Andres G. Aragoneses⚡️

@MyCrypto trying to sell the idea of decentralization to the masses is like trying to sell the idea of opensource to business people; it will never work; you cannot sell them a solution without mentioning them the problem. So explain "lack of counterparty risk" and "lack of vendor lock-in"

@SuperJMN opensource can be real work too haha

Fair point! for public repos the "save CI minutes" angle mostly disappears. That was honestly my starting itch (a couple of private repos eating the free tier), and over time I realized it wasn't really the interesting part. What's left, for me at least, is more mundane: hardware I control for things like Android or AOT builds, caches that stay warm between runs (NuGet, workloads, SDKs), and not having to wire self-hosted runners per repo. Plus the build description (deployer.yaml) is the same locally, on a Pi, or on a hosted runner, which I find handy. But yeah, if your builds fit comfortably in hosted runners and you're on public repos, there's not much here you don't already have for free. It's mostly useful if you happen to want a homelab doing real work

I got tired of CI/CD platforms charging by the minute. So I'm building my own fleet in .NET. A plain Raspberry Pi 4 ARM64 worker can now build and publish: Windows .exe, .msix macOS .dmg Linux .deb, .rpm, AppImage Android .apk Your hardware. Your builds. Zero €/min. github.com/SuperJMN/Dotne…

@SuperJMN well, even if I was interested in what you mention now, I think I'm simply not affected by your initial motivation: costly CI/CD minutes? well, only for private repos, but I just mostly public ones myself

Hey Andrés! It's a good point, and one I've been thinking about. Right now Fleet/Deployer doesn't emit any provenance: the coordinator already knows commit SHA → worker → artifact hash (it's all in the job timeline), but nothing gets signed or attested. Adding it is very doable. The architecture actually fits well: workers stay fungible (they just build), and the coordinator can act as the root of trust that emits the attestation after validating what the worker produced. Two realistic levels: 1. GitHub Attestations via gh attestation generate after each release — Sigstore-backed, verifiable with gh attestation verify, almost free to add. 2. SLSA L3 with in-toto provenance signed by the coordinator — works for any artifact (NuGet, APK, etc.), not just GitHub releases, but a real chunk of work (key management, attestation format, verifier). Honest answer: today Fleet is basically a project I use myself, so I haven't built it yet. I'd love to know which one you'd actually need: GitHub Attestations, or full SLSA L3, before I commit to it. If there are a few people who'd use this, I'll prioritize it.

@lucasmeijer I don't use sandbox anymore, but use unix permissions, see: github.com/badlogic/pi-mo…

if I do want to lego my own together, which lego blocks should I look at? what do you like for sandbox? what do you like for agent loops? what do you like for "somehow persist my sandbox state"

Who has a really nice setup like this? - Cloud based agents - Kanban board-esque overview - Full control over agent loop

Wtf? Just use PPQ. - No KYC - Account optional - Pay privately via crypto

@nadanada_me @levelsio In your case you don't need to be an app indeed. And I completely understand the aversion to app review processes

@knocte @levelsio If you build any freedom tool, you’re very limited by App Store policies and can be cancelled at any time. So for us, it wouldn’t work via an app.

Young people (20s) barely use websites, 70% prefer apps, they mostly exist inside apps and games not websites

@nadanada_me @levelsio I think it depends. As for me I hate websites that could be better with apps (e.g. Yadda phone, Nomad list), but I hate apps that shouldn't exist at all (e.g. why the hell do I need an app to charge my car??) or apps that would be better as websites (e.g. news, X itself, etc)

@levelsio We hate apps.

@cqwww @BobSummerwill @levelsio @onlinedopamine Do you secure it in any way? Like containers, sandboxing or hosting it in VPS?

@BobSummerwill @levelsio @onlinedopamine I just dropped the GitHub repo for Hermes into pi, asked what functionality from it would be useful for my workflow, and it integrated it for me. I've never seen the Hermes tui or run the CLI myself. I just live in pi, prompting my way into the future 🤓

what model are you guys currently using with openclaw? i switched to codex after anthropic cut of access it's been absolutely horrible so far all of my cron jobs broke, tool calling seems non existent, and it's horrendous at following instructions PLS SEND HELP

@gniting @steipete @arunoda yeah pls 🙏😅

@steipete @arunoda “Heartbeat you can disable”? Explain pls?

OpenAI is not in position to burn cash for OpenClaw usage. They will do same what Google and Anthropic did, sooner or later. Fix the tool :)

@bu7emba @drorsos @fredchuuu @steipete nevermind, no one seems to understand the term 'sic' these days sigh

@knocte @drorsos @fredchuuu @steipete Yes and even if its value is empty string CC apparently still tries to use it!

Anthropic now blocks first-party harness use too 👀 claude -p --append-system-prompt 'A personal assistant running inside OpenClaw.' 'is clawd here?' → 400 Third-party apps now draw from your extra usage, not your plan limits. So yeah: bring your own coin 🪙🦞

@bu7emba @drorsos @fredchuuu @steipete You had an 'ANYHROPIC_API_KEY' (sic) env var?

@drorsos @fredchuuu @steipete Yes it was this one! Even if the value is empty string it still picks up. Unfortunately the error message wasn't helpful.

@WWOCrypto @SimplyBitcoin exactly, I was looking for this answer; Samson just lost all credibility in my view

@SimplyBitcoin Banks, company systems are private, servers doesnt even have internet. You can not crack a system that is unreachable, or they can block you after couple of wrong tries. But in bitcoin everything is public. They can try unlimited times, essentially how btc works

Worried about quantum computing breaking Bitcoin? Samson Mow says worry about your bank, the military, and nuclear launch codes first. The media forgot to mention that part.

@nopara73 nah, April's fools I bet

Found an annoying bug in Calendly, that when I moved an entry to the last place, it bounced back. Then I thought I asked Claude in the browser to move it. It encountered the same bug, then did some JS magic, fixed the website code and made it work🤯

@steipete @BentoBoiNFT you sure? I was thinking delegating to clankers the backporting of commits and resolution of conflicts would make it bareable hehe

@knocte @BentoBoiNFT the amount of extra work this would take makes it unfeasible.

OpenClaw is amazing as a concept but I am getting genuinely sick and tired of at least one thing breaking every single day

@steipete @BentoBoiNFT would it help to have a different release management strategy? e.g. after version X released, only allow bugfixes for X.Y versions (with a stable_X branch) and only allow new features in main branch

@BentoBoiNFT Wanna help or just complain?

@steipete @alexgrama plugins or extensions? or is it the same thing? (in browsers it's not)

@alexgrama Working on plugins, and it’s a larger refactor. Almost there.

What’s happening? @steipete

@nadanada_me oh, and "nada nada" can also mean "swim swim" (imperative)

@nadanada_me although the correct expression in Spanish would be "Nada de nada". Because "nada nada" just means "nothing nothing" and it sounds weird, sorry, haha

📢 LNVPN is now nadanada.me. Same services. Same privacy. Same team. New name, because we're more than a VPN now. "Nada nada" = Spanish for "nothing at all." That's what we know about you. 🤷 ⚡ Lightning still gets 5% off and always will.

@MattAhlborg agreed, and btw, talking about development workflows, AutoClaw is not just useful for OpenClaw, but also for coding agents, see awto-pi-lot: github.com/badlogic/pi-mo…

Every hour you invest in improving your AI development workflow, is ten hours paid to you in return down the line. Do it!

@kyle_mccleary @kalomaze sorry I'm very new to harnesses, my point is, if I do `git clone foo && cd foo` and then run your harness from that dir, it's 100% safe that the AI won't touch any other files than within that dir, right? or do I need to opt-in some setting first

@knocte @kalomaze Yes, but you can have it in a real workspace just fine, just like codex. You can use template images with gvisor, firecracker, or docker as the sandbox engine. There are merits for each, e.g. theorem proving needs low latency lean sandbox which Firecracker is best suited for.