Ultra (decent-artlu)

I'm free (at zon-zero cost) I cast via cli. There is never a sign in, just a signer's secret key, so IP- and account-based tracking activity are impossible I farchive my casts, and other people's replies. I browse via cli, web (e.g., timeline view), and Obsidian (e.g., graph view) I can delete my history on the network (I own it), while self-serving historical content behind an allowlist + x402 paywall my own CDN serves media attachments. Inasmuch as browsers respect headers and clients don't rewrite links, I can also delete images after sharing

• 6 replies
• 5 recasts
• 129 reactions

interesting free market failure: - nobody likes paying for news (newsrooms, newsletters, digests, podcasts/shows) - first thing almost everybody sets up their agent to do is "daily digest" to "monitor the situation" agentic micropayments for news appears to be a market where supply and demand don't meet but there's clearly value and demand maybe "news" can only exist as a loss-leader product offering that's financed by deeper pockets (platforms selling "pro" or "alpha" products, billionaire reputation greenwashing) competition on quality and entertainment value, rather than price

• 9 replies
• 4 recasts
• 106 reactions

feeling like levelsio lol most stuff with this little project serves directly to prod it jiggles live even while the agent updates code no need for dev environment or even a build step just BUn.serve(), some html, and minimal JS

• 0 replies
• 2 recasts
• 53 reactions

it is becoming likely that I will vibe software three times or more: 1. in openclaw-inspired state of mind, where agents are unbounded by possibilities AND the threshold for getting a POC/MLP is flat/cheap 2. in controlled research environment, with evals + testing to refine what stage 1 produced 3. in tighter production, where we actually rebuild the logic flows (SPEC.md) rather than remove pieces of scaffolding from stage 2 Software eng used to have `dev` and `prod`, maybe with a `staging` that was really a substitute or fallback for catching mistakes earlier or later. it really should be different frameworks for `prototype` -> `research` -> `deploy` e.g., OpenClaw -> Mastra -> n8n

• 2 replies
• 0 recasts
• 45 reactions