redfloatplane

Yeah. The times I have let claude off the read-only leash, it's gone fine for me too (with stern warnings not to do anything stupid, and a close eye). But that's not really solving the same problem as this project, I guess. From what I can see this is using a safer and more reproducible method (and not k8s native, so it feels a little foreign to me).

Clever solution. I think ops (like this) and observability will be pretty hot markets for a while soon. The code is quite cheap now, but actually running it and keeping it running still requires some amount of background. I've had a number of acquaintances ask me how they can get their vibe coded app available for others to use.

I really like this idea. I do a lot of kubernetes ops with workloads I'm unfamiliar with (and not directly responsible for) and often give claude read access in order to help me debug things, including with things like a grafana skill in order to access the same monitoring tools humans have. It's saved me dozens of hours in the last months - and my job is significantly less frustrating now.

Your method of creating ansible playbooks makes _tons_ of sense for this kind of work. I typically create documentation (with claude) for things after I've worked through them (with claude) but playbooks is a very, very clever move.

I would say something similar but as an auditable, controllable kubernetes operator would be pretty welcome.

Hey, you do you, I’m glad you appreciate my perspective. I wasn’t trying to catch you out but I see how it came across that way - I apologise for my edit, I had hoped the ;) would show that I meant it in jest rather than in meanness but I shouldn’t have added it in the first place.

As I said in my comment, no shade for writing the code with Claude. I do it too, every day.

I wasn’t “irked” by the readme, and I did read it. But it didn’t give me a sense that you had put in “time and effort” because it felt deeply LLM-authored, and my comment was trying to explore that and how it made me feel. I had little meaningful data on whether you put in that effort because the readme - the only thing I could really judge the project by - sounded vibe coded too. And if I can’t tell if there has been care put into something like the readme how can I tell if there’s been care put into any part of the project? If there has and if that matters - say, I put care into this and that’s why I’m doing a show HN about it - then it should be evident and not hidden behind a wall of LLM-speak! Or at least; that’s what I think. As I said in a sibling comment, maybe I’m already a dinosaur and this entire topic won’t matter in a few years anyway.

I think you’re right and it was OpenCode. The semantic collisions are going to becpme more of a problem in the coming Cambrian explosion of software

Wow, thanks for posting that, news to me! In this case I don’t understand why there was a whole brouhaha with OpenClaw and the like - I guess they were invoking it without the official SDK? Because this makes it seem like if you have the sub you can build any agentic thing you like and still use your subscription, as long as you can install and login to Claude code on the machine running it.

I don’t want to come off like I’m shitting on the poster here. I’ve definitely made that kind of careless mistake, probably a dozen times this week. And maybe we’re heading to a future where nobody even reads the readme anymore because they won’t be needed because an agent can just conjure one from the source code at will, so maybe it actually straight up doesn’t matter. I’ve just been thinking about what it means to release software nowadays, and I think the window for releasing software for clout and credit is closing, since creating software basically requires a Claude subscription and an idea now, so fewer people are impressed by the thing simply existing, and the standard of care for a project released for that aim (of clout) needs to be higher than it maybe needed to be in the past. But who knows, I’m probably already a dinosaur in today’s world, and I really don’t mean to shit on the OP - it’s a good idea for a project and it makes a lot of sense for it to exist. I just can’t tell if any actual care has gone into it, and if not, why promote?

I think these days if I’m going to be actively promoting code I’ve created (with Claude, no shade for that), I’ll make sure to write the documentation, or at the very least the readme, by hand. The smell of LLM from the docs of any project puts me off even when I like the idea of the project itself, as in this case. It’s hard to describe why - maybe it feels like if you care enough to promote it, you should care to try and actually communicate, person to person, to the human being promoted at. Dunno, just my 2c and maybe just my own preference. I’d rather read a typo-ridden five line readme explaining the problem the code is there to solve for you and me,the humans, not dozens of lines of perfectly penned marketing with just the right number of emoji. We all know how easy it is to write code these days. Maybe use some of that extra time to communicate with the humans. I dunno.

Edit: I see you, making edits to the readme to make it sound more human-written since I commented ;) https://github.com/gavrielc/nanoclaw/commit/40d41542d2f335a0...

Yes. Factually wrong and also numerically wrong - Clawdbot -> Moltbot -> OpenClaw, changing names twice, not thrice. To shitpost a little - an LLM editor would have caught that for you, Gary.

For a little while recently I did a weekly vibe game jam for myself using TIC-80 and Claude Code and had the best time.

I actually went to college for game development but never really built any games for myself because the time investment was just too large and the payoff too small. I really think that agent-driven development is the way for games, especially small games and to prototype gameplay mechanics. You no longer have to worry about how to factor your codebase when you just want to see if some idea works. This is especially the case when you put yourself in the constraints of a little virtual machine where you don't have to care much about assets, which are now definitely the bottleneck for this kind of thing.

These games are all unfinished, riddled with bugs, and almost none of them are actually fun (I like Traffic and Shapeship the most) but the thing is that I absolutely loved making them in a way that I haven't enjoyed making anything on a computer in a long while. It was so exciting to see doors that I felt were long since closed to me be blown wide open by agentic development.

https://redfloatplane.lol/blog/07-tic80/

https://redfloatplane.lol/arcade/

Since we're all posting about our favourite email provider, Purelymail has been one of my best discoveries of the last year or so. Ten dollars a year (though it's expected that price will go up) for as many mailboxes as you like. There's a webmail too in case you don't want your own IMAP client. I migrated every email I had (except an unmigrateable @gmail.com) to Purelymail over Christmas and I couldn't be happier.

Appreciate that. I skimmed it and put it on my reading list for when I have a little more brainpower. I think it will go quite well with a few related In Our Time episodes. I’ve started with one about Authenticity, Heidegger and St Augustine. If you take the view that high-level LLMs can be seen as a novel kind of being, there are a lot of very interesting thoughts to be had. I’m not saying that’s actually - or genuinely - the case, before people start to flame me. But I do think it’s a fruitful thing to think about.

The wikipedia page Signs of AI Writing is quite a good one: https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing

But it's a game of whackamole really, and already I'm sure I'm reading and engaging with some double-digit percentage of entirely AI-written text without realising it.

Seems like a postprocess step on the initial output would fix that kind of thing - maybe a small 'thinking' step that transforms the initial output to match style.

Fair cop, I completely missed that!!

Perhaps so, but there are only 5 uses of 'authentic' which I feel is almost an exact synonym and a similarly common word - I wouldn't think you need a thesaurus for that one. Another relatively semantically close word, 'honest' shows up 43 times also, but there's an entire section headed 'being honest' so that's pretty fair.

I expect they co-authored the constitution and other prior 'foundational documents' with Claude, so it's probably a chicken-and-egg thing.

The constitution contains 43 instances of the word 'genuine', which is my current favourite marker for telling if text has been written by Claude. To me it seems like Claude has a really hard time _not_ using the g word in any lengthy conversation even if you do all the usual tricks in the prompt - ruling, recommending, threatening, bribing. Claude Code doesn't seem to have the same problem, so I assume the system prompt for Claude also contains the word a couple of times, while Claude Code may not. There's something ironic about the word 'genuine' being the marker for AI-written text...

I used this about a year ago when I went through a short Rails phase. I was a bit surprised not to see more Rails-specific UI libraries considering how batteries-included the rest of the framework is, and at the time I didn't really 'get' tailwind. I'm not in a Rails phase anymore, but nice work on the library!

The repo was created in May 2023, and it seems like the bulk of commits were made in 2024, before vibe coding was really a thing. I think it's pretty harsh to dismiss projects in this manner.

Hmm, that benchmark seems a little flawed (as pointed out in the paper). Seems like it may give easier problems for "low-resource" languages such as Elixir and Racket and so forth since their difficulty filter couldn't solve harder problems in the first place. FTA:

> Section 3.3:

> Besides, since we use the moderately capable DeepSeek-Coder-V2-Lite to filter simple problems, the Pass@1 scores of top models on popular languages are relatively low. However, these models perform significantly better on low-resource languages. This indicates that the performance gap between models of different sizes is more pronounced on low-resource languages, likely because DeepSeek-Coder-V2-Lite struggles to filter out simple problems in these scenarios due to its limited capability in handling low-resource languages.

It's also now a little bit old, as with every AI paper the second they are published, so I'd be curious to see a newer version.

But, I would agree in general that Elixir makes a lot of sense for agent-driven development. Hot code reloading and "let it crash" are useful traits in that regard, I think

Putting aside the execution:

It's interesting to see people creating and 'selling' agent skills. This one asks for donations, but I was expecting to see a stripe link and 'download for 4 dollars, yours forever' (personally I think that would convert better...)

I wonder if there will be full-blown skill marketplaces soon. Would that be a way for some experts to recoup some (presumably very small portion) of the income they might lose due to generative AI market effects?

Mine is https://redfloatplane.lol, I’ve got a blog and a little game arcade :)

I agree!

I tend to think this product is hard for those of us who've been using `claude` for a few months to evaluate. All I have seen and done so far with Cowork are things _I_ would prefer to do with the terminal, but for many people this might be their first taste of actually agentic workflows. Sometimes I wonder if Anthropic sort of regret releasing Claude Code in its 'runs your stuff on your computer' form - it can quite easily serve as so many other products they might have sold us separately instead!

It seems there's at least _some_ mitigation. I did try to have it use its WebFetch tool (and curl) to fetch a few websites I administer and it failed with "Unable to verify if domain is safe to fetch. This may be due to network restrictions or enterprise security policies blocking claude.ai." It seems there's a local proxy and an allowlist - better than nothing I suppose.

Looks to me like it's essentially the same sandbox that runs Claude Code on the Web, but running locally. The allowlist looks like it's the same - mostly just package managers.

I do get a "Setting up Claude's workspace" when opening it for the first time - it appears that this does do some kind of sandboxing (shared directories are mounted in).

Agents for other people, this makes a ton of sense. Probably 30% of the time I use claude code in the terminal it's not actually to write any code.

For instance I use claude code to classify my expenses (given a bank statement CSV) for VAT reporting, and fill in the spreadsheet that my accountant sends me. Or for noting down line items for invoices and then generating those invoices at the end of the month. Or even booking a tennis court at a good time given which ones are available (some of the local ones are north/south facing which is a killer in the evening). All these tasks could be done at least as well outside the terminal, but the actual capability exists - and can only exist - on my computer alone.

I hope this will interact well with CLAUDE.md and .claude/skills and so forth. I have those files and skills scattered all over my filesystem, so I only have to write the background information for things once. I especially like having claude create CLIs and skills to use those CLIs. Now I only need to know what can be done, rather than how to do it - the “how” is now “ask Claude”.

It would be nice to see Cowork support them! (Edit: I see that the article mentions you can use your existing 'connectors' - MCP servers I believe - and that it comes with some skills. I haven't got access yet so I can't say if it can also use my existing skills on my filesystem…)

(Follow-up edit: it seems that while you can mount your whole filesystem and so forth in order to use your local skills, it uses a sandboxed shell, so your local commands (for example, tennis-club-cli) aren't available. It seems like the same environment that runs Claude Code on the Web. This limits the use for the moment, in my opinion. Though it certainly makes it a lot safer...)

> but typically don't use flashcards

Can you elaborate on this? I watch an unhealthy amount of University Challenge and I assumed that the vast majority of contestants would use flash cards as a trivia retention tool. Most people I've met who need to rely on large amounts of accurate but relatively dispersed knowledge (law students, say, or specific historical professions) use flash cards in one way or another. It surprises me greatly that 'professional quizzers' wouldn't. Perhaps _some_ of them wouldn't - I'm sure as with anything there are some who are preternaturally excellent.

I suppose in terms of catastrophe resilience repairability would be important, although how do you repair a broken GPU in any case. Probably cold backup machines is probably the more feasible way to extend lifetimes.

And yeah - I was thinking that actually power efficiency isn’t really a massive deal if you have some kind of thin client setup. The LLM nodes can be at millraces or some other power dense locations, and then the clients are basically 5W displays with an RF transceiver and a keyboard…

An entertaining thought experiment :)

Yes, I didn't do a great job of managing my language in that post (I blame flu-brain). In the case where _someone_ is going to be reading the code I output, I do review it and act more as the pilot-not-flying rather than as a passenger. For personal code (as opposed to code for a client), which is the majority of stuff that I've written since Opus 4.5 released, that's not been the case.

I'll update the post to reflect the reality, thanks for calling it out.

I completely agree with your comment. I think the ability to review code, architecture, abstractions matters more than the actual writing of the code - in fact this has really always been the case, it's just clearer now that everyone has a lackey to do the typing for them.