Tech blog

localai

Offline voice assistant, NO dedicated GPU, less 15Gb RAM

Most voice assistants are walkie-talkies. You press, speak, wait. This one keeps the microphone open while it answers, so you can interrupt mid-sentence, change your mind, or correct yourself and get a reaction within one audio buffer. Everything runs locally. No cloud, no accounts, no round-trips, skills

linux Featured

Which is the Best Linux terminal? Alacritty, Kitty, Ghostty, Foot, it's not easy

Every few months someone on Reddit declares that the terminal they use is the fastest one. Then someone else posts a benchmark, an argument starts, and everyone goes back to the terminal they already had installed. I spent the last few days collecting benchmarks and reading those arguments on Reddit

linux Featured

The Linux I always wanted: LabWC with real Wayland blur

I've been after this setup for a long time and it finally works. LabWC, Noctalia shell, and a terminal with actual blur behind it, done through Wayland itself instead of dragging X11 back in. The Linux ecosystem is fantastic. Open source, freedom, modularity. The catch is that sometimes

linux

Panico paura. 79 bug trovati nel kernel Linux dalla AI

Greg Kroah-Hartman · Kernel Recipes 2026 · 29 settembre 2026 · 57 min Video completo — Kernel Recipes In sintesi Anthropic ha annunciato 79 vulnerabilità trovate nel kernel Linux da un suo strumento basato su LLM. Kroah-Hartman, che ha avuto accesso ai dati grezzi, li ha esaminati uno per uno: alla fine

pi agent

Pi Coding Agent and Jev integration use cases

Pi 1.0 hooks the tool_call lifecycle from its extension surface, so a decision model can sit at fixed points in the loop instead of being consulted after something already happened. Jev answers three kinds of question about a piece of state: pick one option from a list, return

llm

Does harness evolution generalize? Notes on Google's RRSI paper

Google's RRSI paper starts from an uncomfortable observation. Harness evolution works a lot like training a model on your eval set: an agent's harness gets edited task by task, the benchmark score climbs, and nobody checks what happens on a benchmark the loop never saw. When

llm

pi vs pi-fff: measuring what the extension actually costs

I have been running @ff-labs/pi-fff in pi for a while, mostly on the strength of the claim that fuzzy search over a pre-built in-process index beats spawning rg and fd on every tool call. That claim is easy to believe and easy to test, so

llm

How to Wire Jev Into Pi coding agent without Losing Your Mind

Jev is everywhere right now. Every timeline I opened last week had it. Half the posts were demos, half were dunking on the hype. And somewhere in the noise, Pi users were quietly building the most useful stuff I've seen. Pi is the minimalist coding agent. Badlogic'

llama

How I got 2.2x more tokens per second from llama.cpp on Intel Arc

I spent a weekend benchmarking llama.cpp on my Xiaomi Book Pro 14. The machine has Intel's Core Ultra X7 358H and Arc B390 integrated graphics, with 30 GiB of unified memory. That last part matters. It means the GPU can address most of system RAM, which changes

intelligenza artificiale

Peter Thiel: l'AI anticristo, il declino dell'Europa e l'America

Peter Thiel è tornato a parlare, e come sempre lo ha fatto in modo disturbante. Il cofondatore di PayPal e Palantir si è seduto con Mathias Döpfner, CEO di Axel Springer, per un'ora e mezza di conversazione in occasione dell'Axel Springer Award 2026. Il risultato è

llm Featured

AssBench is all You need to benchmark LLM Harness Intelligence /s

AssBench is at assbench.com with a simple idea: "One prompt, Multiple models, Multiple dates." Every model gets the same task, produces one HTML file, and you can open the results live and compare two to four side by side. The task is a joke with a point.

linux Featured

Il visualizzatore DEFINITIVO per foto e video su Linux. Carosello

Hai scattato delle foto e dei video dei tuoi ultimi viaggi, ma attualmente non c'è un modo comodo per visualizzare sia foto che video in una cartella. Carosello è un visualizzatore di immagini e video super veloce, scritto in rust e gtk4. Supporta gesture multitouch, pannello a scoomparsa

intelligenza artificiale Featured

Come mai le Big Tech USA vogliono rallentare? Sta per scoppiare la bolla AI ?

Le aziende che costruiscono i modelli AI stanno chiedendo a gran voce regolamentazione. Ma dietro le parole rassicuranti su "sicurezza" e "responsabilità" si nasconde qualcosa di molto più concreto: la paura di cause legali devastanti e la speranza di uno scudo governativo. Il 17 settembre 2026,

linux

4-Finger Swipe to Change Workspace in labwc on Wayland

If you use labwc on a laptop, you probably miss the 4-finger swipe to switch workspaces from GNOME or macOS. labwc 0.20 has no gesture support in rc.xml, and the usual answer for Linux gestures, touchegg, does not run on a pure Wayland session. Here is a

labwc

Fixing Nautilus Cut & Paste in labwc on Wayland

So you're running labwc as your Wayland compositor, and you've picked Nautilus as your file manager. Everything looks great, until you try to move a file. You press Ctrl+X, hop over to the destination folder, hit Ctrl+V... and nothing happens. No error, no feedback,

linux

Vibecoding Isn't a Crime. How to Deal with Flatpak Flathub Refusals and Actually Distribute Your App on Linux

You built something. Maybe you used Cursor, Claude Code, or just had a long conversation with an LLM that ended up generating most of your codebase. Your app works. It solves a problem. You packaged it as a Flatpak because that's what you're supposed to do

coding

Programmare non è più come prima

Ben lo dice chiaro: non gli importa più dei linguaggi di programmazione come un tempo. E non è una frase fatta. Parla per esperienza diretta, dopo anni passati a combattere online per il framework migliore. Fino a poco tempo fa, scegliere se usare React o Svelte, TypeScript o Go, REST

linux

DHH: Omarchy, l'Agent OS che vuole rivoluzionare Linux

Intervista con David Heinemeier Hansson, creatore di Ruby on Rails, sulla sua distribuzione Linux e il futuro dell'informatica agentic. Cos'è Omarchy Omarchy è la nuova distribuzione Linux di DHH. Un sistema operativo Arch-based che integra agenti AI direttamente nell'esperienza desktop. Il primo vero

labwc

Fixing Nautilus Cut & Paste in labwc on Wayland

So you're running labwc as your Wayland compositor, and you've picked Nautilus as your file manager. Everything looks great — until you try to move a file. You press Ctrl+X, hop over to the destination folder, hit Ctrl+V… and nothing happens. No error, no feedback,

benchmark

Panther Lake Qwen3.5-9B on the same Core Ultra X7 358H: Xiaomi Book Pro 14 (CachyOS) versus MSI Prestige 14 Flip

We cloned 2604272-SYST-LLAMAPT89 from OpenBenchmarking.org, an MSI Prestige 14 Flip AI+ run of pts/llama-cpp-2.5.0 on the same Intel Core Ultra X7 358H, and compared it to local Phoronix Test Suite 10.8.6 results on the Xiaomi Book Pro 14. Same Qwen3.

llama

Benchmarking llama.cpp Backends on Intel Panther Lake: Vulkan vs SYCL vs OpenVINO vs CPU

I benchmarked the Qwen3.6-35B-A3B-MTP-UD-Q4_K_M.gguf model (a 35B Mixture-of-Experts model with 3B active parameters, 22 GiB quantized) across all available llama.cpp compute backends on an Intel Panther Lake laptop. The goal: find out which backend delivers the best inference

linux Featured

Xiaomi Book Pro 14 2026, Intel X7 358H + B390 - Linux Driver Fixes

Xiaomi Book Pro 14 2026 works very well with Linux but some tweaks aren't provided by mainstream Linux distros. Date: 2026-09-12 System: CachyOS (Arch-based), kernel 7.1.5-arch1-2 (6.18.48-1-cachyos-lts as fallback) Notes: currently the only Linux kernel where

xiaomi

Xiaomi Book Pro 14 (2026) has an hardware backdoor?

a deep analysis The Claim: Xiaomi's official product page for the Book Pro 14 (2026) states: "关机状态，也能远程唤醒操控电脑" — Even when powered off, you can remotely wake and control the computer. "内置小米网络模组，关机状态下，也能远程唤醒，出门在外没带笔记本，也能从容应对突发状况。" — Built-in Xiaomi network module, can remotely wake even

llm

What Can You Actually Do With a Local LLM?

Everyone talks about ChatGPT, Claude, and Gemini. But a growing number of people are running large language models entirely on their own hardware. No API keys. No subscriptions. No data leaving their machine. The question is simple: what can you realistically do with a local LLM? More than you think.

harness

Best AI Agent Harnesses and How I Use Them

A practical breakdown of the four main AI agent harnesses one developer uses every day for coding, automation, and personal productivity. The Four Harnesses Pi: The Minimalist Open Source Agent Pi is a minimal agent harness. Small codebase, fully open source, MIT licensed. The best thing about it is that