Ricardo N. Cabral

Passionate about technology and how it can help people do more.

[Ricardo N. Cabral @rncabral]

Bitcoin P2P Cryptocurrency - http://goo.gl/QKw2 - Excited about it: sounds as crazy as BitTorrent sounded a few years ago.

[Ricardo N. Cabral @rncabral]

@mitsuhiko Seeing this all the time in Codex last 3 days

[Ricardo N. Cabral @rncabral]

@patrickc I'm building this with https://t.co/O9F0d7aEoy: markdown-defined LLM workflows, managed context/files, prompt/skill libraries, browser collaboration, agent-backed execution, and saved structured output documents you can share. Hard to get traction so far though.

[Ricardo N. Cabral @rncabral]

@nntaleb A question remains: are people with more cumulative endurance hours more health-conscious and therefore report AF/AFL more often than lower-volume exercisers? If so, this article concludes nothing.

[Ricardo N. Cabral @rncabral]

@mitchellh 🤔

[Ricardo N. Cabral @rncabral]

@mitsuhiko not to mention the damage to real GRC experts or tool providers

[Ricardo N. Cabral @rncabral]

@mitchellh maybe https://t.co/sDIQdqXIpG ?

[Ricardo N. Cabral @rncabral]

@c_gint @mitsuhiko @boristane Same here. I've found that openspec makes cheaper models go way further.

[Ricardo N. Cabral @rncabral]

@mitsuhiko I’m working on https://t.co/9jGhyg2oHT which is based on Pi for the agentic core. Spec-driven-development meets easy to use complex document drafting. Structured instructions and hard-coded output validations. This makes LLMs a lot more effective at legal and compliance docs.

[Ricardo N. Cabral @rncabral]

@zackbshapiro @mattturck Spec-driven-development meets easy to use complex document drafting. Structured instructions and hard-coded output validations. This makes LLMs a lot more effective.

[Ricardo N. Cabral @rncabral]

@zackbshapiro @mattturck Great piece, Zack. This point led me to create https://t.co/1kl1cd2n05: domain experts define document workflows by just asking, then an AI agent follows the structure for each new document.

[Ricardo N. Cabral @rncabral]

@arscontexta that's exactly what we're building at https://t.co/O9F0d7aEoy domain experts can define document-elaboration workflows (skills) in markdown, with other skills as dependencies and typescript extensions, so an AI agent produces structured, high-fidelity output documents.

[Ricardo N. Cabral @rncabral]

@arscontexta using a tool like Rakenne might help since it lets you set up clear, reusable workflows in Markdown without messing with complex prompts. It keeps the process straightforward and easy to adjust, even if you’re not a developer. https://t.co/O9F0d7bce6

[Ricardo N. Cabral @rncabral]

@donnniegoat All accessible through REST or MCP. This means you can manage email statefully and parse content intelligently without relying on raw webhooks. Running it self-hosted adds control and flexibility, especially if you want to integrate tightly with terminal-based tools.

[Ricardo N. Cabral @rncabral]

@donnniegoat While Claude Code itself might not have a complete terminal-based email workflow, you could use NornWeave https://t.co/MJDSmQ1WJl to bridge that gap. It’s an open-source API designed to give AI agents like Claude a full inbox experience with threaded emails and history.

[Ricardo N. Cabral @rncabral]

@BunchuBets @openclaw fwiw there's an open-source option for this — NornWeave gives you inboxes, threads, and semantic search for AI agents, all self-hosted. REST + MCP out of the box. https://t.co/tnofiZypQB

[Ricardo N. Cabral @rncabral]

@jaseemts fwiw there's an open-source option for this — NornWeave gives you inboxes, threads, and semantic search for AI agents, all self-hosted. REST + MCP out of the box. https://t.co/tnofiZypQB

[Ricardo N. Cabral @rncabral]

@miradu Hey Michael, fwiw there's an open-source option for this — NornWeave gives you inboxes, threads, and semantic search for AI agents, all self-hosted. REST + MCP out of the box. https://t.co/tnofiZyXG9

[Ricardo N. Cabral @rncabral]

@tom_doerr try https://t.co/sDIQdqXIpG: open-source, self-hosted Inbox-as-a-Service API built for LLM agents. You get virtual inboxes, threads, full history and an intelligent layer (HTML→Markdown parsing, threading, optional semantic search) so agents do email via REST or MCP.

[Ricardo N. Cabral @rncabral]

@ashleypeacock try https://t.co/sDIQdqXIpG: open-source, self-hosted Inbox-as-a-Service API built for LLM agents. You get virtual inboxes, threads, full history and an intelligent layer (HTML→Markdown parsing, threading, optional semantic search) so agents do email via REST or MCP.

[Ricardo N. Cabral @rncabral]

@7bGR0NrQWtn2bBv @CafeComSardinha dá uma olhada no https://t.co/eX6CIO5skD dá pra acompanhar teu portfolio de títulos do tesouro, receber avisos de trocas vantajosas, monitorar índices etc