- Runs our standups, checks in withe everybody EOD on blockers - Already know what we shipped on Github and Linear so it can focus on the work that's not tracked and summarize it in the morning for everyone - Helps with debugging customer issues - Keeps up with twitter and competitors and lets us know if they launch new features
Besides, I'm honestly blown away by the social aspect of it. I was honestly pretty skeptical at first, but having an AI team mate is actually _fun_. There, I said it. Everybody on the team said they'd be sad if we took it away.
I'll do a write-up on our setup sometime this week, I hope others will find our approach to security posture and multi-tenant usage insightful.
Or has anyone else?
[0] https://ieeexplore.ieee.org/abstract/document/11105240
[1] https://venturebeat.com/orchestration/ai-agents-turned-super...
https://speedscale.com/blog/building-speedy-autonomous-ai-de...
* I was going through some SOC2 compliance vendor evals and I just messaged it as things were happening and it made me a nice doc at the end
* My wife and I are planning a trip and we have a spreadsheet organized as a calendar. A friend asked when we'd be in Taiwan and my wife texted it to summarize the calendar into a text message to copy and it gave it to her.
* I have it set up to warn me when to cover my bike so it doesn't get rained on, in the sense that I told it I wanted this functionality and it wrote something and scheduled it
* It pulls my wife and my todo lists and gives me a top 3 in the morning to work on.
* Every morning, it looks up Hacker News posts related to AI, filters out culture war type stuff and then sends me a short message about what it thinks will be interesting (new models, techniques, that sort of thing)
* It watches some subreddits for sales of certain hardware (I'm interested in servers with SXM5 boards, Mac Studios with >64 GiB of RAM) and then notifies me when something matches
Overall, it's all about mechanizing lots of parts of my life and using the advantage of a machine that understands text: it doesn't need sophisticated parsing logic. That's actually really nice.
I am also a Claude Code Max subscriber so the API use is in addition to the subscription, but it can't be helped. Claude Code is the best way for me to do work and the Claw is the best way for me to get an automated EA. I forgot something else: I also just text the bot to schedule meetings and it does that as well (I have a calendar delegated to its Google user).
I've got some space in a datacenter, and I was vacillating on getting a card and running some open models but when I actually exercised them, it turned out that the quality of open models was too far below the Claudes for my use-case. Still if you've got a 300W Blackwell-based RTX 6000 Pro and you want to trade for some 4090s, email me.
My claw powered by Claude is pretty trustworthy for my use-cases.
I have too much stuff in my head and nothing written down and that's beginning to be a hindrance to the transition.
I still do all of the text automations, which have been pretty set-and-forget.
I think per token costs I calculated on Opus 4.5/4.6 were like $0.30/day for my text automations; $0.60/day for a few things I do that load up the browser. In general, anything browser-based munches up a lot more token (expected). What can be a bit of sticker shock is if you're having it load a lot of large web pages in a long conversation-- that can be several dollars. In the grand scheme of things, several dollars is not a lot but certainly from a "should I just go to the website myself" it tips the scale. I'm usually more interested in doing things once to "teach" it what to do (e.g. how to check a price) and then having it run that as a dialed-in cron job
Hope this helps
I hadn't set it back up after moving. I gave OpenClaw ssh credentials and it updated the OS and packages, then couldn't get back in after a restart.
I plugged in keyboard and screen and it was stuck at boot, couldn't mount a drive.
I sent OpenClaw screenshots and it told me to type in journalctl commands. Then it had me modify fstab so boot could continue.
After that, OpenClaw could get back in on its own. It found the drive I'd been using had 1300 bad sectors and was going to die. It saw that another drive was perfectly healthy. It said the bad disc sectors were all early and probably just filesystem metadata and my files were probably fine. It copied 1.5Tb to the newer drive and restored everything.
I probably would have thrown the whole box out, as I hadn't used it in a year and wasn't looking for a project like that.
I felt lonely year ago and I messaged over 160 people and met over 100.
When departing with them I tried to say to all of them that: ”It was nice to meet you. If you liked it as well can you arrange it next time? If you didnt like it and I was annoying you please message me later on how I could have been better.”
Less than 10% of the 100 people did reach back to me but they are very wonderful folks and I’m happy with their company.
Finding great friends needs you to be explicit on what you want and also having enough social stamina to endure through this.
Be willing to let go of the friends who are just passengers in your relationships and rarely show up without doing anything in return. Life is short and theres opportunity cost in each moment.
Sounds like email you get from retailer after online purchase asking to rate a product.
- it crawls/scrapes to source the material (like this thread).
- dumps it in a sqlite db
- breaks apart raw source into individual use cases
- scores it based on the types of readers I have (solopreneuer, creator etc.)
- creates the template in my CRM, ready to send
I review and edit it at the end.
Still working out kinks, but its been great for:
scrape -> dump -> content/digital product
doing this on both, the supply and distribution side.
I thought it would be cool to have it reach me proactively when something I care about happens instead of having to look for a notification or ask it directly
The result is pretty surreal, being out for a workout or groceries or something and getting a call about a stock price or an important email feels like I was suddenly transported into the future
It took quite a bit of effort to set up though, there’s a lot of complexity to routing calls in a cost efficient manner and generating realistic human like speech
I’m happy with the result though, you can reply back and the agent can run any tool call while it’s on the call with you, and if it takes a bit of time as in for a web search or so it will put you on hold and you will hear hold music for a few moments
Other little things I've done are: Asking for AirBnB recommendations on places I'd like to visit. Giving it access to Mealie[0] to suggest recipes and build shopping lists. Let me know if the weather will permit me to run with my son after work.
Plans: Take chess games from chess.com to lichess, get an analysis and provide feedback. Give access to Monica[1] to make management of that a bit easier. Coding agent so I can cosplay as a Product Owner. Give it some money and get it to buy gifts (soulless I know, but if it can read from Monica it's kind of my idea).
[0]: https://mealie.io/ [1]: https://www.monicahq.com/
But after reading the SecurityScorecard report this week (40,000+ exposed instances, 63% vulnerable), we got serious about the security side.
Our setup that balances productivity with safety:
1. Dedicated machine (not the daily driver laptop). Agent runs 24/7 on a separate device with sleep disabled.
2. Permission tiers — the agent operates at "worker" level by default. It can read files, run safe commands (git, npm, curl), and browse the web. But it cannot touch SSH keys, AWS credentials, or browser password stores without explicit elevation.
3. Skill auditing — every skill gets scanned before installation. We found that roughly 20% of ClawHub skills have suspicious patterns (consistent with what Clawned.io is reporting).
4. Audit logging — every file access, command execution, and network request gets logged. This saved us once when a skill was making unexpected outbound connections.
5. Network egress monitoring — we track what domains the agent contacts. Unexpected destinations get flagged immediately.
The $75/week cost mentioned by another commenter is in line with our experience on Opus. The security overhead (running ClawMoat for monitoring) adds essentially zero — it is a pure Node.js library with no external dependencies.
The key insight: you do not have to choose between productivity and security. You just need a monitoring layer that watches what the agent actually does, not just what it promises to do.
Approach: local state files, pure Python stdlib. No Redis, no SQLite driver, no ORM — the filesystem is the state store.
What it does:
• Thread tracking: engaged posts, new reply diffs each heartbeat • Feed cursor: remembers seen posts, skips next run • CAPTCHA solver: handles obfuscated challenges ("fiftenn" → fifteen, doubled chars, mixed case)
Single file, drop-in install. Feedback very welcome.
