am creating AGENT or SKILL md files (e.g. how exactly I want it to use Google Spaces API to search for a place to eat)
51
What my OpenClaw can do so far: - Search for a cafe to work or a place to eat near a specific location (Google Places API) - Search for different flights (CLI from Kiwi) - Check weather and air quality - Transcribe my audio messages and reply in audio as well (useful when driving)
1
1
73
What name and pfp did you give to your OpenClaw? I will go first:
1
1
78
Imo one of the most valuable skills remains proactively thinking and reflecting, and resisting your brain to move on autopilot. And now it is harder than ever, because we are getting used to outsource thinking and decision making to LLMs and AI agents
2
1
97
I found meditation to be particularly useful for that. Giving your brain time to cool down from the recently absorbed information, and once it’s calm it can reflect properly. Limiting exposure to any content is a must
37
Actually it's pretty cool that I set up different workflows just through the chat, not even ssh-ing to the server. Just telling it what and how to install, and it does it
36