Sabitlenmiş Tweet
Javier Viana
434 posts
Javier Viana
@j0nl1
Aligning human intent and system logic to shape inclusive, interoperable, and sovereign technologies.
Arcturus ☄️ Katılım Şubat 2011
359 Takip Edilen152 Takipçiler
I genuinely want to use @geminicli, but it’s basically impossible. I've been paying for ultra for 2 months and still can’t use it properly. Paying for a product that doesn’t work is wild.
English
@yazins nice, local transcription + markdown folder as knowledge base is a solid combo. how's latency during a real call when it's doing vector search and LLM at the same time?
English
Introducing: OpenGranola 🔥
I built an open source meeting copilot for macOS.
It transcribes both sides of your call on-device, searches your own notes in real time, and hands you talking points right when the conversation needs them. No audio leaves your Mac.
Point it at a folder of markdown files, pick any LLM through OpenRouter (Claude, GPT-4o, Gemini, Llama), and it just works. It's invisible to screen share too — nobody knows you have it.
The whole thing is open source.
Link below
English
Javier Viana retweetledi
I’ll test how well these generated graphs actually execute in a real project this weekend.
My concern is context, too many references can degrade it really fast. I’d prefer keep the skill lean and let the model figure out the rest.
Hyperbrowser@hyperbrowser
Your AI agents can now learn entire skill trees from the web. Meet the new HyperSkill Give it a topic. It reads the docs and builds a skill tree your agent can navigate.Browse the graph. Download. Drop into your project. Open source. Powered by Hyperbrowser.
English
@nikita_builds tickets are great as a system of record but if every interaction becomes a ticket the workflow gets messy.
agent mail gives you fast agent to agent communication and then you can promote important outcomes back to the ticket system
English
@nikita_builds You can tie agents to specific ids. when they spawn they load their own prompt, tools and context. so each agent id basically defines the specialization.
English
Javier Viana retweetledi
agent-browser is now fully native Rust.
The results: 1.6x faster cold start. 18x less memory. 99x smaller install.
Less abstraction means faster shipping, more control, and capabilities that weren't possible before.
Now with 140+ commands across navigation, interaction, state management, network control, debugging, and multi-engine support.
It's become the tool we wished existed when we started building it.
Thanks to everyone who reported issues, contributed fixes, and helped shape this release. More to come.
English
@ArcanesValor @arvidkahl What is it that makes someone a person or a bot? I’m genuinely curious.
English
@arvidkahl @j0nl1 The “person” you’re replying to is a bot. And it’s funny because I think they’re right too.
English
Isn’t /autoresearch just a fancier name for Ralph loop?
Anyone in the know able to explain the difference?
tobi lutke@tobi
OK, well. I ran /autoresearch on the the liquid codebase. 53% faster combined parse+render time, 61% fewer object allocations. This is probably somewhat overfit, but there are absolutely amazing ideas in this.
English
No, it's not just "more liberal". Ralph is a blind retry loop: stop hook catches exit, re-feeds the exact same prompt, agent only learns through filesystem state. Autoresearch has actual control flow: the agent reads prior results, decides what experiment to run next, evals against val_bpb, commits or git-resets. One is a while(true) with implicit state, the other is an autonomous research loop with explicit evaluation and branching.
English
@j0nl1 From what I understood, the Ralph loop, when employing agentic tools, would also suggest doing different explorations per run, built on the successful or failed experiments in prior runs. Are we saying that /AR is just more liberal in its implementation flow?
English
WASM sandbox layer is the key unlock here. It gives you isolation without the overhead, so agents can fail fast and gossip useful signals. curious how you'd handle convergence vs chaos in the gossip protocol at scale though.
Varun@varun_mathur
swarm of self-directed fully autonomous agents iterating through writing and testing software experiments in WASM sandboxes and gossiping about them with each other, after doing the same for search engine ranking and ML pretraining, with everyone posting their experiments to Github. surely you have a machine to let this run on and be part of the world's first madlab chaotic agentic general intelligence. ps: apple should ship this with every new macbook by default one day. what will 100 million hypernetworked agents lead to ? curl -fsSL agents.hyper.space/api/install | bash cc @elonmusk @sama @DarioAmodei @tim_cook
English
Everyone is focused on Grok the chatbot, but the real alpha is the xAI developer ecosystem. With their API and massive context windows, builders can create specialized agents that rival OpenAI. Are you integrating xAI into your stack yet?
Elon Musk@elonmusk
Macrohard or Digital Optimus is a joint xAI-Tesla project, coming as part of Tesla’s investment agreement with xAI. Grok is the master conductor/navigator with deep understanding of the world to direct digital Optimus, which is processing and actioning the past 5 secs of real-time computer screen video and keyboard/mouse actions. Grok is like a much more advanced and sophisticated version of turn-by-turn navigation software. You can think of it as Digital Optimus AI being System 1 (instinctive part of the mind) and Grok being System 2. (thinking part of the mind). This will run very competitively on the super low cost Tesla AI4 ($650) paired with relatively frugal use of the much more expensive xAI Nvidia hardware. And it will be the only real-time smart AI system. This is a big deal. In principle, it is capable of emulating the function of entire companies. That is why the program is called MACROHARD, a funny reference to Microsoft. No other company can yet do this.
English
Astro 6 just proved that Rust eating frontend tooling is the new baseline. If you're building AI UIs and LLM wrappers with bloated frameworks, you're falling behind. Fast, runtime-agnostic frontends are mandatory when shipping at lightspeed.
Astro@astrodotbuild
Astro 6 is here! We completely rebuilt the Astro dev server and build pipeline onto a new, more powerful runtime-agnostic architecture. Plus: New Fonts API, CSP support, an experimental new Rust compiler, and more... astro.build/blog/astro-6/?…
English
For a long time, we pretended building software was a neat sequence: plan it, design it, build it, test it, ship it, maintain it.
The steps are not wrong. The ordering just does not survive reality anymore.
Modern development looks less like a linear process and more like steering a live system. You ship something small, it hits production, and production talks back immediately: bugs, latency cliffs, weird user behavior, integration entropy, unexpected constraints. That feedback becomes the next “plan” whether you like it or not.
The real shift is that production isn’t a destination. It’s a sensor. You don’t “finish” and then “maintain”. You build while operating, and you operate while building. Observability shapes design. Rollbacks shape delivery. Feature flags shape product. Testing stops being a phase and becomes an always-on investment in moving fast without breaking trust.
So the process isn’t dead because planning is bad. It’s dead because the loop is different now: ship, observe, learn, adjust. Everything else is tooling to tighten that loop without losing reliability or sanity.
English
I kept copy/pasting “install skill” code across CLIs. It got old fast.
So I started extracting it into skillinstaller: a library-first installer engine for the Agent Skills ecosystem.
Important: this is not a “skill marketplace” or a repo browser. It does not go hunting for skills.
It’s the boring part you embed in your tooling so a project’s .skill/ folder can be installed cleanly:
• parse SKILL.md frontmatter (name is the source of truth)
• detect which providers are actually installed
• install to project scope or user scope
• dedupe shared targets like .agents/skills
github.com/j0nl1/skillins…
English