I think everybody who has basic understanding of programming and deployment better should stick to some AI coding agent like Claude Code, Codex, OpenCode etc etc.
I don't think I'm missing out by not using OpenClaw & Co.
I suppose it could be a lot more useful if I actually gave it access to any of my personal data (it lives in a heavily resource-limited container), but there's absolutely no way I'm letting that hot mess of a walking, talking CVE anywhere near my data. It's somehow both horribly insecure and extremely prone to locking me out because of several competing security/permission models fighting it out and gridlocking each other.
Code quality and the issue tracker of the repo are a big mess; for example, the local "memory" retrieval functionality is completely broken for some trivial reason that has been reported and auto-closed about five times (automatically, of course).
In summary: Brilliant idea, terrible execution. Can't wait for the first big tech player I trust enough (or at least one that has my data already anyway) to actually make it a product. I'd use it in a heartbeat.
There are only a few primitives:
1. session history
1a. context map + rendered context map (think of a drive partitioning scheme, but for context -- you can specify what goes into each block of context and this gets built before being sent out for inference).
2. agent definition / runtime
3. workflow definition / runtime
4. workflow history
5. runtime history (for all the stuff session and workflow history fail to capture because they are at a lower level in the stack)
That's it. Everything else builds on top of these primitives, including
- memory (a new context block that you add to a context map)
- tool usage (which is a set of hooks on inference return and can optionally send the output straight back for inference -- this is a special case inside the inference loop and so just lives there)
- anything to do with agent operating environment (this is an extension of workflows)
- anything to do with governance/provenance/security (this is an extension of either workflows and/or agent operating environment... I haven't nailed this down yet).
I suppose I should say something about how agents and workflows work together. I've broken up 'what to do' and 'how to think' into the two primitives of 'workflow' and 'agent' respectively. An agent's context map will have a section for system prompt and cognitive prompt, and an agent can 'bind' to a workflow. When bound, the agent has an additional field in their context map that spells out the workflow state the agent is in, the available tools, and state exit criteria. Ideally an agent can bind/unbind from a workflow at will, which means long-running workflows are durable beyond just agent activity. There's some nuance here in how session history from a workflow is stored, and I haven't figured that out yet.
Generally, the idea of a workflow allows you to do things like scheduled tasks, user UI, connectors to a variety of comms interfaces, tasks requiring specific outputs, etc. The primitive lays the foundation for a huge chunk of functionality that openclaw and others expose.
It's been fun reasoning through this, and I'll admit that I've had an awful lot of FOMO in the mean time, as I watch so many other harnesses come online. The majority of them look polished, and are well marketed (as far as AI hype marketing goes). But I've managed to stay the course so far.
I hope you find your ideal fit. These tools have the potential to be very powerful if we can manage to build them well enough.
Nightly, I have OpenClaw pull the latest changes from a private GitHub repo that is my Obsidian notes vault. It then looks to see which new notes have been added and then runs a "create flashcard" skill to extract and author useful flashcards for spaced-repetition practice. I then gave it access to a custom web-based spaced-repetition flash card application that I built a few months ago. It uses an API to insert new cards, check to see when I last reviewed cards and optionally send me a nudge reminder if I haven't reviewed cards in a few days.
It is a nice workflow that has been working well. I go to class, take notes in Obsidian and check in my changes. By the next day, when I open my flashcard app on my phone, I have new cards to review from yesterday's class.
I cringe at my old boss’s handiwork in Claude and power automate sometimes and go “you know I could just do that in a script and a cron job and it would be completely bullet proof, right?”
Then he just shoots back with, “yeah but now I don’t have to ask you.”
This is my kludge, there are many others like it but this one is mine.
But I'm exploring setting up Hermes from scratch so my family can interact with it in a group chat.
I'm running half my company with Nanoclaw. Same idea, and has some benefits, but I live in CC all day so it's marginal (except for the fact that my laptop has to be on)
In a few years, it's become so easy to falsify articles, falsify comments, falsify images, difficult to really even trust responses online anyways. As far back as 2016, Microsoft already had bots deployed online that could respond 96,000 times [1] in 16 hours all over social media. Remember Tay? [1][2]
[1] https://www.theverge.com/2016/3/24/11297050/tay-microsoft-ch...
[2] https://en.wikipedia.org/wiki/Tay_(chatbot)
Even official government responses.
The British Royal family went to falsification immediately. [3] Note child's broken fingers bent sideways (lower left, didn't even get circled)
[3] https://inews.co.uk/news/signs-princess-kate-royal-family-ph...
The White House is posting altered arrest images of people. [4]
[4] https://www.theguardian.com/us-news/2026/jan/22/white-house-...
Can't trust this stuff much anymore. Obvious caveat with this post.
The whole thing is incredibly buggy. The dashboard is horrible, with page after page with similar-looking settings and what feels like hundreds of things I will never use. The categories in the dashbaord are also unintuitive. It's the kind of thing an AI would put together if it got very vague instructions. It doesn't scream quality and thoughtfullness. Not a bit.
IronClaw is much more promising imo. Trying it out right now. Much less issues so far.
It really is eye-opening how powerful it is once you connect your life to it. The biggest improvement is actually the trivial stuff — emailing contractors, accountants, etc. I no longer need to open another app, navigate six separate steps, and organize everything on the fly just to send one message. I stay in one place, organize my thoughts about what I’m trying to do, and the execution happens automatically.
For example, open claw will find the relevant threads, lookup the relevant details via web search, present them to me, give me a draft, I can review it, edit it, and send. 2 minutes instead of 20.
As a software engineer using AI daily, I think this is the real unlock — staying in a single context and not getting lost in the long tail of trivial details that fragment your attention before you ever get to the actual work.
The mental model shift is important too. It’s not that AI does all the work and we end up with agents emailing each other. It’s that AI automates the steps required to achieve your goals, so you can focus on the substance — like actually writing the email instead of navigating to the place where you write the email.
I also have GitHub and Vercel wired up, which means I can jot down an idea for a little productivity app for me and my family and it just appears a few minutes later, deployed and ready to use. That alone has been worth the setup.
No one can tell me a compelling use case. The whole thing seems designed around getting people to burn more tokens.
Exactly. These companies are only hyping openclaw so that we continue to spend hundreds of dollars a day worth of tokens on their infrastructure.
That’s why companies like Nvidia, OpenAI, Anthropic and Google and many others all want you to spend more on tokens on openclaw and they don’t care if it has no use-case.
All I see is this: Almost no-one other than the hosting providers and course sellers are making money on openclaw and its clones but not those who are running openclaw itself.
What a scam.
Schedule jobs on my job management tool (I vibed it) using a custom MCP. I’ll ask it questions like “what jobs are on today” etc. start the job, complete the job etc.
It will watch Gmail using a MCP for work orders from local real estate agents, where it will schedule quote visits.
After the quote visit I add the photos to telegram (the channel I happen to use) where I then ask it to analyse the photos.
Claude in NanoClaw does a good job of figuring out what needs to be done, but it doesn’t always get it quite right, so I use intake-api which is a “session inbox” [1] that generates a form and uploads it to Cloudflare along with the images and puts a link to the form in the chat so I can make adjustments to the annotated images of the property. After I’m happy, I click submit.
I’ll then go back to the telegram channel and let it know I have submitted the form. Claude will then pull the JSON payload back down into the session and integrate it into the quote.
It will ask me questions about what the job will cost etc, and anything else it thinks it needs to know.
It will then generate a full PDF proposal using Latex between 14 and 32 pages long depending on how many photos were taken.
There are sections with terms and conditions as well as a bit of sales guff etc. as well as quite a nice cover page with their contact details and mine etc.
When I’m happy I’ll ask it to create a draft email in Gmail with the attached pdf proposal.
I’ll review it quickly and then send it.
I also have the Xero MCP setup so I can ask it to create invoices and contacts etc.
I do all of this when getting in and out of my truck.
It’s freed up my home life to spend more time with my children and my Mrs.
There is alot of scope for small business owners who need these sorts of agentic assistant tasks and with my Agentic CRM I’m see a glimpse of the future for guys like me I think.
Intake API:
Can't think of much use for it at the moment but I have it just read and summarise my email, calendar events and git repo in a daily briefing format, it only has readonly access to both, as I dont trust it to do stuff for me or on my behalf.
The briefing thing is nice though not super useful.
I could have piped the results of that search into `claude -p` and had it do the analysis "real-time" and only alert me about things I would be interested in. That's closing the loop in a similar way people use OpenClaw for. But I'm just not interested. It adds more failure points and conditions. The automated things should be as simple and predictable as possible. This may change after months or years more of LLM development or even just me refining my working config. But not yet.
SunshineTheCat•1h ago
Oddly enough, rarely, if ever do they say what specific things they're using it for and how it's saving them time.
I remain interested in it, however, I've still awaiting an actual use case that can't be handled by some other tool/service that does it better/faster.
lyfeninja•1h ago
I'm letting it mature a little before dipping my toe in. I've seem some horror stories, like it deleting repos, system files, and whatnot.
SunshineTheCat•59m ago
I think once I see someone post a use case that I could actually see saving me some serious time, I'll take the plunge. Until then, I'll just let people continue to say how great (or terrible) it is.
chrisjj•1h ago
People? Or bots.
gessha•1h ago
esseph•41m ago
pxc•48m ago
What it adds is making this kind of thing easy for normies, even if it's neither the best way to do things nor very difficult for hobbyists to do using existing tech.
Maybe it's the wrong approach, maybe what people really want is more deterministic software that they use agents to help write. But this kind of thing can maybe serve as a prototyping phase for that. Perhaps in the future, people's assistants will offer to "solidify" frequently used workflows into software that minimizes or eliminates the LLM's role. For existing Claude Code users, its like "please just skip to that step! its cheaper and more secure and more reliable". But to many people who are interested in automation, perhaps that seems out of reach as a first step.