A CLI-driven VPS provider focused on persistent NixOS and Ubuntu VMs with static IPs, snapshot support, and scaling up to 60 vCPUs. Positions itself as infrastructure you provision declaratively rather than clicking through a cloud console.
Explore the radar’s memory
Every pick we’ve ever shipped, plotted as a living constellation. Picks cluster by topic; threads between clusters reveal cross-topic connections. Zoom in to expand a cluster into its picks — drag, filter, click through to the brief.
// Browse
All picks
Archive results
-
machine0 – Persistent NixOS VMs You Control from the CLI
-
Can Europe train a frontier AI model on the compute it owns?
An open-source project scoping whether the compute Europe actually owns could train a frontier-scale AI model. More research artifact than shipping tool, but relevant for anyone following the sovereignty-compute debate or thinking about distributed training infrastructure.
-
I built 80 mini-games using Fable before it was shut down
A collection of 80 browser-based mini-games generated with an AI game-creation tool before it shut down, now running as a free open beta. More of a curiosity about AI-generated game breadth than a shipping tool.
-
Lathe – Use LLMs to learn a new domain, not skip past it
Generates source-backed, hands-on tutorials for any technical topic and makes you actually type the code rather than copy it. Built on the premise that LLMs are better tutors than autocomplete shortcuts. For a founder or engineer who keeps meaning to properly learn something but keeps skimming instead.
-
Formally verified polygon intersection – Opus 4.8 oneshots, prev failed
An open-source repo containing the first formally verified polygon intersection algorithm, completed with AI agent assistance after previous model versions failed. Mainly interesting as a benchmark data point on where current frontier models land on hard formal-verification tasks.
-
Continue? Y/N: A 60-second game about AI agent permission fatigue
A 60-second browser game that puts the player through a rapid-fire sequence of AI agent permission prompts, satirizing the approval-loop tedium of current agentic systems. A pointed commentary dressed as a game, worth sharing with your team if permission fatigue is a real friction point in your stack.
-
Tiny-vLLM – high performance LLM inference engine in C++ and CUDA
A C++ and CUDA reimplementation of vLLM's inference engine, stripped down for clarity and performance study. Useful if you are learning inference engine internals or need a minimal base to fork, less useful if you need production throughput today.
-
Forge – Guardrails take an 8B model from 53% to 99% on agentic tasks
An open-source reliability layer that wraps local LLM tool-calling with domain-agnostic guardrails — retry nudges, step enforcement, error recovery, and VRAM-aware context management. Claims to lift an 8B model from 53% to 99% on agentic tasks. Worth checking if you are running self-hosted agents and tired of flaky tool-call chains.
-
Agent.email – sign up via curl, claim with a human OTP
Gives AI agents their own email inboxes with a signup flow designed for agents, not humans — authenticate via curl, claim with a human OTP. Useful for anyone building agents that need to send, receive, or act on email without hijacking a human inbox.
-
Id-agent – Token efficient UUID alternative for AI agents
Generates shorter, token-efficient identifiers as a drop-in alternative to UUIDs, specifically for use inside AI agent contexts where every token in a prompt costs money. A small but real optimization for builders running high-volume agentic pipelines.
-
Find the best local LLM for your hardware, ranked by benchmarks
An open-source tool that matches your hardware specs to the best-performing local LLM based on actual benchmark data. Useful for anyone trying to run inference locally without spending an afternoon reading model leaderboards and doing RAM math.
-
Show HN: Rocky – Rust SQL engine with branches, replay, column lineage
A Rust-powered SQL engine that brings Git-style branching, time-travel replay, and column-level lineage to data pipelines, Rocky is aimed at data engineers and SaaS founders who want the safety and observability of enterprise tooling without the enterprise price tag or the bolted-on complexity.
-
Show HN: Drive any macOS app in the background without stealing the cursor
A weekend-built tool that lets AI agents click, type, and fully control any macOS application inside a background virtual display, leaving your real cursor completely free. Developers building agentic workflows, desktop QA bots, or coding assistants will want to see how it sidesteps the biggest headache in GUI automation.